Fintech In The News

NYSE Partners With Securitize to Develop 24/7 Tokenized Securities Platform | WSJ

The New York Stock Exchange signed a memorandum of understanding with Securitize, the BlackRock-backed tokenization firm, to co-develop what NYSE is calling its “Digital Trading Platform” — a blockchain-based venue designed for 24/7 trading and on-chain settlement of tokenized U.S. equities and ETFs.

Under the deal, Securitize becomes NYSE’s first “digital transfer agent,” meaning it will be responsible for minting blockchain-based shares — converting traditional stocks and ETFs into tokens that carry full shareholder rights, including voting power and dividend access. As Securitize CEO Carlos Domingo put it, most tokenized equity products today are functionally derivatives or price trackers that don’t confer actual ownership. NYSE is building native tokenization within a regulated securities framework.

The platform’s features include round-the-clock trading, instant settlement, fractional share purchases, dollar-denominated orders, and stablecoin-based funding. The technical backbone pairs NYSE’s existing Pillar matching engine with blockchain post-trade infrastructure designed to support multiple chains. ICE is simultaneously working with BNY and Citi to enable tokenized deposits across its clearinghouses, so clearing members can manage margin and funding requirements outside traditional banking hours.

This announcement came just days after the SEC approved Nasdaq’s competing tokenized securities framework, and weeks after ICE’s minority investment in crypto exchange OKX at a $25 billion valuation.

Nasdaq Seeks to Build Crypto Into Wall Street’s Market Plumbing | Bloomberg

While NYSE is building a separate blockchain venue from scratch, Nasdaq is taking the opposite approach: embedding crypto infrastructure directly into the existing pipes that Wall Street already uses to manage risk, collateral, and surveillance across traditional asset classes.

Nasdaq has partnered with digital asset technology firm Talos to connect crypto trading and risk-management tools with Nasdaq’s Calypso platform — the same system banks and brokers use to manage collateral and surveillance across stocks and bonds. Through the deal, Talos’ institutional client base (hedge funds, retail brokers, etc.) will gain access to Nasdaq’s traditional risk-management stack for their crypto operations, while Nasdaq gets a pathway to serve the crypto market using infrastructure it already sells to the financial system’s largest participants.

This comes on top of two other major Nasdaq moves this month. First, the SEC approved Nasdaq’s framework to trade certain tokenized stocks and ETFs on blockchain rails alongside traditional shares — a regulatory milestone. Second, Nasdaq tapped crypto exchange Kraken to distribute stock tokens globally, giving tokenized Nasdaq-listed securities a pathway to Kraken’s international user base. Nasdaq also partnered with Talos separately to create a tokenized collateral management solution, addressing one of the key structural barriers that has prevented institutional adoption of tokenized assets.

The philosophical difference between NYSE and Nasdaq is worth noting. Nasdaq is layering tokenization onto existing clearing infrastructure — tokenized shares still trade through brokers, still settle via DTCC, with blockchain used mainly as an alternative record of ownership. This uses blockchain’s efficiency gains while keeping existing intermediaries in place. NYSE is building a parallel blockchain-native venue that could eventually bypass traditional settlement entirely.

This Week’s Media

Both recommended interviews come via Plaid’s co-founders: CEO Zach Perret and former Plaid executive, now founder and CEO of Column, William Hockey. First, Perret on TBPN (link) discussing credit cards and their competitive position vis-a-vis stablecoins for agentic payments:

“For as far as I can see, which in this environment is maybe a year or something like that, I think agents using credit cards is going to be by far the primary thing that happens. The reason for that is merchants accept credit cards. Merchants don’t accept stablecoins, for the most part, for most of the things that a consumer wants to do. Now, over time, certain types of merchants might start to accept stablecoins. The question is: will the companies that care about making credit cards really easy be able to make them sufficiently easy in the amount of time before merchants actually start thinking about accepting stablecoins? And my hunch is, you’ve seen, I think Ramp actually launched an agentic credit card... I know there’s a bunch of other agentic credit cards that are out there, and I can’t imagine that there will not be 5,000 YC companies launching agentic-focused credit cards soon. So I think we will see more and more options pop up. That’s not to say I’m not a believer in stablecoins. I think it’d be super cool, but I just don’t think it’s realistic in a short time frame.”

Second, Hockey’s interview with Patrick O’Shaughnessy on Invest Like the Best (link). The whole interview was great and a recommended listen, but I thought this comment on the US banking and Federal Reserve technology infrastructure, and why money movement constraints are a feature not a bug:

“The other thing I tell people is there’s this narrative like, “Oh, financial services is fundamentally broken.” Our institutions are actually pretty damn good. There’s this narrative sometimes people talk about, it’s like, “Oh, US financial services. It’s built on COBOL and stuff like this.” It’s just not, it’s like a fun talking point.

I rack my own hardware at the federal reserve and all these places. There’s no COBOL in a lot of these places. It’s actually pretty good. I’ll go on the record talking about this. The Fed is a pretty good tech team. Their systems are actually pretty good. You think about the US right now, through the Fed, has a capability to move money and to clear money through all these institutions, 24/7. Faster than stablecoins, faster than crypto. Right now, as we speak. We’ve had that for decades. Systems are very good, extremely reliant. They’re fantastic. I think it’s very challenging for Silicon Valley to build something better.

The problem isn’t in the fundamental infrastructure, it’s in our implementation of it. The reason why community banks can’t send money 24/7 isn’t because the technology doesn’t exist at the Fed. It’s because there are constraints in those business models that make it very challenging and something to do.

Give you an example. If you can send money out of your community bank 24/7 – well, that bank could run on a weekend. I don’t know if you guys have ever like been to a rural community, the community bank with 50 people – you can’t get people to work on the weekends.

If you’re JP Morgan, if you’re Stripe, yeah, you can manage 24/7 liquidity. But if you’re a small community bank, you can’t have that. It’s why I think people conflate “We don’t have something, we don’t have access to something,” with “We don’t have the fundamental ability to do that.” But actually the reason is implementation, not the underlying infrastructure.”

What I Wrote This Week

What CLARITY Means for CRCL/COIN; Tether’s Audit | March 24, 2026

The CLARITY Act’s proposed yield ban would prohibit stablecoins from offering interest “directly or indirectly,” potentially putting COIN’s USDC rewards program in the crosshairs. Banning rewards would be a near-term boost to COIN’s EBITDA — eliminating the $127M in Q4 USDC Rewards would have boosted COIN’s adjusted EBITDA by 23%.

For CRCL, the near-term economics don’t change (it still keeps ~37% of Reserve Income), but mid-term growth could stall if removing yield makes USDC less compelling in developed markets where holding dollars alone isn’t a sufficient value proposition. If the proposed bill stays as is, prohibiting yield sharing removes the most plausible mechanism for getting stablecoins into consumer wallets at scale, likely relegating them to back-end settlement infrastructure rather than a disruptive front-end payment method.

Meanwhile, Tether announced a Big Four audit (FT later reported the auditor is KPMG) — worth watching given CRCL has arguably been gaining international share in part because of Tether’s compliance gaps.

CLARITY Act

Below is a good summary of what’s happening behind the scenes from Crypto in America’s Eleanor Terrett:

According to an internal stakeholder email shared with me, the proposal would prohibit platforms from offering yield “directly or indirectly” for holding a stablecoin or in a manner that resembles a bank deposit. The restriction would apply broadly to digital asset service providers (exchanges, brokers, etc.) and their affiliates to limit workarounds, and would bar anything “economically or functionally equivalent” to interest.

The proposal would also permit activity-based rewards tied to user activity, including loyalty, promotional, or subscription programs, provided they are not deemed economically or functionally equivalent to interest. It would also direct the SEC, CFTC, and US Treasury to jointly define permissible rewards and establish anti-evasion rules within one year.

If passed, prohibiting offering yield “directly or indirectly” would seemingly end the “rewards” workaround CRCL/COIN have been using to pass yield from CRCL → COIN → end users. And barring anything “economically or functionally equivalent” to interest is purposefully vague and broad. The bill delegates enforcement to a joint SEC/CFTC/Treasury rulemaking within 12 months, meaning the actual boundaries won’t be known until well after the bill passes.

Either way, COIN’s current rewards program for purely holding USDC appears to be squarely in the crosshairs. I’m sure COIN is working on a structure that wouldn’t be classified as sharing interest, but that structure is TBD and the likelihood of success is unknown.

In Q4, CRCL generated $770M in revenue, of which $733M (or 95%) was “Reserve Income” — revenue generated from the interest earned on USDC reserves. CRCL shares this yield with partners like COIN and Binance through distribution partnerships. In total, CRCL passed through 63% of its Reserve Income — and 60% of its total revenue — to distribution partners, retaining the remaining 37%. This is essentially CRCL’s gross profit.

COIN captures the lion’s share of the pass-through, receiving ~50% of CRCL’s Reserve Income. COIN then uses some of these fees (which it recognizes as Stablecoin Revenue) to incentivize Coinbase One subscribers, offering 3.5% APY on USDC balances. In Q4, COIN reported that it spent ~35% of its Stablecoin Revenue on what it calls “USDC Rewards” — $127M out of $364M in Stablecoin Revenue.

Immediate Implications

• For CRCL: There’s no change to its current distribution agreements with COIN and others, so its Net Reserve Margin should stay the same.

• For COIN: It will still collect ~50% of CRCL’s Reserve Income as Stablecoin Revenue. However, it will need to rethink how it passes USDC Rewards through to end users and Coinbase One subscribers. If COIN stops rewards altogether, this could increase near-term profitability. Had the $127M in Q4 USDC Rewards gone away entirely, adjusted EBITDA would have increased 23% to $693M. (COIN CEO Brian Armstrong himself has publicly acknowledged that a yield ban would make COIN more profitable short-term)

Mid-Term Implications

• For USDC holders in the US and other developed markets, receiving no rewards from COIN and others makes the stablecoin less valuable, which could slow the growth in stablecoin supply. In emerging markets, holding the dollar still has value without being paid yield. Total stablecoins are up 34% year-over-year but only up ~2% since October. A further slowdown in growth would directly and negatively impact CRCL and COIN revenue.

  

• For COIN, being prohibited from offering USDC yield reduces the core value proposition of Coinbase One. If COIN cannot find a comparable way to incentivize subscribers, it could see slower sub growth and associated trading volume and revenue. The permitted activity-based carveout (payments cashback, trading fee rebates, DeFi lending on Base) offers a path forward, but none of those individually replaces the simplicity of “hold USDC, earn 3.5%.”

• This also calls into question how the terms of future distribution agreements might be affected. If COIN can no longer pay USDC rewards, is CRCL willing to give up ~50% of its Reserve Income? That will seemingly depend on whether COIN can find a workaround to pay "rewards" within the confines of the law.

Long-Term Implications: US Payments Disruption Case More Challenging

The more interesting longer-term question is how this impacts the case for stablecoins to disrupt traditional card-based payments. I've been skeptical stablecoins would disrupt Visa or Mastercard anytime soon; however, one potential disruption scenario would be to use rewards to encourage consumer stablecoin adoption and essentially create a better debit alternative by passing at least a portion of reserve yield through to the consumer via transaction-based rewards (similar to cashback).

Stablecoins are at a competitive disadvantage versus credit cards given the same interchange-based model those in favor of stablecoins deride provides variable unit economics (i.e., the 2-3% per transaction) to fund rewards and issue credit. However, this is not the case for debit — largely due to the Durbin Amendment capping interchange for large issuers at $0.21 + 5 basis points of the purchase value. This could in theory allow a stablecoin-based incentive mechanism that is unlikely to match traditional credit cards but would be an improvement over debit.

The challenge is that even this path requires using balance-based revenue (reserve income) to fund activity-based rewards — which may still fall in the gray zone of the CLARITY Act's "economic equivalence" standard. And prohibiting yield sharing more broadly removes the most plausible mechanism for getting stablecoins into consumer wallets at the scale needed to make any of this work. Without widespread adoption, the front-end disruption case likely stalls. What does seem likely is that for consumer-based payments in the US, stablecoins will be relegated to better back-end settlement infrastructure — a role Visa, Mastercard, and Stripe are already co-opting — rather than a disruptive front-end payment method.

On Tether

Separately, Tether announced it is undergoing a Big Four audit (though it didn’t name the auditor). I think CRCL’s selloff is at least in part due to this news and the expectation that it could signal increased competition from Tether’s USDT. This is less about US competitive dynamics and more about international ones.

Tether’s GENIUS Act-compliant stablecoin, USAT, has gained little traction since launching in January. USAT’s market cap is currently ~$20M vs. USDT’s $184B and USDC’s $81B. USDT is ~64% of total global stablecoin supply; USDC is ~28%.

Tether’s issues with US regulators are well documented — the $41M CFTC fine, the $18.5M NYAG settlement, and the reported DOJ probe into sanctions and AML violations. But Tether, for all its dominance globally, has run into obstacles with foreign regulators as well. Tether did not pursue MiCA compliance in the EU, resulting in USDT being delisted from every major European exchange — Binance, Kraken, Crypto.com, Bitstamp, OKX — between January and March 2025. CRCL, by contrast, is fully MiCA-compliant through its French EMI license. Same story in Japan, where USDC was the first foreign stablecoin permitted by the FSA.

An audit is a long way from regulatory compliance. But if it signals that Tether is moving in that direction, it at least raises the question of increased competition with CRCL in the regulated international markets where USDC has been gaining share precisely because of Tether’s compliance gaps.

Deal Room

Mastercard Acquires BVNK for Up to $1.8B | Mastercard

Mastercard announced a definitive agreement to acquire BVNK, a London-based stablecoin infrastructure company, for up to $1.8B (including $300M in contingent payments). This is the largest stablecoin acquisition to date, eclipsing Stripe’s $1.1B purchase of Bridge last year. Coinbase nearly bought BVNK for ~$2B before those talks collapsed around November. Mastercard had also been in discussions with Zerohash for $1.5–$2B, and that fell through too.

BVNK processes roughly $30B in annual stablecoin payments across 130+ countries on all major blockchain networks. BVNK reportedly only generates ~$40M in revenue^[1] – valuing the company at 45x revenue. The deal rationale is less about current financials than the technical capabilities BVNK provides, including giving Mastercard the ability to plug 24/7 stablecoin settlement directly into its existing payment rails — think cross-border transfers, B2B payments, and remittances.

As Jorn Lambert, MA’s Chief Product Officer, put it: most financial institutions and fintechs will eventually provide digital currency services, whether via stablecoins or tokenized deposits. Mastercard wants to maintain its current positioning in card-based payments as the orchestration layer, whether fiat or stablecoins. This acquisition provides Mastercard with the technical capabilities to offer fiat-to-stablecoins on/off-ramps. Combine this with the Crypto Partner Program launch the week prior (85+ companies) and it’s clear Mastercard is willing to invest significant capital to position it favorably as stablecoins further integrate into traditional payments.

Kraken IPO Frozen | CoinDesk

Kraken has shelved its multibillion-dollar IPO, just four months after confidentially filing with the SEC and closing an $800M round at a $20B valuation (including $200M from Citadel Securities). The exchange is reportedly waiting for market conditions to stabilize before revisiting a listing. BitGo, the only digital asset company to list this year, has seen its stock drop 44% from its offering price. For context, 2025 saw 11 crypto IPOs raise a combined $14.6B, including several digital asset treasury companies (DATs). The Kraken IPO would’ve been a marquee listing for blockchain financial infrastructure, with the market structure bill nearing passage. We’ll see if this is a short-term delay or a longer shelving. Not everyone is pulling back — Securitize is still pushing ahead with its SPAC, including a $225M PIPE.

Fintech In The News

Tempo Launches Mainnet and Machine Payments Protocol | CoinDesk

Tempo, the payments-focused blockchain incubated by Stripe and Paradigm, launched its mainnet along with the Machine Payments Protocol (MPP), an open standard co-authored with Stripe that enables AI agents to pay for services autonomously. Visa contributed card-based payment specs to the protocol as well. Tempo raised $500M at a $5B valuation in 2025 from investors including Thrive Capital.

MPP is payment-method agnostic — it supports stablecoins on Tempo, Stripe-powered card payments, and other rails. An agent can request a resource, the service responds with a payment request, the agent authorizes from its wallet, settlement is instant, and the resource is delivered. The “sessions” feature allows pre-authorized spending limits with streaming micro-payments batched for efficiency.

The competitive landscape for agentic payments protocols is crowded, Coinbase’s x402, Google releasing its own scheme in September (supporting cards + stablecoins), and the Ethereum Foundation backed ERC-8004. But Tempo/Stripe’s advantage is distribution — Stripe already processes payments for millions of merchants. If MPP becomes the default protocol, Stripe’s install base will play a pivotal role.

OpenAI Rolls Back Instant Checkout | Wired

OpenAI is phasing out its Instant Checkout feature and moving to a merchant-controlled app model within ChatGPT. Walmart revealed that conversion rates for purchases completed directly inside ChatGPT were 3x lower than products that clicked out to Walmart’s own site. Only about 30 Shopify merchants and ~200K Walmart products ever went live. OpenAI hadn’t built systems for sales tax collection or fraud prevention. The word “unsatisfying” was used by a Walmart EVP.

I’m in the camp that OpenAI, Anthropic, and Google will all figure out native-LLM commerce over time — the data advantage and consumer intent signals are too powerful to ignore. But this episode shows the importance of focus. Doing e-commerce well is hard even if you own the top of the funnel. OpenAI learned the same lesson as every marketplace before it: the last mile of commerce is brutally difficult.

Walmart is now embedding its own Sparky chatbot directly into ChatGPT (and soon Gemini), keeping full control of the shopping experience while riding on the distribution of consumer AI apps. Conversion rates for Sparky-through-ChatGPT are reportedly running at ~70% of Walmart.com direct — far better than Instant Checkout’s numbers. Imagine telling someone back when OpenAI announced Instant Checkout that things would be going so poorly that they would end up letting Walmart’s Sparky into their checkout flow.

CFTC Issues No-Action Letter for Phantom | CFTC

The CFTC’s Market Participants Division issued a no-action position to Phantom Technologies, the self-custodial crypto wallet provider. The letter states that MPD will not recommend enforcement action against Phantom for failure to register as an introducing broker, provided Phantom meets certain conditions around its software that facilitates user trading with registered FCMs, introducing brokers, and designated contract markets.

An important clarification: this is a no-action letter, not a formal ruling or regulation. The CFTC is saying it won’t pursue Phantom under current law, not that it’s establishing a new policy. That said, the significance for the self-custodial wallet ecosystem is real. This provides a workable framework for wallet providers to enable derivatives trading access without triggering broker registration requirements — a practical on-ramp for DeFi-to-CeFi interoperability that the industry has been navigating cautiously.

Figure Launches Forge | Figure Chairman Mike Cagney’s X @mcagney

Figure (Nasdaq: FIGR) launched Figure Forge, a tokenization service for real-world assets running on the Provenance blockchain. The service converts illiquid assets — real estate equity, auto loans, equipment financing — into fungible digital tokens that can be used as collateral in decentralized finance (DeFi) protocols. The first application is a partnership with Agora Data to tokenize auto loan portfolios originated through independent car dealers.

Figure has been quietly building one of the more interesting onchain financial infrastructure stacks in the market. Between Figure Connect, FGRD (the first public equity security fully issued, traded, and settled on-chain), the OPEN network, and now Forge, Figure is assembling a vertically integrated blockchain capital markets platform. Forge is the latest piece connecting real-world origination to onchain liquidity.

S&P 500 Goes Onchain via Hyperliquid | S&P Global

S&P Dow Jones Indices licensed the S&P 500 to Trade[XYZ] to launch the first officially approved S&P 500 perpetual futures contract on the Hyperliquid blockchain. Non-U.S. investors can now take leveraged, onchain S&P 500 exposure 24/7 using real-time index data.

This is a landmark for the convergence of traditional finance benchmarks and onchain infrastructure. S&P DJI licensing its flagship index to a decentralized exchange is a strong signal of institutional adoption. The practical implications are continuous price discovery, 24/7 hedging, and global accessibility without the friction of traditional futures exchanges.

Data That Matters

Worth keeping in mind as we enter a more uncertain macro environment: the fundamentals of payments companies tend to be more resilient than investors assume. Payments companies generate most of their revenue from charging a percentage fee on nominal spend, insulating their revenue to some degree from broader macro slowing, especially if nominal and real PCE diverge due to higher inflation. And historically, real PCE tends to hold up relatively well in downturns. Of course, every situation has its nuances and this just focuses on the revenue side of the equation, ignoring the negative impact a downturn would have on valuations. The data below – provided by Morgan Stanley – is meant to illustrate the overall resilience of the payments business model in tough economic times.

^[1] CoinDesk

This Week’s Reads

Stablecoins Are Not Free — Why They Are A RAIL in Consumer Payments | Starpoint by Tom Noyes

Tom Noyes’ blog Starpoint is recommended reading for anyone trying to learn the nuts and bolts of payments. His post today on stablecoins is no exception. Noyes’ view aligns closely with my own on the stablecoin debate. For consumer payments, stablecoins are likely to become another rail and used as plumbing, with the existing four-party model remaining intact with Visa and Mastercard retaining their position as the orchestration layer. Stablecoin adoption as a payment instrument seems more likely to gain traction in other segments — agent-to-agent (A2A), business-to-business (B2B), micropayments, and cross-border. Payments are not a monolith, even if the sector is commonly framed by investors through the consumer-to-business lens. None of this is to say stablecoins won’t see massive adoption — just not in the manner many assume.

“There’s a narrative running through payments circles right now that goes something like this: stablecoins will replace card rails because they’re cheaper, faster, and programmable. Stripe makes acceptance easy. Card networks are too slow to innovate. Machine-Machine payments need programmability. GENIUS Act passed. The future is obvious...I keep coming back to the same conclusion: stablecoins are not a replacement for cards, but rather another rail with cards retaining their role as the layer of abstraction for multiple networks (as they do today). They will do well where cards don’t play (micropayments, B2B and uncarded markets).”

Fintech In The News

Mastercard launches new Crypto Partner Program | Mastercard

Mastercard has launched a formal Crypto Partner Program — a global initiative bringing together more than 85 crypto-native companies, payments providers, and financial institutions to co-develop the next generation of onchain payment products. The program reflects a core belief that the next phase of onchain payments will be built through collaboration. Participants will engage with Mastercard teams on the design and direction of future products and services, including solutions that aim to combine the speed and programmability of digital assets with established card rails and global commerce flows. The partner ecosystem spans every layer of a crypto payment stack — blockchain rails including Solana, Polygon, and Cosmos, and custody infrastructure from Fireblocks, BitGo, and Anchorage Digital. Mastercard’s objective is to embed itself as the authentication, governance, and settlement layer for the onchain commerce era.

Ramp Launches Agent Cards | Ramp

Ramp officially launched its Agent Cards feature on March 11, designed specifically for AI agents that need to make purchases on behalf of users or businesses. The feature gives agents the ability to spend — governed by real spend limits, merchant controls, and full visibility into every transaction — without exposing sensitive card information. At the core of the system is a per-transaction tokenized credential architecture powered by Visa’s Intelligent Commerce protocol and Trusted Agent Protocol, ensuring each credential is temporary and transaction-specific; credentials can be generated via API, MCP, or CLI, making it adaptable for developers at any scale.

Stanley Druckenmiller: “I assume our whole payment systems will be stablecoins in 10 or 15 years.”

In a Morgan Stanley interview released March 13, Duquesne Capital founder Stanley Druckenmiller said blockchain and stablecoins are “incredibly useful in terms of productivity” and “assumes” the entire global payment system will run on stablecoins within 10 to 15 years, calling them more efficient, quicker, and cheaper than existing infrastructure. Druckenmiller critiqued broader crypto as “a solution looking for a problem” — drawing a sharp line between stablecoins and blockchain as infrastructure and crypto as a store-of-value asset.

Data That Matters

Given the topic du jour, I thought it would be valuable to reference data on where the stablecoin industry currently sits. Visa’s Onchain Analytics Dashboard created in partnership with Allium, estimates adjusted stablecoin volume to arrive at a more like-for-like transaction volume figure reported by Visa, Mastercard, and other payments companies. Visa estimates that ‘retail-sized’ transaction volume — its proxy for consumer-based payment volume — was $5.2B monthly, or $62.4B annualized. That represents roughly 0.3% of Visa and Mastercard’s combined payment volume of more than $23T in 2025. McKinsey, in partnership with Artemis Analytics, also published a helpful report on this topic.

Source: Visa Onchain Analytics Dashboard

*From Visa’s Onchain Dashboard: Visa adjusts stablecoin transaction volume. Only categories including centralized exchanges, decentralized exchanges, Lending, Mint/Burn, etc. and addresses, and their associated transactions, are included if the address has not sent more than 1,000 transactions or $10m in transfer volume in a given 30-day period. This removes high-frequency and high-volume trading wallets, high-frequency and high-volume smart contract addresses, bot related activity. Retail Sized are transactions that fall under any of the Adjusted Categories and that are also less than $250.

Deal Room

PayPay IPO Follow-up

Quick follow-up on the PayPay IPO covered in the March 11 edition. PayPay’s stock finished at $18.16 on March 12, up 14% from the $16 IPO pricing. The stock closed Friday at $21.14, up 32% in its first two trading days. Solid investor demand for PayPay in what was a volatile week given the Iran conflict.

The Deal Room

PayPay IPO

Japanese payments company PayPay is expected to IPO today, with Reuters reporting the IPO is expected to price below at $16 per share, below the low-end of the $17-20 per share target range — or a ~$10.7B market capitalization. The SoftBank-backed company was aiming to raise $1.1B in the IPO at the high-end of the valuation, which implied a $13.4B market capitalization.

PayPay will test the waters when most payments companies have had a rough go of it as of late with the Global X Fintech ETF FINX down ~15% year-to-date. PayPay has a unique combination of growth and profitability. In the quarter ending December 2025, PayPay grew gross merchandise volume 22% year-over-year, with revenue growing 24% and adjusted EBITDA growing 63% year-over-year. PayPay has generated material operating leverage over the past several quarters with adjusted EBITDA margins increasing from 7.2% in CY Q4 2023 to 24.3% in the same quarter in 2024 and 31.9% in 2025. Japan’s payment market is largely cash-based, unlike other developed economies, with ~48% of the market still paid for in cash.^[1] PayPay has a dominant position with monthly transacting users accounting for 42% of Japan’s smartphone population and its code-based payment share at 64% in 2024.^[2]

Using rough estimates that assume the current revenue trajectory holds with flat-to-up margins, if PayPay prices at $17 and can grow EBITDA 20–30% in 2026, the stock trades at 14.2–15.4x forward EBITDA. This compares to ~11x 2026 EV/EBITDA for the ~13 fintech companies I follow with expected revenue growth of 20-30% in 2026.

Fintech In The News

ICE Invests in OKX at $25B Valuation | CoinDesk

ICE is making a strategic investment in OKX at a $25B valuation. As part of the investment, ICE will license OKX’s spot crypto prices to launch crypto futures products, while OKX will offer ICE futures and tokenized equities to its customers in the U.S. This is ICE’s second major crypto investment following its $2B investment in Polymarket this past October that valued the blockchain-based prediction market company at a $8B pre-money valuation.

Nasdaq Partners With Kraken in Plan for 24/7 Tokenized Stock Trading | WSJ

Just a few days after the ICE/OKX announcement, Nasdaq announced a partnership with Kraken to power 24/7 tokenized stock trading. As tokenization goes mainstream, legacy financial institutions, including exchanges, are opting to partner rather than build in order to prioritize speed to market. These are starting as minority investments and partnerships, but it seems like the natural end state will be a wave of consolidation as legacy providers try to cement their existing position in the market.

OpenAI Scales Back Shopping Plans for ChatGPT | The Information

OpenAI is positioning itself more as a lead generation, marketplace model than a merchant of record. The early phases of “agentic” commerce looks a lot like the ecommerce consumers are used to with a different demand aggregator at the top of the funnel. As we saw from Google in the past decade, it continued to make a push deeper into the commerce funnel, especially in sectors like travel. Settling as a tax on demand wasn’t for a lack of trying to move into transactional commerce. My guess is this will be similar for agentic commerce: where the respective models start today is unlikely to be where they end up over time (or at least, it won’t be for a lack of trying). I discussed more thoughts on agentic commerce and how it could develop here: The (Ecommerce) Agents Are Coming

Robinhood Ventures Fund I (RVI) Declines 11% in its Trading Debut | CNBC

Robinhood’s Venture Fund I (RVI) declined 11% in its public market debut at the New York Stock Exchange and has traded at these levels since the public listing last Friday. The fund offers exposure to notable private companies such as Revolut and Databricks, but is missing some of the largest private companies, including SpaceX or Stripe (the latter which Robinhood said “is coming soon”). Robinhood’s CEO Vlad Tenev is publicly focused on increasing retail investors’ access to the private markets. I have several questions about this structure, which I outlined in a recent note.

Revolut OCC Bank Charter Filing | PYMNTS

Revolut filed for a national bank charter with the OCC and for deposit insurance with the FDIC, signaling it is no longer content to operate in America through partner arrangements. A successful application would give Revolut direct connections to Fedwire and ACH, and the ability to offer personal loans and credit cards.

What I Wrote This Week Month

This section will typically cover the prior week. For this inaugural edition, I’ve included the past several weeks, which omits articles mentioned above.

Stripe vs. Adyen Comparison (2025) | February 24, 2026

As a public company, Adyen faces the scrutiny of the public markets (including my post-earnings writeup). Stripe has largely avoided this by staying private and offering employees and investors liquidity through annual tender offers. In the past few years, Stripe has avoided the distractions that come with being publicly traded – and has the benefit of seeing its valuation appreciate materially, while other fintech peers see their stocks under pressure.

Adyen and Stripe are a duopoly in cloud-native ecommerce merchant acquiring, with Braintree and Checkout.com distant 3rd and 4th players. Unfortunately, for public markets investors, we only get to compare the financial performance of Adyen and Stripe once a year when Stripe releases its annual letter. And even more unfortunate, is that Stripe only releases a handful of data points like total processing volume and some related revenue or growth metrics. However, this is the closest look we get at comparing the past year’s performance of these payment giants. Here is my take on the results.

Payments: Assessing the AI, Agentic, and Stablecoin Risks for Merchant Acquirers | February 5, 2026

Narratives surrounding AI, agentic commerce, and stablecoins are becoming more prevalent in fintech and payments specifically. The regulatory and compliance aspects of payments makes it more likely that AI is an accretive innovation to existing incumbents versus disruptive and the existing payment four-party model is well-positioned to capture the growth in agentic commerce.

Figure (FIGR), Part 1: Building the Fannie Mae & Freddie Mac of Non-Qualified Lending February 4, 2026

Figure is one of the most fascinating publicly traded fintechs currently. The company has audacious goals to recreate much of how the capital markets works on the blockchain. Figure started with HELOCs but is quickly expanding into other asset classes, including auto loans. It’s not just the promise of big dreams - Figure is executing against this, which shows up in the numbers. In Q4, Figure grew transaction volume 131% year-over-year. The company recently released its February monthly metrics, showing volume grew 127% year-over-year.

My “Part 2” on Figure is long overdue.

Data That Matters

Visa & Mastercard’s US Payment Volume Growth Decomposed: PCE Growth vs. Digital Payments Penetration Gains

The graph below decomposes the total payment volume growth of Visa (V) and Mastercard (MA) into adjusted personal consumption expenditures (PCE) growth and digital payments penetration gains. I adjusted PCE to exclude healthcare, housing & utilities since these are mostly non-addressable for typical consumer card payments. The proxy for digital payments penetration is calculated by subtracting adjusted PCE growth from V and MA total volume growth. Since the two operate a duopoly, end-market growth is adequately captured through adjusted PCE growth.

Notably, in the three years preceding the pandemic (2017-2019), digital payments penetration accounted for 60%+ of V & MA’s US growth. Since 2022, it’s ~20%. At V’s February 2025 investor day, it stated US digital payments penetration was 60%+ — my rough math puts it over 70%. With ~80% of volume growth driven by end-market growth, this increases the inherent cyclicality in their US businesses. I explained this in more detail in a follow-up X post:

If you think about $V and $MA long-term volume growth drivers, it’s been a combination of personal consumption expenditures (PCE) — or the amount consumers are spending on goods and services regardless of what payment instrument they’re using — and digital payments penetration increasing — or the increasing number of consumers using digital/card payments versus cash or check (or a cardholder who already has a card using it for a higher percentage of their purchases over time).

For a long time, $V / $MA volume growth was more evenly split between end market / PCE growth and increasing digital penetration. This allowed them to sustain higher than industry growth (ie PCE) — typically ~2x that rate — and allowed both to maintain a certain level of growth in cyclical downturns (PCE falls but digital penetration gains carry more weight so they still grow) but given growth has converged more to PCE growth and digital penetration is going to be less of a driver moving forward, $V / $MA volume growth will more closely reflect PCE, which means it’ll be more heavily tied to macro economic up/down cycles — and generally, will grow slower going forward assuming PCE remains ~stable.

Notably, this says nothing about the investing merits of V and MA. Both have meaningful international exposure that grows faster than the domestic business – and revenue from value-added services (VAS) is an increasingly important driver of revenue growth (both have done a good job here), but understanding the building blocks of their US growth is an important aspect of thesis.

Block (XYZ) Operating Efficiency

Alongside Q4 earnings, Block announced a layoff affecting 40% of its workforce, with CEO Jack Dorsey citing advancements for the company in leveraging AI internally as a key consideration. Among investors, this sparked a debate about whether AI was the reason or XYZ was an overbloated organization and “AI” was a convenient excuse to reduce its workforce. Dorsey responded to an X post claiming “managerial incompetence” from overhiring during Covid by saying:

yes we over-hired during covid because i incorrectly built 2 separate company structures (square & cash app) rather than 1, which we corrected mid 2024. but this misses all the complexity we took on through lending, banking, and BNPL. and that we’re now targeting $2M+ gross profit per person, 4x our pre-covid efficiency, which stayed flat at ~$500k from 2019 until 2024. we have and do run an efficient company... better than most.

Below is the operating efficiency Dorsey cited by year. Operating efficiency was range-bound from 2019-2022 but is on pace to increase 4.2x since then based on 2026 guidance and double year-over-year. Furthermore, XYZ screens as the most “efficient” fintech in its peer set. I put efficient in quotes, because gross profit per full-time employee is a blunt instrument to gauge operating efficiency. Ultimately, I think XYZ was too bloated and its operating efficiency was insufficient for a business generating over $10 billion in gross profit. It’s also likely seeing some early proof points of efficiency from AI. Lastly, other fintechs are even more bloated, which is likely foreshadowing a focal point for the sector as we move through 2026.

**Caveats are I used $XYZ’s post-RIF headcount; Adyen’s mid-point of their FY guide while the rest I assumed flat headcount. This penalizes Adyen while benefitting XYZ, but the chart is meant to be directionally illustrative, not a precise ranking.

.

^[1] PayPay’s F-1 IPO prospectus.

^[2] PayPay’s F-1 IPO prospectus.

Comparing to Adyen, Stripe grew volumes materially in 2025 (34% YoY) no matter how you look at Adyen’s volume - reported volume growth was 8% YoY or 21% when adjusting for the loss of Cash App business.

2025 was the first year Stripe surpassed Adyen’s volume at ~$1.9T vs. Adyen at ~$1.6T. I converted Adyen’s volume to USD to make it more comparable.

Total processed volume only tells part of the story though considering enterprise volume for Stripe is I believe ~50-60% of their total volume with the remainder SMB. Adyen is essentially all enterprise - meaning their enterprise volume is still likely ~40% larger than Stripe (but these are just my estimates not reported figures).

Both Adyen and Stripe stated the 2025 cohorts were their largest ever. Stripe benefiting disproportionately from new AI company formation: “In 2025, many more new companies joined Stripe than ever before, with more than half of them (57%) based outside the US. This new cohort is by far the highest performing and fastest moving we’ve ever seen, growing around 50% faster than the 2024 cohort. The number of companies reaching $10 million ARR within 3 months of launch was double the 2024 count.”

Net take rates are meaningfully different for Adyen and Stripe. Enterprise net take rates are similar for both (~16-20bps) but Stripe’s blended take rate is likely ~2x higher given their SMB mix (SMB net take rates are likely above 50bps+) putting consolidated take rates at ~35bps for Stripe. Additionally, non-payment revenue is becoming more meaningful for Stripe - the letter cited a target of $1B+ annual run-rate in 2026. Non-payment revenue is likely LDD% of total revenue for Stripe (ADYEN gives little disclosure here).

In total, net revenue for Stripe at ~$6.8B versus Adyen at ~$2.8B. Adyen likely has higher EBITDA margins than Stripe (53% in 2025 vs. unknown for Stripe) but a significantly smaller revenue base.

Stripe tendered shares at $159B this month, up 74% from the ~$91.5B Stripe tendered shares at in February 2025. This compares to Adyen’s current market cap at ~$35B (down ~47% over the past year). Stripe’s 350%+ valuation premium vs Adyen is widest it’s ever been.

One could argue Stripe would be valued higher than Adyen if it were public today (growing faster, more tangible AI benefits to growth) but the gap feels too wide to me. Adyen is expected to grow EBITDA ~20% YoY in 2026 and trades for 12.7x 2026 EBITDA. Let’s assume Stripe grows revenue at ~30% in 2026 and put the EBITDA margin range at 30-50% (wide range I know) - this puts Stripe’s 2026 EBITDA at $2.7-4.4B* or a multiple at 36-59x.**

*This accounts for SBC to try to make like-for-like with Adyen.

**Cash is not reported by Stripe but the company is likely in a net cash position, meaning the actual multiple would be lower than this range. This is meant to be directional.

I previously wrote brief thoughts on agentic commerce (which was more from a payments perspective), concluding that “it seems inevitable that AI agents will make purchases on behalf of consumers in the not-too-distant future.” I wish I’d phrased that differently. While agents completing much of the purchase journey end-to-end is plausible in the near term, the reality will be far more nuanced — both by category and by the specific criteria that define each shopping experience.

How agentic commerce evolves will depend on advancements in LLMs, which parts of the value chain agents control, and how both shape consumer behavior. What I’m confident in is this: agentic commerce is unlikely to evolve in one monolithic way — just as ecommerce hasn’t, and in-person retail before it. The nuance will live in the category. Agentic commerce for groceries will look fundamentally different from booking a hotel or buying clothing. Even within a single platform, categories are treated very differently — consider Amazon’s grocery strategy versus its apparel vertical.

The Current State

In September 2025, OpenAI announced its Agent Commerce Protocol (ACP); Google followed with its Universal Commerce Protocol (UCP) in January. Both attempt to solve the coordination complexity inherent in multi-party commerce. Without universally accepted protocols, every player would need bespoke agreements with each other — an NxN problem that would make agentic commerce dead on arrival. ACP and UCP try to solve this by creating unified standards for all parties to follow.

Neither protocol is guaranteed to succeed. Notably, Amazon is holding out, prioritizing its closed ecosystem and protecting its first-party data and advertising business. The walled-garden versus open-protocol tension we see across the internet today may well persist in an agentic world. But given the scale of both Google and OpenAI, these protocols seem likely to anchor significant portions of how agentic commerce develops.

In both ACP and UCP, the merchant remains the merchant of record — responsible for everything that happens after the consumer initiates a purchase. The architectures differ meaningfully, though. ACP is structured with ChatGPT as the front-end aggregator owning discovery, with OpenAI taking a ~4% transaction fee for that aggregation role.1 Google’s UCP is more open-source in nature, providing a fuller framework across the commerce journey — including post-purchase elements like returns, an area where ACP offers comparatively little guidance.

The current iteration of both shopping experiences feels less like agentic commerce and more like search-led commerce with conversational LLM-based discovery replacing keyword search. The checkout experience remains largely unchanged, payments infrastructure is identical to traditional ecommerce, and the merchant retains responsibility for logistics and post-purchase operations — as the Etsy and Wayfair examples illustrate.

OpenAI’s ChatGPT Instant Checkout Experience with Etsy

Source: OpenAI

Google’s AI Mode Shopping Experience with Wayfair

Source: Google

When most people imagine agentic commerce, they picture something more expansive: a consumer gives an initial set of instructions and an agent handles the entire discovery and transaction process with minimal human interaction, coordinating directly with merchant agents to complete the purchase. Platforms like OpenClaw are providing early signals of what this could look like — both the promise and the risks, with legitimate security concerns already surfacing. But if history is a guide, that hypothesized end-state is likely years away, at least for certain categories. Even if adoption accelerates, the impact on the broader retail and ecommerce landscape will likely be iterative rather than existential for existing players. After two-and-a-half decades, ecommerce is still less than 20% of total retail sales and the most powerful paradigm shift in ecommerce – mobile – accounts for less than 50% of all ecommerce transactions, nearly two decades after the iPhone was launched.2 If agentic follows a similar adoption curve, it becomes another channel merchants must plan for alongside in-store, desktop, and mobile.

There are a few trajectories agentic could take:

• AI reduces friction within the existing checkout flow — auto-filling personal and payment information — showing up as higher conversion rates with no material structural change to the underlying experience.

• LLMs improve the discovery layer without replacing the transaction. Instead of searching separately for hotels and flights, a consumer inputs travel dates and destination, and the LLM assembles a potential itinerary.

• Agents go beyond discovery and handle price comparison and review aggregation, filling the checkout basket while the consumer retains final approval over payment.

• Agents handle the entire end-to-end purchase process with little-to-no human input — the scenario most likely to emerge if personalized agents gain broader traction.

Even short of a fully agentic experience, AI’s impact on ecommerce fundamentals could be substantial. Cart abandonment remains structurally high at ~77% globally.3 The conversion rate for the average ecommerce transaction sits under 3% — with some categories closer to 1% — meaning the overwhelming majority of visitors never complete a purchase.4 Ecommerce return rates run over 19%.5 Better product matching and predictive analytics have clear roles to play in all three even without an agent handling the entire commerce journey.

Agentic Suitability: Shopping Friction

Understanding how far into the purchase funnel agents will go — or more accurately, how far their human will allow them to go — depends in large part on the friction in the shopping process. A useful lens is to evaluate consumer categories across four criteria:

Purchase Frequency measures how often a consumer buys within a given category. Higher frequency gives an agent more opportunities to learn preferences, but it also deepens loyalty to the platform or retailer delivering the product — which can actually reduce reliance on a third-party LLM as the interface.

Average Order Value reflects the average dollar amount per transaction and serves as a reasonable proxy for per-transaction decision stakes and consumer risk tolerance. The higher the AOV, the more likely the consumer wants to be consulted before purchasing.

Physical Value Chain Reliance captures how much of the customer value proposition depends on real-world logistics, discovery, or in-person fulfillment. The more the experience lives in the physical world, the less likely agents are to own the end-to-end process. An LLM can aggregate supply — but can it deliver as quickly, cheaply, and reliably as Amazon or DoorDash?

Decision Specifiability is perhaps the most important and underappreciated criterion: the degree to which a consumer’s purchase preferences can be fully expressed as objective, machine-readable parameters before the transaction occurs. A decision that can be parameterized has a far higher probability of being delegated to an agent than one that relies on aesthetic judgment or subjective evaluation. Reordering the same grocery list weekly sits at one end of this spectrum; selecting a piece of jewelry sits at the other.

*The above examples are meant to be illustrative.

Categories with low purchase frequency and high AOV — a vacation to Thailand, say — present a specific set of dynamics. The infrequency makes it hard for brands or platforms to build loyal repeat relationships, while the stakes make consumers want to shop around, price compare, and retain final approval. These categories are natural candidates to replace traditional search with conversational LLM-based discovery, but they are unlikely to become fully agentic. The discovery layer will probably be agent-led; the final decision, including payment confirmation, will remain human (think of this as “human-aided agentic”).

However, this likely won’t be the case for all travel. The frequent business traveler who books the same hotel every week is highly likely to have their agent — or Hilton’s in-app agent — handle the booking directly, with a card on file and Hilton Honors rewards intact. One scenario involves infrequent, high-stakes friction. The other involves frequency with almost none.

High-frequency, low-AOV categories appear best suited for a fully agentic experience. But these are also the categories where dominant retailers and marketplaces are best positioned to own the customer relationship precisely because of that frequency. If you order groceries from Amazon every week, you’re starting your journey in Amazon — not in ChatGPT or Gemini or Claude. If you order from your favorite local taco spot through DoorDash every Friday, you’re going back to DoorDash. For these categories, Amazon’s Rufus or DoorDash’s native agent is far better positioned to own the interface than a general-purpose LLM. This is agentic commerce — just not the version most people are imagining, and far less disruptive to incumbents than the headlines suggest. Here, LLMs function more as a new customer acquisition channel, structurally analogous to Google search or paid social.

These categories also tend to have material physical-world components that are critical to the customer experience. The degree to which this matters depends on whether the incumbent actually owns that physical experience. DoorDash’s delivery network is a structural differentiator because DoorDash controls it end-to-end. The lodging experience from a hotel booked on Booking.com, by contrast, is non-exclusive to Booking and largely separate in the consumer’s mind — making the physical world less of a structural advantage for Booking, even though the stay itself is central to why the customer booked in the first place.

Media subscriptions — Netflix, Spotify — represent a distinct case worth noting. The subscription model effectively internalizes the discovery problem rather than outsourcing it to an LLM, which insulates these platforms from third-party agent disruption at the top of the funnel. Their opportunity lies in building proprietary in-platform agents that improve content discovery within a walled ecosystem.

Will Agentic Commoditize Marketplaces?

The core value proposition for most marketplaces is reducing search costs through supply aggregation. Given how LLMs work, it’s worth asking whether certain marketplaces become commoditized over time — specifically those whose value is concentrated in discovery rather than physical execution. Travel will likely be one of the first categories where this theory is tested. That’s not coincidental — travel was among the first verticals to move online, precisely because the product is inherently digital at the point of discovery: you research, compare, and book well in advance of “consumption” (i.e., the stay).

Booking.com and Expedia dominate US online travel, including hotels. Their respective moats today are supply-side scale and continuous iteration on the discovery and checkout experience, which yields conversion advantages that they reinvest into supply acquisition and demand generation. Booking.com, in particular, has managed the Google ecosystem better than perhaps any other marketplace and is one of Google’s largest advertisers. But in a world where hoteliers are actively standardizing for agentic connectivity — APIs, structured data, direct feeds — does the value of supply aggregation start to erode? Specifically, if agents are sourcing the best inventory on criteria — pricing, reviews, proximity to destination, etc. — does this neutralize the advantages that have been created through supply aggregation (i.e., historically built through performance marketing and conversion optimization)?

Booking’s CEO Glenn Fogel addressed this directly on a recent earnings call:

“Look, we’ve been competing, so to speak, with a top-of-the-funnel player who’s very big for a very long time and people go there, but yet then they come to us direct many of them. Why? Because we are offering them something of value. That’s what I believe is going to continue. And by the way, it’s been very successful for that large person at the top of the funnel, Google. We’ve both done very well. They’ve made very good money setting up an auction format for taking advertising, and we have been more than happy to pay a lot of money as have competitors of us. It’s worked out really well for them. It’s worked out really well for us. And by the way, some of these other large language models when they’re looking at this, they may say, that’s the right way to do it. Not go down the funnel, don’t become a merchant of record, take it high above, don’t deal with the mess of the day-to-day in and out that we have to deal with all the time, successful for us, successful for them. I think that’s the way it’s probably going to go in the end, but that’s my opinion.”

Fogel’s broader point — that LLMs will likely monetize at the top of the funnel through an advertising model, much as Google has, rather than competing downstream in the complexity of payments and operations — is reasonable and probably correct. Google’s own history with travel OTAs is the proof of concept: it built an enormously profitable business sitting above the transaction, never becoming a merchant of record itself. However, depending on how agents source and rank inventory, it could result in more hotels going direct — cutting out middlemen like Booking.

Wyndham discussed its efforts on its Q4 call this week:

“We continue to work with public LLMs, including Google AI Mode and ChatGPT, to establish direct connections of our hotel data to eliminate the need for these models to scrape our sites. In November, Google selected Wyndham as one of a handful of partners to take part in an agentic booking experience on AI Mode in search. Soon, guests will be able to discover Wyndham properties through natural conversational interactions while our connected systems enable seamless direct bookings within AI Mode. We’ve also successfully connected to Anthropic’s Claude.”

Fogel cited being the merchant of record and handling payments as a key advantage:

“You want to be merchant of record, payments — that’s really complex. We have over 100 different payment methods, more than 50 currencies... regulations on the payment here is tremendous... You want to be involved in the travel business and you want to be dealing with merchant of record... that’s just the EU. And then, of course, the national regulations — we’re dealing with over 200 countries... Do we think that the large language models are going to be entering and want to enter down the funnel down to where we are? I don’t think so.”

I agree with Fogel that LLMs are unlikely to go deep into the funnel and participate after the transaction, but supply aggregation as an advantage is likely to diminish over time, especially as hotels modernize their infrastructure and establish direct feeds to LLMs. Today, Booking’s access to over four million properties — particularly independents running outdated property management systems — is genuinely hard to replicate. The long tail of independent hotels will take years to convert; however, the top six hotel chains already account for nearly 60% of US hotel rooms, and they are moving quickly.6 This doesn’t mean Booking and Expedia disappear. But their value proposition gradually shifts. If their differentiation shifts down the value chain — from discovery and demand generation toward payments infrastructure and customer service — the result is likely take rate compression. The typical marketplace captures 15-25% of the transaction value versus 3-6% for merchant-of-record / payment providers.

This dynamic is likely to play out across categories: agents will go as far into the funnel as friction, stakes, and specifiability allow. In the near term, the clearest winners are incumbents who own frequency and the real-world part of the value chain. The ones who are most at risk are those who rely on differentiating through discovery where LLMs can erode their current supply aggregation advantages.

• Transaction volume: €398.4 billion vs. €408.9 billion consensus (missed by ~3%), up 15% YoY (FX was likely a ~400 basis point headwind).

• Net revenue: €672.3 million vs. €688.0 million consensus, up 15% YoY / 19% FX-neutral (missed by 2% in constant currency).

• 2026 revenue guidance was revised from “low-to-mid twenties” at the November investor day to 20-22% FX-neutral.

• EBITDA margins are expected to be flat in 2026, largely due to the company hiring a net new 550-650 full-time employees. This represents a 12-14% YoY increase in headcount.

This results in 2026 consensus revenue / EBITDA / diluted EPS likely being revised down by ~2%/~4%/~5%. A few notes on highlighted cells below:

• Updated revenue guidance at 20-22% likely results in consensus at best anchoring to the midpoint 21%, but with another negative surprise, the market will be skeptical Adyen can achieve this range. This is especially true for the approximately 20% targets in 2027/2028 that Adyen set three months ago at its investor day.

• Updated EBITDA estimates assume margins are flat YoY, per company outlook. There could be some conservatism on margins given the company is guiding to flat margins with FTEs expected to grow 12-14%, unless wage inflation ticks up in 2026. Achieving Adyen’s 55%+ margin target by 2028 would require ~60% incremental margins in 2027/2028 assuming 20% net revenue growth vs. 53% in 2026 and 69% in 2025.

• Diluted EPS is estimated to grow slower than EBITDA since interest income is ~25% of net income and, at best, is likely to be neutral to growth and would become a headwind if rates are cut further.

Sources: Adyen public filings, Bloomberg, Ottavi estimates.

This is another quarter of negative estimate revisions. I posted the following chart in my last writeup (Adyen: Its Past & (AI) Future). If we assume that consensus estimates for 2026 EPS settle around the €39.47 that I’m estimating then this is ~16% below the level after Adyen’s Q4 2024 earnings in mid-February 2025. A stock down ~49% during that period may seem extreme but consistent negative revisions can result in the multiple overshooting to the downside (as can also happen on the upside when estimates exceed expectations), especially when considering the compounding impact lower current numbers have on future estimates and lost investor credibility. And Adyen’s multiple was at a high starting point at just under 50x next-twelve-month EPS. This is before incorporating the current market backdrop that is punishing companies for earnings misses.

Source: Bloomberg

The market is grappling with whether the consistent revisions are signs of increasing competitive dynamics (like we saw with Braintree in 2023), execution issues, a more saturated market, or a management team that over-promises. I don’t believe competition is a major concern, but it’s likely a combination of the other factors. Either way, it’s evident that Adyen’s management has created credibility issues with the market. Over the past two years in most quarters there has seemingly been a reason why numbers that miss expectations should be explained away. These include, eBay share shifts, “APAC merchants” (i.e., Temu and Shein) facing de minimis tariff headwinds in the US, and losing (or more likely, removing) Cash App.

2026 was supposed to be a cleaner story with the lapping of the negative impacts from each of these, aided by a record 2025 new customer cohort (the company states a new cohort typically contributes a mid-single-digit percentage to net revenue growth in its second year). All considered, this gave management confidence to issue a 20%+ revenue guidance for 2026-2028, but with 2026 expected to be “low-to-mid-twenties.” However, just three months later, the company revised the range lower to 20-22% FX-neutral, which Adyen’s CFO attributed to better understanding customer priorities through their year-end planning conversations:

Thanks for giving me the chance to clarify, because, again, I find this important. It’s not that we’ve lowered our view of our growth for 2026. It’s that we understand where the priorities of our customers will be in a deeper way and can share that. And it’s not about shifting priorities.

It’s about understanding where the focus of our customers will be in the following year, in any given year. I can give a couple of examples where things change from year to year. One of the reasons that we’re seeing strong growth in Latin America, for instance, is not only that we’ve been investing a lot in the product over the past years and we’ve built up a great team there and we have a strong offering, but also because, as there was some geopolitical tensions, some of our international, especially retailers, shifted some of their focus more towards that market. That’s one example of where you might see a bit of a shifting priority.

But it’s not that priorities have shifted in a big way and that’s what’s led to a lowering of our view. It’s just we have a good reflection, a refined reflection of our expectations for 2026, and that’s what we share. Now, those opportunities have a balance of when they drive revenues. Some of them drive revenues more short term.

Customer plans change and the macro remains dynamic, but this seems like an unforced error that further degrades credibility further. To be fair, some of the headwinds Adyen has faced are out of its control (i.e., tariffs), but it aggressively guided again when it didn’t need to. Adyen could’ve stated that it was targeting approximately 20% revenue growth over the next three years and that it would provide a tighter range on the Q4 call, setting guidance at 20-22%, but instead introduced the preliminary range it had to walk back. In addition to the revised guidance, the market is also grappling with the fact that Adyen’s Q4 growth was already below the bottom end of this range at 19 YoY FX-neutral.

Adyen is Squarely in Act 2

Putting management communication aside, Adyen’s core Digital business — where it processes online payments for enterprise merchants — is maturing. Digital net revenue grew 5% YoY on a reported basis, 10% FX-neutral, and 11% FX-neutral excluding Cash App. Further adjusting for the APAC retailer headwinds would get Digital FX-neutral underlying revenue growth to ~13% YoY. Adyen also called out an impact from a large Digital customer migrating to Unified Commerce as it rolled out in-person payments with Adyen. It’s unclear the impact this had, but maybe a few percentage points. Digital’s underlying FX-neutral net revenue growth was in the teens throughout 2025 and that pace seems likely to persist given Adyen’s relatively high penetration of enterprise ecommerce. To achieve its ~20% revenue growth goals, Adyen will need Unified Commerce and Platforms to make up for Digital’s shortfall. It was notable that Adyen positioned Unified Commerce as one of its two differentiators in the opening page of its H2 2025 shareholder letter:

Looking ahead, two factors increasingly define our differentiation. First, Dynamic Identification adds an intelligence layer to our platform, enabling real-time decisioning that improves conversion, reduces cost, and manages risk with greater precision as our customers scale across channels. Second, our Unified Commerce offering creates a structural advantage that is difficult to replicate at scale. By connecting in-person and digital data on a single platform, we provide a complete, end-to-end view of the customer journey. Together, these capabilities position Adyen to operate where commerce is most complex and most valuable, accelerating our progress toward becoming one of the largest fintech players in the world.

This is not to say Digital won’t continue to be a meaningful driver of growth, but it seems likely Digital will be a sub-20% revenue grower going forward (and management seems to be comfortable admitting that). This requires investors to believe that Unified Commerce and Platforms, which collectively account for 46% of net revenue, can offset a likely ~teens growth rate in the more mature Digital segment. For example, if Digital grows 10-15% in 2026, Unified Commerce and Platforms would together need to grow 25-33% to achieve the low-end of Adyen’s 20-22% range. Achievable given both segments exited 2025 already there on an FX-neutral basis, but highlights Adyen is squarely in its ‘second act.’

Fortunately for Adyen, it is well positioned for both, especially Unified Commerce because 1) Adyen is uniquely differentiated as the only next-generation processor that does offline/online well at scale; 2) it allows Adyen to address the in-person payments segment of the market which is ~80% of retail sales; and 3) the installation of terminals results in stickier customer relationships and wallet-share that can approach 100% vs. 40-60% for Digital.

Other Earnings Takeaways

Some other takeaways from earnings:

AI

Adyen announced its “Personalize” launch, which is a new module within Adyen Uplift. Personalize leverages its Dynamic Identification tool and allows merchants to dynamically alter the payment experience in real-time to increase conversion and lower costs. Pilots showed up to 6% higher conversion and up to 3% lower costs. I continue to believe that Adyen’s innovation with AI should help it distance itself from legacy processors and allow it to capture incremental market share.

Competitively, I continue to believe that AI will be a net tailwind to Adyen’s business (AI-enabled Dynamic Identification and fraud tools) rather than a headwind. However, there is the “AI risk” that Stripe is capturing much of the AI volume growth currently given most AI startups are launching on Stripe. This isn’t atypical – Adyen doesn’t service startups and focuses on enterprise and historically has won customers from Stripe and others as their business scales enough to graduate to “enterprise.” But in the interim, Stripe’s growth will likely outpace Adyen which creates the narrative that Adyen is “losing” to Stripe.

Customer, Geographical, and Product Expansion

Adyen announced expansion with Uber and Starbucks. Starbucks rolled out Adyen in Europe after being introduced to Adyen through Starbucks’ local franchisee in Mexico, Alsea. This mirrors a typical wallet-share expansion I discussed in my Toast example. Ultimately, Adyen’s goal is to win the big prize: capturing an increasing share of international and eventually, domestic processing. Starbucks has more than 41,000 stores and Adyen is cited as powering 943 of them across the UK, Austria, and Switzerland in its earnings release, with the total count likely closer to 2,000 when including Alsea. Winning the other international markets and eventually, the US business is how Adyen goes from ~5% wallet share to 50%+.

Our partnership with Starbucks demonstrates how Adyen supports high-traffic environments and complex, multi-market operations where speed and reliability at the point of sale are critical. In 2025, Starbucks selected Adyen to power in-store payments across 943 stores in the UK, Austria, and Switzerland, supported by 2,375 Adyen payment terminals. This European rollout builds on an existing partnership with Alsea, Starbucks’ licensee, where Adyen already supports Starbucks’ payments in Mexico — extending a single, unified platform across regions. The transition to Adyen’s platform was executed at pace and with minimal disruption. Over 900 stores were live in just seven weeks, with up to 125 stores onboarded in a single week, all during business hours.

Adyen also called out expansion in Latin America several times. This is likely supporting TikTok Shop and Temu’s expansion into Brazil and other Latin American countries.

Lastly, issuing continued to be a bright spot in the quarter, with volumes growing 8x YoY. Adyen called out corporate travel booking platform, Navan, as a case study. Navan’s issuing growth with Adyen grew 5x YoY in H2, with further expansion cited.

In 2025, issuing reached an inflection point, with volumes growing 8x YoY as platforms embedded cards directly into their core workflows. This shift supports a wide range of operational use cases, from supplier payouts to employee expenses, while reducing reliance on external banking relationships. Growth also accelerated across capital and account products. Capital loan volume more than doubled YoY with over 80% of businesses returning for subsequent loans. Recent launches such as Fresha Capital7 and YetiPay8 extend these capabilities into vertical SaaS platforms, where fast, embedded financing directly supports business growth.

Where does Adyen go from here?

All told, I don’t believe any of these concerns are disqualifying to an investment, especially with the company positioned to grow EPS at least in the high-teens percent compounded over the next three years with the stock at 23x/19x 2026/2027 EPS and sentiment poor. The growth drivers are in place for Adyen to sustain 20%+ revenue growth and on an underlying basis (ex-one-time issues cited), Adyen’s volume growth is well into the twenties currently. Adyen remains the dominant enterprise ecommerce merchant acquirer, is well-positioned to capitalize on AI, and is uniquely positioned to capture meaningful share in Unified Commerce and Platforms, but management has credibility to restore over the coming quarters.

Revisiting Adyen’s Past

Before getting into possible future scenarios, I think it’s valuable to briefly review the past since it provides useful context on why the stock is where it is today.

From the company’s June 2018 IPO until mid-August 2023, Adyen’s stock generated a 42% compound annual return, driven by a 36% revenue CAGR and 39% EPS CAGR.^[1] The stock was rewarded for this growth and execution—its average NTM EBITDA and P/E multiples were ~65x and ~90x, respectively, during this period (Covid-era valuations also played a role).

Source: Bloomberg

Adyen’s H1 2023 earnings on August 17, 2023 changed everything. Adyen reported significantly slower than expected growth, driven by increasing competitive intensity from PayPal’s Braintree (one of Adyen’s largest US competitors). Under PayPal’s old leadership (now, old-old leadership), PayPal pivoted to use Braintree as a loss-leader in many instances to win enterprise business, with the goal of bundling PayPal’s other higher-margin products (branded checkout, Hyperwallet, etc.). This uneconomic pricing resulted in Adyen losing wallet share among several key large enterprise clients, including eBay.

Adyen was in the midst of a multi-year investment cycle, scaling its engineering and sales teams materially, and the company experienced negative operating leverage as a result of slower revenue growth. For the first time since the IPO, questions about Adyen’s advantages and longer-term growth and profitability were raised. Adyen’s stock fell 39% the day of its H1 2023 earnings.

Within a few months after the selloff, the stock recovered much of its losses as the company hosted its first investor day in November and provided a Q3 update. However, since then Adyen’s stock has been mostly range-bound, with valuation multiples in a consistent downtrend. The stock currently trades at~27x NTM earnings on consensus numbers, which is the same as Adyen’s trough earnings multiple after the H1 2023 selloff.

Source: Bloomberg

Adyen shares are down ~17% year-to-date and ~25% over the past year. This is in part due to negative earnings revisions, with 2026/2027 adjusted EBITDA and EPS all down 10-14% from the recent peak in mid-March 2025. More recently, general payments sector volatility and emerging concerns about AI disruption have sent the stock down further.

Source: Bloomberg

Unpacking an Adyen Long Thesis

All told, the selloff has pushed Adyen to levels where the prospective returns are potentially more compelling than they have been in a while. At Adyen’s investor day in November, the company shared updated financial targets that include 2026 revenue guidance of low-to-mid twenties and for growth in 2027/2028 to be approximately 20%, with more specific guidance given annually.

Adyen increased its target EBITDA margins to 55%+ by 2028 and reaffirmed capital expenditures to 5% of net revenue, consistent with historical levels. If one believes that Adyen can achieve its guidance then the forward returns are potentially attractive, depending on the exit multiple assumed. Adyen trades at ~27x 2026 EPS for a stock that will likely compound EPS in the mid-20%s through 2028 if it achieves its guidance.

The analysis below assumes a year-end 2027 exit on 2028 numbers. The EPS estimates are informed by a few key variables, including 2026-2028 revenue CAGR and incremental EBITDA margins. All other income statement assumptions (tax rate, interest income) are in line with current trends. If one believes Adyen can compound revenue at 20-22% through 2028 then an EPS range of €62.69-65.90 is plausible. For this example, assuming 23-27x 2028E EPS would generate an IRR of 12-26% by year-end 2027. As of this writing, Adyen’s stock trades at €1,156.

Note: these assumptions are meant to be illustrative to frame the potential stock returns in different scenarios and are not a recommendation or investment advice.

Source: Adyen public filings, Ottavi estimates

The core building blocks to achieve 20%+ revenue growth in 2026 and ~20% in 2027/2028 are predicated on ~40% from existing customer market volume growth (assuming “high-single-digits” is ~8% growth) and ~60% from new wins, expanding the wallet share of existing customers, and adoption of newer products (financial products)—offset by Adyen’s tiered pricing model.

Simplistically, for market volume growth, one can use ecommerce growth as a proxy for Adyen’s digital volume and Visa and Mastercard’s volume growth as a proxy for point-of-sale volume, which was 23% of Adyen’s total in Q3 2025. Visa and Mastercard are expected to grow volume ~9-10% in 2026, while global ecommerce is forecast to grow 7.2% in 2026.^[2] A more granular analysis would examine sales trends of Adyen’s largest merchants or focus on trends in Adyen’s largest verticals, like ecommerce, luxury and food & beverage. Assuming consumption trends stay consistent, at least in 2026, Adyen’s market volume growth assumptions appear achievable, but remain subject to the macro environment.

Source: Adyen’s November 2025 investor day

Adyen’s Wallet-Share Gains

Adyen estimates that its wallet-share for its oldest cohorts is 40%+, but still with room for expansion, while earlier cohorts can be <20%. Adyen’s wallet share can be as high as 100%, but varies depending on the merchant type and whether Adyen is processing online, offline—or both (i.e., “Unified Commerce”). Offline relationships tend to move closer to 100% over time due to the installation of hardware terminals in-store, while online is typically split between 2-3 processors, with one processor capturing the majority of the volume, while the others are leveraged for redundancy and for negotiating lower prices at renewal.

Source: Adyen’s November 2025 investor day

A typical merchant lifecycle might resemble something like this: a large multinational consumer brand based in the US with operations across the US and Europe might start processing with Adyen in the UK to test Adyen’s performance, particularly on key metrics like authorization rates. Adyen’s net take rate for this merchant processing just in the UK might be, for example, 25 basis points, but to incentivize the merchant to shift more wallet share to Adyen across other European countries and the US, Adyen will offer different pricing tiers if volume targets are achieved. If Adyen’s wallet share increases from, say, 10% for the UK-only to 50% as the merchant goes live with Adyen across all of its markets then the 25 basis-point take rate might be reduced to 15 basis points. Adyen generates more revenue per merchant and the merchant saves on per transaction costs. Adyen’s marginal costs to process an incremental transaction are nearly zero, so it will gladly exchange lower revenue per transaction for higher overall revenue and profit per merchant. Adyen’s tiered pricing and wallet share gains work hand-in-hand.

A more concrete real-world example of this dynamic is Toast. Toast uses Adyen as its acquirer in international markets (the UK, Ireland, and Canada), while it uses Worldpay and Chase in the US (process note: Toast identifies which merchant acquirers it uses in its Merchant Service Agreement for each country). Toast is likely using these international markets as a testing ground before deciding whether to expand with Adyen in the US.

It’s plausible that eventually Toast will want Adyen to be one of its two acquirers in the US and reduce the total number of global acquirers from three to two. My guess is that Worldpay would be the most at risk since Toast likely uses Chase for other banking services and has a subsidized rate. If so, Adyen’s expansion to Toast’s US business would unlock an addressable ~$200 billion in incremental payments volume.^[3] In this example, Adyen likely only captures 5-10% of Toast’s volume, but there’s a defined pathway for this to increase to 50%+ over time if it becomes one of Toast’s two processors in the US.

Navigating AI

An increasingly important consideration for Adyen’s ability to retain and grow with existing customers, and win new ones, is its ability to leverage AI and adapt to the changing ecommerce landscape as agentic commerce emerges as a new sales channel. At least part of the reason for Adyen’s recent selloff is that the market is assessing whether Adyen will be a winner or loser in this new and evolving paradigm.

As I briefly highlighted in my recent post (Payments: Assessing the AI, Agentic, and Stablecoin Risks for Merchant Acquirers), the agentic debate largely comes down to whether one thinks agents will leverage the existing payment rails or use new ones (i.e., protocols using stablecoins)—and how the value chain will evolve in the payments stack within agentic commerce. My working hypothesis is that the existing rails are best positioned to win in agentic for the majority of traditional consumer purchases.

For Adyen, agentic commerce and AI are intertwined. First, ~77% of Adyen’s payment volume is transacted online, where fraud is 8x higher than in offline channels, with authorization rates 6 percentage points lower than in-person transactions.^[4] With the proliferation of AI-enabled fraud by bad actors, protecting personal information, while correctly verifying legitimate transactions is arguably more challenging than ever. The most prevalent and likely use case for Adyen with AI is creating more robust fraud tools (in addition to leveraging it for more efficient internal engineering), which is where its AI-enabled product suite—Adyen Uplift—comes into play. Adyen Uplift reduces false positives by 42% and delivers conversion improvements of 1-6%.^[5] Adyen is well-positioned to capitalize on AI given its single tech stack that it’s built entirely in-house—with no M&A— and its immense scale of €1.4 trillion in payments volume, with data spanning offline/online, across verticals, and geographies.

Adyen used the following example to show the type of results conversion increases from Adyen Uplift at the company’s November 2025 investor day:

Source: Adyen’s November 2025 investor day

Agentic Commerce Increasing Payments Complexity

The agentic landscape is quickly evolving, including which standard will help govern agentic economics, roles, and rules. Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol (ACP) are two protocols jockeying to set the industry’s standards. Ultimately, for Adyen, this increasing complexity and commerce fragmentation should be a net benefit that reinforces its value proposition as it integrates with all the different payment modalities and enables merchants to focus more on what they do best. Merchants want to minimize the number of processors it uses and will favor those that reduce complexity in selling across channels.

Adyen framed this in their November 2025 investor day as being “the universal translator”:

So in this new ecosystem, we see Adyen as a universal translator for agentic commerce. We see this as the bridge for merchants to be able to manage and process any payment they want through any AI protocol, through any AI agents. One API, one platform, full control. And this isn’t just about payments. It’s about data integrity. It’s about trust. It’s about identity. When you combine all these three things together, it’s not about this one-off transaction. It’s about loyalty with that customer. And so while others are focusing on single components, authentication discovery, we’re focusing on the [rails] that pull all these components together in a seamless fashion for our merchants.

Source: Adyen’s November 2025 investor day

In the US, there has been a valid argument over the past several years that merchant acquiring in ecommerce was becoming commoditized, particularly since most online purchases run on Visa and Mastercard rails. However, agentic commerce could reverse—or at least slow—this trend, as payments become increasingly complex.

Either way, the application of AI and enabling agentic commerce should allow Adyen (and Stripe) to further distance themselves from legacy providers who cannot innovate fast enough to meet the increasing demands from merchants. For Adyen, this should at the very least be supportive of its wallet-share gains and new wins assumed in its mid-term targets and could even accelerate momentum.

^[1] Adyen; revenue and EPS CAGRs are from 2018-2023.

^[2] Bloomberg, Shopify, eMarketer.

^[3] Bloomberg consensus estimates for Toast payment volume in 2026 is $232 billion, with majority from domestic.

^[4] Visa’s Global Product Drop. April 04, 2025.

^[5] Adyen

In total, all public merchant acquirers are down over the past year, with relative underperformance versus the S&P 500 ranging from 42% ($ADYEN.NA) to 89% ($FISV).

Source: Bloomberg

An increasing number of companies trade for single-digit P/E multiples, which implies the market believes these companies are facing terminal risks and will not be able to sustain even modest levels of EPS and FCF per share growth. This post focuses on the existential risks (AI, agentic, stablecoins) rather than any specific individual stock.

Quick Payments Primer

As a quick primer, most US-based digital payments are transacted through the four-party model popularized by Visa ($V) and Mastercard ($MA). These networks sit at the middle of the ecosystem as the orchestration layer, setting rules and facilitating communication between merchant acquirers and card issuers. American Express ($AXP) and Discover (now part of Capital One, $COF) operate closed-loop models where they control all major aspects of the payment process. This write-up focuses on the merchant acquirers (“banks” in the graphic below), which include Adyen ($ADYEN.NA), Block ($XYZ), Fiserv ($FISV), Global Payments ($GPN), PAR Technology ($PAR), Shift4 ($FOUR), Stripe (private), and Toast ($TOST).

Source: Marqeta

Acquirers contract directly with merchants, simplifying the process for them to accept payments. Acquirers handle funds flow, fraud detection, and chargebacks. In exchange, acquirers capture a net take rate of approximately 10–120 basis points. The graphic below is from Goldman Sachs; it is somewhat dated (2022), but the figures have not changed substantially, and the general premise still holds: the take rate spectrum increases as you move from enterprise clients to SMBs. Similarly, the more of the payments stack a processor controls in-house, the more economics it captures.

For example, Toast ($TOST) operates as a payments facilitator (PayFac), meaning it handles the front-end but partners with an acquirer (such as Worldpay or Adyen) and therefore gives away some of its economics. By contrast, Shift4 ($FOUR) handles acquiring in-house and captures more of the payments economics than Toast, even in a common vertical like restaurants.

Source: Goldman Sachs

Source: Goldman Sachs

While all the businesses listed above operate in the payments space, there are nuances in their models. Some subsidize hardware costs ($FOUR), while others generate a meaningful portion of their revenue from software ($TOST, $SHOP). Some focus on organic growth, while others pursue an acquisitive strategy.

Historically, merchant acquiring has been a competitive business and the industry continually moves toward commoditization. Even with these challenges, merchant acquirers have shown to have retentive merchant bases, stable take rates (on a like-for-like basis), and generate significant free cash flow. However, the payments landscape is experiencing more potential change over the coming years than it has in maybe the last few decades. There are three fundamental potential changes happening to the ecosystem: AI-Driven Innovation, agentic commerce, and stablecoins.

AI-Driven Innovation

If AI significantly reduces the cost of software development, this could theoretically create an increasingly competitive playing field (in a market that is already crowded). However, building a payments company is more challenging than “vibe-coding” a competitor. Creating a viable payments operation requires the ability to facilitate money movement. Even if a startup takes the PayFac route (partnering with an acquirer), it still must register with different payment networks, partner with an acquiring bank that underwrites payment risks, and meet extensive compliance requirements (e.g., PCI DSS Level 1 Certification, state money transmitter licenses, KYC, AML). While not impossible, these requirements demand significant capital and resources, creating meaningful barriers to getting started. Some acquirers, like Adyen ($ADYEN.NA), hold banking licenses and are integrated directly with central banks, providing additional competitive advantages.

Furthermore, in physical retail, cloud-based acquirers have taken significant share from legacy providers (NCR, MICROS), but they have achieved this through distribution networks. Selling physical hardware to SMBs requires “feet on the street.” In payments, distribution has proven to be a relatively defensible strategy. Legacy merchant acquirers still retain something like half of the US restaurant market even with the success of $TOST, $XYZ, $FOUR, and $PAR. Even if a vibe-coded competitor integrated payments, it would likely need to build out a salesforce to execute on its strategy.

Alternatively, AI may provide a net benefit to acquirers through improved fraud detection. Adyen Uplift—Adyen’s AI module—leverages over $1 trillion in annual payment volume to reduce fraud and increase conversion rates. In select case studies, $ADYEN has found authorization rates increase by 6% with Uplift. While fraud is not as significant a concern in physical retail, it remains prevalent in e-commerce. Although Adyen Uplift is still in early stages, this capability is expected to be a tailwind to monetization and should allow Adyen to capture greater wallet share versus subscale processors.

One risk to existing acquirers would be if those who have historically charged premiums for software revenues experience deflation—or impaired pricing power—because competitors that have historically had weaker R&D innovation see the playing field leveled by AI-generated code (consider $FOUR versus $TOST).

Agentic Commerce

It seems inevitable that AI agents will make purchases on behalf of consumers in the not-too-distant future. Unless robots advance to the point of physically purchasing goods for us, this is primarily a consideration for e-commerce acquirers.

Agentic commerce would present potential risk if it disintermediated acquirers from the checkout process. However, this scenario seems unlikely. Current evidence suggests that agentic commerce will leverage existing payment rails rather than circumvent the current system. For example, ChatGPT Instant Checkout—which allows consumers to purchase products within the ChatGPT experience—requires consumers to click a “buy” button to confirm payment details and complete the purchase, following the standard e-commerce checkout process. Currently, ChatGPT Instant Checkout is powered by Stripe and Shopify (which itself uses Stripe), with payment fees comparable to traditional e-commerce purchases (2.9% + $0.30). This excludes the fee ChatGPT charges in addition to this for demand generation.

ChatGPT Instant Checkout experience

Source: ChatGPT

One counterargument is that the current experience is human-led and feels analogous to purchasing in any other e-commerce marketplace, rather than a truly autonomous agent completing the purchase for the consumer—which is where the market will inevitably evolve to. That is valid, but it still seems more likely than not that agents will leverage existing infrastructure rather than build entirely new payment rails. One could argue that the proliferation of agents will actually require more safeguards around payment identification, which is precisely where existing payments infrastructure is well-positioned. For example, $V launched Visa Intelligent Commerce in April 2025, which works to identify agents and provide them with tokenized credentials to reduce fraud and facilitate seamless future purchases. If agentic commerce rides the $V / $MA rails then acquirers should remain well-positioned as the four-party model remains intact.

However, agentic commerce does open the possibility for new types of consumption that do not currently exist at scale. Consider micro-transactions: in-app purchases or access to individual articles behind hard paywalls. If micro-transactions end up being significantly smaller than the average purchase today (e.g., $0.50 for an article), the economics of the existing payments model do not work. Using Stripe’s 2.9% + $0.30 pricing, $0.31 of the $0.50—or 62% of the transaction value—would go to Stripe, leaving little for the merchant. While the 2.9% + $0.30 rate is for SMBs, even interchange++ pricing (common for enterprise merchants) would still consume much of the transaction value given there is typically a fixed cost component that goes to the acquirers. Ultimately, the pricing model would need to change. Either transactions would need to be batched, or models would need to evolve. This could open bottom-up disruption from stablecoins, which offer an orthogonal approach to disrupting payments given traditional acquirers likely cannot address these volumes through their existing economics.

Stablecoins

Stablecoins have scaled rapidly in recent years, bolstered by the passage of the GENIUS Act in July 2025. Stablecoins promise cheaper, faster, programmable money that has the potential to disrupt global payments. According to Artemis Analytics, stablecoin transaction volume is at an approximately $35 trillion run-rate. However, much of this volume currently facilitates or settles exchange trading volumes rather than payments for products or services. Retail-based stablecoin payments are still nascent compared to Visa and Mastercard’s combined global payments volume, at ~$76 billion, which is 0.3% of Visa and Mastercard’s combined global payments volume of $23+ trillion in 2025. It seems that stablecoins are most likely to address cross-border, business-to-business, and treasury money movement rather than consumer-based-payments (in the US).

Even in a scenario where stablecoin payments take meaningful share from traditional card-based payments, volume shifts would likely be gradual rather than abrupt. In such a scenario, merchants will not want multiple acquirers. This is particularly true for physical commerce merchants that tend to route 100% of their volume through one acquirer, which is integrated with (or even provides) their payment checkout hardware.

It seems more likely that stablecoins become another payment rail that acquirers accept on merchants’ behalf. The increasing complexity is actually a net benefit to acquirers who can simplify the process for merchants to accept multiple payment methods (cards, BNPL, account-to-account, Venmo, stablecoins). An increasing number of merchant acquirers already provide the ability to pay with stablecoins at checkout.

Stablecoins could also benefit merchants through faster settlement. Shift4 ($FOUR) announced in December 2025 that it is allowing merchants to settle with stablecoins instead of bank transfers, enabling 24/7 fund movement.

This was a quick overview of how I view AI, agentic, and stablecoins but is by no means an exhaustive analysis of each. Some of these areas are quickly evolving so my opinion may change as the facts do. Any thoughts or pushback is welcomed.

Figure Technology Solutions (ticker: FIGR), hereafter referred to as “Figure,” is a company disrupting the non-qualified mortgage market through its blockchain-based solutions, with nascent expansion into the broader credit and equity markets. Figure sits at the forefront of the tokenization trend, which could see trillions of dollars of assets move on-chain over the coming years.

No matter the asset class, there are commonalities in Figure’s approach: use vertical integration and blockchain to reduce costs through automation and disintermediation, while scaling transactional liquidity. In HELOC, Figure in-housed the loan origination system, appraisals, and lien registry, while automating income verification and providing e-close capabilities. This reduced the origination process from 40+ days to a median of 10 days, while cutting origination costs 90%+ versus the average mortgage process.^[1]

Typically, Figure demonstrates proof of concept by being its own first “customer.” In HELOCs, for example, Figure spent years originating mortgages on its own balance sheet before allowing partners to white-label its technology and originate their own branded mortgages using Figure’s loan origination system (LOS). The company is doing the same with its On-chain Public Equity Network (OPEN) by launching a tokenized version of its stock natively on-chain that is fungible with its NASDAQ-listed security.

As of Q3, Figure had 246 origination partners, up 46% sequentially, and 118% year-over-year, with partner-branded origination volume accounting for 76% of total Consumer Loan Marketplace Volume for the trailing-nine-months—either through white-labeling Figure’s technology and/or selling mortgages on Figure Connect (the company’s transaction marketplace where originators and investors can buy and sell loans), or both.

Source: Figure’s Q3 2025 earnings presentation

Ultimately, the goal is to turn Figure Connect into a “TBA market” equivalent, providing the non-qualified mortgage (non-QM) market a liquid secondary market akin to what Fannie Mae and Freddie Mac offer for qualified mortgages (QM). Figure also recently started scaling several adjacent and interrelated businesses. Democratized Prime is Figure’s blockchain-based competitor to traditional prime brokerage and overnight funding markets, which will be a critical component to its end-to-end mortgage marketplace. OPEN aims to capitalize on the stock tokenization still in its infancy, but is positioned to grow substantially, especially once the crypto market structure legislation is passed. YLDS, Figure’s fixed-income security that combines the protections of a stablecoin but can pay yield (SOFR minus 35 basis points), has shown initial signs of scaling, with its in-circulation market cap growing from $21 million in November to $369 million currently.^[2]

Figure’s net take rate has averaged approximately 6% of Consumer Loan Marketplace Volume, comprising:

● Ecosystem fee: ~3% on mortgages originated through the platform

● Gain-on-sale: Margin captured when mortgages are sold to investors

● Partner fees: Ecosystem and technology fees from origination partners

● Securitization fee: 40 basis points on mortgages Figure securitizes

● Servicing fee: 25 basis points on mortgages Figure services

Figure also earns interest income on cash and other liquid assets. For Democratized Prime, Figure charges 50 basis points annually on total assets.

Even with its nascent businesses providing little financial impact today, Figure’s core HELOC business is scaling fast—and profitably. In 2025, the company generated $8.4 billion in Consumer Loan Marketplace Volume, up 63% year-over-year. The company also pre-announced Q4 results on January 12, showing Consumer Loan Marketplace Volume further accelerated to 131% year-over-year, up from 40% in Q2 and 70% Q3. This acceleration reflects both growth in origination volumes and scaling Figure Connect, which accounted for 46% of Figure’s total Consumer Loan Marketplace Volume after launching in June 2024. Consensus estimates expect Figure to generate $694 million in revenue in 2026, up 36% versus 2025, and $364 million in adjusted EBITDA, up 48%.^[3] The company’s expected adjusted EBITDA margin of 52% demonstrates the scalability of its model. The stock trades at ~23x 2026 EBITDA and ~42x 2026 earnings. In a future post, I’ll outline the financial details of an investment thesis, including detailing why I believe consensus estimates are conservative.

Figure was co-founded by Mike Cagney and June Ou. Cagney currently serves as Figure’s Executive Chairman and focuses on bigger picture strategic initiatives, while CEO Michael Tannenbaum oversees the company’s day-to-day operations. Cagney and Tannenbaum first worked together at SoFi, where Cagney was CEO and Tannenbaum was CRO.

The US Mortgage Market

The US mortgage market originated approximately $1.8 trillion of loans in 2025, divided into two broad categories: Qualified Mortgages (QM) and Non-Qualified Mortgages (non-QM). QMs are loans the US federal government guarantees through government-sponsored entities (GSEs) Fannie Mae and Freddie Mac. To qualify—or “conform”— mortgages must meet certain criteria established by the government. In 2024, ~95% of mortgage originations were QM with non-QM making up the remainder. Non-QM’s share has increased from nearly zero a decade ago.^[4]

For 2026, the FHFA (Federal Housing Finance Agency) raised the baseline conforming loan limit for one-unit properties to $832,750, up from $806,500 in 2025 (a 3.3% increase). In high-cost areas, the ceiling can reach $1,249,125. Additional conforming requirements include minimum credit scores, debt-to-income ratios, and loan-to-value levels. Loans that fall outside these parameters are classified as non-qualified. Common examples include jumbo loans (those exceeding the conforming limit), debt-service-coverage-ratio (DSCR) loans (common for investment properties), and bank-statement loans (typically used by self-employed borrowers). The non-QM market is where Figure competes today.

The typical mortgage process takes 55+ days to complete, including over 40 days to originate the mortgage and approximately ~15 additional days to sell the loan into the secondary market.^[5] The mortgage process looks something like this:

Much of the US mortgage market volume is originated by independent mortgage banks (IMBs). In 2024, IMBs originated over $1.2 trillion in mortgages, accounting for 68% of the $1.8 trillion total. This is up materially from 24% in 2008 and 55% in 2018 as large banks have pulled back on their mortgage practices. However, most of the loans IMBs originate are first-lien. Their share of subordinate loans—which include HELOCs and other non-qualified products—was significantly lower at 16%.^[6] This is due to the challenging origination economics for non-QM and a lack of a secondary market. This gets to the heart of what Figure’s addressing.

Funding & Warehouse Lines

Because IMBs are not depository institutions—and therefore do not have the same low-cost deposit funding that banks enjoy—they do not retain mortgages once originated and instead, sell them into the secondary market. Without access to deposits, IMBs enter into “warehouse lines” with financial institutions to fund the gap between when a mortgage is originated and when it’s sold in the secondary market. This warehousing period averages ~15 days but can extend to several weeks.^[7]

Without retaining loans, IMBs and other originators that sell mortgages are in the business of capital turnover—maximizing how quickly they can originate, close, and sell a loan to achieve unit profitability. Since revenue is largely defined by the mortgage size, interest rate, and credit quality, the objective centers on minimizing costs. Traditionally, this means leveraging technology to automate and reduce operating costs, increasing throughput to generate operating leverage, and reducing variable costs tied to time—specifically, personnel overhead, warehouse expenses, and the opportunity cost of capital being tied up in a warehouse line rather than originating additional loans.

Rising Origination Costs

Complicating this calculus is the rising costs of mortgage originations. For roughly a decade post-GFC, originators’ average per loan revenue and expenses moved in relative lock-step even as mortgages remained a relatively low-margin product (typically a single-digit net margin).

Source: Freddie Mac

However, the pandemic disrupted unit economics. Initially, originators experienced massive economies of scale as an explosion in origination volumes spread fixed costs over more loans and a higher mix of refinancing activity which tends to have better unit profitability boosted margins. However, as volumes receded, originators struggled to maintain unit profitability, as operating leverage reversed and cost inflation continued to persist.

Much of the time and expense is due to mortgages still being a human-led process, with ~67% of per origination cost from personnel expenses (i.e., underwriting and processing). This is in part due to the processes required by Fannie and Freddie that results in human-aided documentation collection and verification. Furthermore, industry estimates showed that technology-related expenses increased from 2% per loan in 2021 to 4% in 2024.^[8]

This results in many mortgage originators not generating sustainable profitability. The Mortgage Bankers Association (MBA) estimates that, on average, originators <$400 million in annual volume lose ~$1,000 per loan originated. This cohort of ~4,000 institutions originates ~$230 billion in annual mortgage volume—or ~13% of total industry volume. IMBs, on average, generated eight straight quarters of net production losses from Q2 2022—Q1 2024.^[9] Even in good times, net margins are typically in the single-digits.

The TBA Market: The Lifeblood of the Mortgage Market

Once a mortgage is originated, it typically enters a warehouse line and is then sold in the secondary market. The secondary mortgage market is the lifeblood that allows the US mortgage system function. Loans are typically sold to GSEs Fannie Mae and Freddie Mac in the To-Be-Announced (TBA) market, which provide a guarantee in exchange for originators adhering to conforming guidelines. Beyond the guarantee, the market’s standardization enhances transactional liquidity. When including origination volumes transacted and other secondary transactions, the TBA market facilitates $250-300 billion in daily trading volume—many multiples the active mortgages outstanding on an annual basis.^[10]

Gain-on-sale revenue also generates the majority of the origination revenue—typically 70–80% of the total—while origination fees charged to the borrower (including origination, processing, underwriting, and administrative fees) are typically ~10–20%, with warehouse net income making up the remainder.^[11] Warehouse net income represents the spread between interest earned while holding the mortgage and the cost of the warehouse facility. The time to sell a mortgage, or “dwell time,” can extend up to 60 days. In total, an originator can generate revenue amounting to ~300bps of the mortgage amount.

While TBA liquidity is dependable, the process still has inefficiencies. The typical mortgage can spend at least 15 days in the warehouse facility as it goes through the necessary documentation and approval process. This ties up capital, although the incremental interest cost is generally offset by the positive carry from the mortgage’s higher coupon relative to the borrowing rate. An originator that generates $800 in profit per loan can generate 3x the amount of total annual profit by cutting dwell time from 15 days (TBA) to 5 days (Figure). This excludes the additional operating leverage from higher volume turnover.

Figure’s Disruption Begins with an Underserved Market: HELOCs

HELOCs gained notoriety during the GFC as debt balances expanded due to lax underwriting standards, including practices that allowed homeowners to use HELOCs to “piggyback” their existing mortgages—in many cases, to avoid paying private mortgage insurance (PMI) by increasing the down payment amount to the 80% threshold. Following the GFC, HELOC origination volumes declined materially as both bank and non-bank originators pulled back.

Underwriting standards are more stringent today. When used prudently, HELOCs allow homeowners to tap into their home’s equity to pay for renovations, other expenses, or simply to have a revolving credit line available. After more than a decade of consistent decline, HELOCs started to grow again in 2021 as a way for homeowners “locked in” at low mortgage rates to tap into their equity, which is at an all-time high of $34+ trillion in the aggregate.

In Q4, HELOC balances amounted to $422 billion, up 9% year-over-year—amounting to 1.2% of total home equity versus 7.9% at the peak during the GFC and 1.0% at the nadir in Q2 2022.

Figure started with HELOCs because (1) it was an underserved segment of the mortgage market; (2) it could be meaningfully improved through vertical integration and blockchain; and (3) it provided a wedge to build Figure Connect. Additionally, Figure hypothesized that it could create an application and origination process for HELOCs that rivaled that of personal loans—leveraging co-founder Mike Cagney’s experience from SoFi. The result was a process that reduced the median application timeline to 10 days vs. the typical mortgage of 42 days, and cut total origination costs to ~$750.

Source: Figure’s S-1

Beyond the demand-side pull from a significantly better origination process, Figure attracted origination partners because it allowed them to tap into a market that was largely unavailable to them due to suboptimal unit economics. Revenue generated per HELOC originated tends to be lower than for a primary mortgage, largely due to smaller average loan amounts. Using rough numbers and assuming average revenue per originated mortgage of ~300bps: a $100,000 HELOC will generate ~$3,000 in revenue versus ~$12,000 for a $400,000 primary mortgage.

While per-loan operating expenses may be lower—particularly when the HELOC borrower is an existing bank customer—the low revenue base makes it difficult to generate profits without a largely automated process. If the average personnel costs are ~67% of average mortgage’s origination costs—or $7,900—then originators would need to cut this expense by 62% just to get to break-even (assuming the originator could eliminate all other costs). This is a tall—and unlikely—task for nearly all originators.

Figure Connect: Creating the “TBA Market” of Non-Qualified Mortgages

Even if an originator can make the unit economics viable, the lack of a standardized secondary market for non-QMs further compounds the challenge, making loan disposition more difficult and costly. This is why post-GFC, most HELOCs—and subordinate loans generally—are originated by depository institutions that can leverage HELOCs as customer retention tools rather than profitable standalone products. Unsurprisingly, IMBs only account for 16% of subordinate mortgages versus 68% for total originations. Figure’s value proposition has attracted 246 origination partners—and accelerating. This includes ten out of the twenty largest IMBs. Figure’s HELOC market share among IMBs reached an estimated 19% in 2025, up from 13% in 2024. However, its total market share remains relatively low at 4% of the total market.^[12]

Convincing origination partners to adopt Figure’s white-labeled solution was a critical first step, but insufficient for Figure to ultimately create an end-to-end process that rivaled Fannie and Freddie for non-QM. To do so required creating a liquid, standardized marketplace where originators could seamlessly transact with mortgage buyers, which is where Figure Connect comes into play.

Figure Connect allows partners who originate using Figure’s LOS to sell their loans to investors. To create a similar guarantee mechanism, Figure launched a joint venture with Sixth Street that would act as a “guarantor vehicle”—giving originators confidence they could sell originated loans. Figure also paired Connect with Democratized Prime, allowing partners to replace traditional warehouse lines with bilateral, direct connections to lenders.

As it did with the origination process, Figure leveraged vertical integration and blockchain technology to eliminate intermediaries and cut costs. The advantages show up in a few key areas:

Bringing the homogeneity of QM to Non-QM: Figure’s LOS process produces anonymized data that lives on the blockchain. Critically, Figure’s Digital Asset Registration Technologies (DART) displaces the traditionally used Mortgage Electronic Registration Systems (MERS). DART “listens” to a transaction when consummated on Provenance—the L1 blockchain Figure uses (and which it owns ~25% of Provenance’s outstanding HASH tokens)—updating the registry in real-time. This eliminates the paper documentation process of MERS and ensures there’s no double-pledging of loans. The typical QM can have a defect rate in the mid-to-high-single-digits.^[13] Figure reduces this to zero.

Reduction in third-party review costs: Figure estimates saves 80% of securitization costs because its loan data is immutably stored on the Provenance blockchain. Compared to standard securitizations in which ratings agencies may review 100% of originated loans, leveraging the blockchain reduces this to a sampling of only 20% for Figure’s AAA-rated securitizations.

Faster time to payout: Figure reduces settlement time to ~5 days versus 15+ days for Fannie and Freddie, and potentially significantly longer for private non-QM purchases. This increases volume turnover.

Since launching in June 2024, Figure Connect has scaled from near zero to ~90% of total Figure addressable volume (Figure-branded and Figure partner volume) within a year. This means that nearly all mortgages originated by Figure or its partners through the LOS are subsequently sold to investors on the Connect marketplace.

With LOS, Connect, Democratized Prime, securitization capabilities, and in-house servicing, Figure vertically integrates the entire non-QM process. Figure started in HELOCs but recently expanded to DSCR loans and crypto-backed loans. Furthermore, its HELOC product is tapping into more than just subordinate loans, with 20% of its volume being first-lien as of Q3.

Source: Figure’s Q3 earnings presentation

The Big Unlock: Figure Certified

Mortgages originated through Figure’s processes will continue to power Connect ’s growth, but the next major catalyst will come from introducing Figure Certified. While details remain limited, Figure’s November 2025 S-1 provided some additional context:

We are introducing a Figure Certified program that employs proprietary technology and artificial intelligence to validate assets originated by third parties outside of our platform against our LOS underwriting guidelines, such that we can confidently tokenize the loans on the Provenance Blockchain and also support the sale of those assets via Figure Connect. Figure Certified loans would be allowed to transact in Figure Connect, increasing the scope of our marketplace beyond our LOS originated assets without losing the benefits of homogeneous collateral. We are targeting a launch of the Figure Certified program by the fourth quarter of 2025.

Figure was targeting a Q4 launch, but it does not appear to have launched as of today. Once operational, Figure Certified will expand the company’s addressable market to all non-QMs and other types of credit (auto and personal loans). This will allow it to address the entirety of the ~$90 billion non-QM market and create a Connect marketplace that can facilitate multiples of that in transactional volume per year. Eventually, if Figure creates enough of a standardized process, we could see it start to capture some of the QM volume from their origination partners who already leverage Figure for HELOCs. This addressable QM origination volume is in the hundreds of billions per year just from Figure’s existing customer base.

Figure has built the infrastructure to do for non-QM what Fannie and Freddie did for conforming mortgages, and it now offers the end-to-end solution to create a liquid, standardized marketplace that could increase efficiency and expand the overall market. A future post will go into the mid-term investment thesis in more detail, including how I’m modeling growth and different scenarios.

^[1] Figure public filings, MBA

^[2] As of February 4, 2026, per DefiLlama

^[3] Bloomberg

^[4] JPMorgan

^[5] Freddie Mac, MBA

^[6] MBA, HMDA

^[7] MBA

^[8] Fannie Mae

^[9] MBA’s 2025 IMB Fact Sheet

^[10] SIFMA

^[11] MBA

^[12] Bernstein

^[13] Freddie Mac

I’ve worked on the buy-side for a decade and a half—most of it investing in later-stage private and publicly traded companies.

My investing style is rooted in fundamentals. I believe a company’s value ultimately comes down to its long-term free cash flow per share. I get most interested in opportunities where I feel the market has incorrectly valued a company’s free cash flow generation, and I gravitate toward situations where change creates confusion about these long-term prospects.

I started Ottavi to create an “investing journal” of sorts—a living record of my thoughts, ideas, and random musings. By doing so, I hope to foster constructive dialogue, continue learning, and perhaps help others along the way. The purpose isn’t to advise anyone on how to invest, but to share situations and frameworks I find interesting, which may help inform their own opinions.

Ottavi will be my research process deconstructed. It will include long-form industry primers, idiosyncratic situations, ad hoc company or sector updates, and more. Some posts may not make sense in isolation, but hopefully over time, a mosaic forms that displays how I think and invest.

First, I believe some clarifications about the Dot-Com bubble are required, which provide useful context. Corry Wang (@corry_wang on X) had a great thread on this topic back in Jan 2021. This was posted in the throes of the Covid Tech bubble, but is as relevant today. The entire thread is worth reading, but his findings that are most relevant to this post:

• The Dot-Com era was largely a large-cap growth bubble—not an Internet stock bubble.

• The ‘Internet’ subcomponent of Bernstein’s (Wang’s former employer) ‘New Economy’ categorization was only $648 billion, or less than 10% of the New Economy’s total capitalization and about 4% of the total market. Internet fads like Webvan receive considerable attention, but these companies were only a small part of the story. Software and companies related to the telecom/fiber buildout accounted for much of New Economy’s market cap.

• Reflexivity was real, but on a lag.

With this in mind, what’s the relevance to current crypto markets? For starters, crypto’s market cap in today’s terms is similar to the Internet’s market cap back in 2000—around $3 trillion.

Here’s the math: Since 2000, US GDP has compounded at about 4%, while the total market cap of US equities has grown at roughly a 6% CAGR. This corresponds to a US GDP of $30.6 trillion and a U.S. stock market capitalization of approximately $68 trillion. If we assume that the ‘New Economy’ and ‘Internet’ market caps grew at the same rate as the overall stock market, their total values would be $30.4 trillion and $2.8 trillion, respectively—the latter of which is a close approximation to crypto’s widely cited market cap figure.

Sources: Corry Wang, Bernstein, Federal Reserve, Ottavi

‘Internet’ in today’s terms is more meant figuratively to encapsulate the market cap of an emerging theme rather than Internet companies themselves, which are obviously worth much more than $2.8 trillion today.

Additionally, I think it’s a fair assumption that most would view the companies in today’s ‘New Economy’ basket as those benefiting from AI, not crypto. I forget the exact statistic I saw recently, but it put the percentage of the US stock market that benefits from AI at levels similar to the 45% estimated for the ‘New Economy’ back in 2000 (I can’t find the source, so take this with a grain of salt since this may be inaccurate). This includes semis, hyperscalers, infrastructure, energy, and so on. Ultimately, AI is increasingly prominent among investors and in the market. However, AI isn’t the purpose of this post... back to crypto.

Coincidentally or not, the ‘Internet’ market cap of $2.8 trillion in today’s dollars closely matches the total crypto market cap of $3.1 trillion. This includes Bitcoin at about $1.8 trillion, Ethereum at roughly $357 billion, and stablecoins at $313 billion. Approximately 79% of crypto’s total market cap is concentrated in BTC, ETH, and stablecoins, with the remaining 21% held by all other crypto, including altcoins and tokens.

Sources: CoinGecko, Ottavi

However, crypto’s market cap is apples-to-oranges with the Internet’s market cap back during Dot-Com. Firstly, BTC accounts for 57% of crypto’s total market cap, but has solidified a position as a “digital gold”—and therefore, a commodity—rather than equity so should be removed from this calculation (however, the market caps of publicly traded BTC miners who generate revenue and earnings from these activities are included in “Crypto Equities”). Ethereum is included given its infrastructure generates revenue and the value of ETH is tied to this financial activity.

Instead of using the current value of stablecoins, which are tokenized treasuries not equities, it would be more comparable to include the market cap of stablecoin issuers who generate their revenue and free cash flow from stablecoins. Here, the current value of stablecoins would be ~$522 billion versus the $313 billion stablecoins in circulation. This assumes Tether’s recent $500 billion private valuation is correct (for this purpose, we’ll assume it is) and Circle’s (CRCL) publicly traded market cap of ~$22 billion. Tether’s USDT and CRCL’s USDC dominate the global stablecoin market, accounting for ~84% of the stablecoin market cap. Even assuming Tether’s $500 billion valuation is accurate, including the entirety in this calculation may be too generous given that much of Tether’s value is in its USDT stablecoin, which is not GENIUS Act-compliant (we’ll need to see how its GENIUS Act-compliant USAT scales). Still, for now, I’m including Tether in this calculation. As such, the combined market cap of $522 billion for both is a good proxy for the current stablecoin equity market cap.

Additionally, if we assumed market caps for other publicly listed crypto stocks (Coinbase, Figure, Galaxy, Bullish, the DATs including MSTR, and even Robinhood), we would arrive at another ~$310 billion in value. In total, even including all other altcoins in this figure, this would put the “true” crypto market cap at ~$1.8 trillion and well below the $3.1 trillion commonly cited – or ~2.7% of the total US stock market vs. ~4% for ‘Internet’ market cap in 2000.

Sources: CoinGecko, Bloomberg, Ottavi

Ultimately, no matter how you slice it, I think the takeaways are similar:

• Crypto’s market cap is still relatively small when compared to other asset classes at ~$1.8-3.1 trillion.

• From a market sense, crypto still is very much BTC and until the market cap from non-BTC eclipses BTC, many market participants will view BTC and crypto’s success as interchangeable with each other.

• … and since BTC is the majority of the crypto market, other crypto assets tend to be correlated with BTC. Not helping matters are some of the marquee publicly traded crypto champions (e.g., COIN) generate the majority of their revenue and profit from trading crypto like BTC.

This also creates a dynamic in which BTC prices can induce reflexivity across cryptocurrencies. Like Wang’s observation that reflexivity was at play with software and could be at play with crypto—at least the way the industry is currently constructed where crypto asset values—or one’s views of how the industry is doing—is tied to Bitcoin’s price action.

Additionally, if we compare the recent price action in crypto, it fails to match up to Dot-Com. BTC is down ~29% from its year-to-date high, but only down 8% year-to-date and up 100%+ over the past two years. Furthermore, the market cap of “altcoins,” which CoinGecko defines as everything ex-BTC is only down 15% year-to-date and 24% when we exclude the rise in stablecoins, which are up 48% year-to-date. There is undoubtedly pain beneath the surface in many tokens and memecoins, but overall, the market is down, especially when compared to the stock market, but far from Dot-Com levels of pain. Although, much of this value destruction gets the bulk of the focus.

This highlights an important aspect regarding the current dynamic in crypto. Much of the value creation to-date has been in a “store of value” asset in BTC, for crypto to get past being more of a trading instrument to entities with enduring free cash flows, it needs to be leveraged as infrastructure to create better, faster, cheaper products and services. Potentially, the rise in stablecoin market cap growth this year while coins decline is a healthy indicator if stablecoins and other blockchain-based infrastructure can result in better real-world use cases.

Santos made this point in his post:

I think in probabilities. My highest-confidence view is that most businesses adopt crypto over the next 15 years to remain competitive. When that happens, crypto exceeds $10 trillion in aggregate value. Stablecoins, tokenization, users, and on-chain activity will grow exponentially. At the same time, valuations will reset. Large incumbents will fall. Business models that never made sense will be exposed.

This is healthy. It is necessary.

Crypto should become invisible. The more a company makes crypto the product, the more fragile it becomes. The durable winners will bury it inside workflows, payments, and balance sheets. Users should not notice crypto itself but feel the impact of faster settlement, lower costs, and fewer intermediaries.

Crypto should be boring.

The era of airdrops, subsidized demand, broken incentive loops, and hyper-financialization is ending, as it always does when capital tightens.

So, what about crypto’s future?

While this outlook may seem pessimistic, I believe crypto stands at an inflection point. Regulatory clarity—the GENIUS Act passed, CLARITY Act slated for early 2026—paired with stablecoins as superior rails for certain use cases (cheaper, faster, better) has accelerated institutional adoption. It’s early, but adoption is happening and will likely accelerate.

SEC Chairman Paul Atkins recently forecasted much of finance moving onchain in the next few years. This aligns with BlackRock’s Larry Fink and Rob Goldstein’s op-ed in The Economist touting tokenization’s potential to revolutionize finance, alongside traditional firms pledging—and investing—capital in this onchain shift (e.g. ICE’s $2 billion stake in Polymarket).

Blockchain is on the cusp of enabling superior products and redesigning financial infrastructure. Amid coin-centric focus, now is ideal to spot firms leveraging blockchain to create fundamentally superior products and services. Incumbents will adapt and thrive or get disrupted; challengers will win or lose. Ultimately, it will come down to what determines long-term fundamental value—creating superior products and services that produce enduring revenues and cash flows.

Revisiting ecommerce: Myriad players drove global penetration to 20%+ today, fueling trillions in annual sales. Financial services—similarly, a tens-of-trillions market globally—could follow suit. If blockchain achieves comparable adoption by 2050 through superior solutions, it could unlock trillions in equity value. Ecommerce birthed Amazon and other category winners while also helping presumed losers like Walmart continue to grow (now trading at a higher P/E than the disruptor). If this roadmap holds, we’re in the nascent stages of more enduring value creation via blockchain commercialization. Exciting times ahead.

A Quick Primer on Hotel Franchisors

In the US, Marriott (MAR), Hilton (HLT), Choice (CHH), Wyndham (WH), InterContinental (IHG), and Hyatt (H) collectively account for ~58% of the hotel market, measured by total rooms. Over time, the “large 6” have consolidated large brand portfolios through acquisitions and organically creating new concepts. When you stay at a hotel in the US, there’s a greater than 50/50 chance it’s one of the brands owned by these large brands.

Source: STR; as of November 2025

CoStar’s STR, considered the definitive data source for tracking US hotel trends, classifies hotels into six primary chain scale segments for branded properties: Luxury, Upper Upscale, Upscale, Upper Midscale, Midscale, and Economy (with Independents as a separate category). These segments are grouped primarily by the chain’s actual global system-wide average daily rate (ADR) from the previous year, ranking from highest (Luxury) to lowest (Economy). STR updates assignments annually based on ADR performance and other factors. Average ADR for Economy year-to-date through November 29 is $71 versus $160 for the total US and $442 for Luxury.

Source: STR; year-to-date weekly average through November 29, 2025.

Strategically, major players use a multi-brand portfolio approach to maximize their market penetration. Companies like Marriott and IHG cultivate a diverse collection of brands to serve different price points and travel preferences. This allows them to deepen their relationship with geographically focused franchisees, who can develop multiple non-competing properties in the same market (e.g., a Hilton franchisee with a Hampton Inn and a Home2 Suites). Furthermore, this breadth is critical for driving engagement in their massive loyalty programs, ensuring they can capture a traveler’s business regardless of destination or budget.

*above brand segmentation may not be 100% accurate.

Hotel franchisors, such as MAR, HLT, and WH, primarily operate an asset-light business model by licensing their established brand names, trademarks, operating systems, and technology platforms—including central reservation systems, revenue management tools, loyalty programs, and distribution channels—to independent hotel owners (franchisees). In exchange, franchisees pay a relatively small initial franchise fee upon signing the agreement to cover application, inspection, and setup costs, followed by ongoing fees that typically total 8–10%+ of the hotel’s revenue.

The core recurring revenue for franchisors comes from royalty fees, usually 4–6% (or up to 10% for premium brands) of gross rooms revenue, sometimes extended to food and beverage or other sources; additional contributions include marketing/advertising fees (2–4%), reservation system fees, and loyalty program assessments. This structure aligns incentives: franchisors earn more when franchisees achieve higher occupancy and rates across the brand’s ecosystem, while enabling rapid global expansion without capital-intensive property ownership or operational risk.

Franchisees typically sign long-term agreements—often 15-20 years, resulting in a highly retentive customer base. MAR and HLT are best-in-class, with annual room churn is ~1%, while WH’s lower-end segment mix results in a higher, albeit still low, annual churn rate of ~4%. This, plus franchisors’ mostly take rate model, means large brands enjoy a highly recurring revenues base that is indexed to growth in the travel market—a sector that has structurally grown at or above nominal personal consumption expenditures (PCE) over the last 10-20 years. Although franchisors’ revenue from franchisees is contractual, the franchisor’s room revenue tends to be cyclical, driven by macroeconomic factors.

Since franchisees are responsible for the capital outlay and operating the hotel, the franchisor needs to expend little incremental capital to grow revenue and profit, resulting in high EBITDA margins and free unlevered free cash flow conversion. EBITDA margins typically exceed 50%, with larger-scale players—MAR and HLT—generating fee-based EBITDA margins significantly above this. Required capital expenditures consist of software costs (typically in the low- to mid-single digits) and “key money.” Key money is an upfront financial contribution by the franchisor to the owner as an incentive to secure a long-term franchise agreement. This payment, often used in competitive markets to gain brand presence in strategic locations, helps fund renovations, development costs, or the owner’s capital stack, and is typically amortized over the agreement term or requires partial repayment upon early termination. Key money varies more by franchisor. HLT is at the low end, with total key money likely to be at ~3% of fee-based revenue in 2025, while MAR and WH are trending toward 7-8% of fee-based revenue. Even after accounting for these capital outlays, unlevered free cash flow conversion can exceed 80%. Altogether, franchisors’ highly cash generative business models results in returning nearly all—or more than all—of their free cash flow to shareholders through buybacks and dividends.

The large 6 franchisors weren’t always asset-light. In fact, while most generate most of their EBITDA from franchise fees today, this was not the case a decade ago. Each has spun off business units (timeshare, REITs) or sold owned assets to simplify the story and create a more predictable EPS/free cash flow per share story.

The combination of growing units, indexed to travel, high free cash flow conversion, and capital returns yields a free cash flow per share algorithm that’s attractive and has historically generated compelling shareholder returns. Over the past five years, only WH and CHH trail the S&P 500.

Source: Bloomberg

At the core is a 6-9% increase in room-based fees from 4-6% net unit growth and 2-3% revenue per available room night growth (essentially, end-market growth). Mix-shift to franchisees with higher royalty rates and ancillary fees (think, co-brand credit card fees), can push total fee-based revenue growth to 8-11%. With slight operating leverage and ~100% of free cash flow returned via buybacks and dividends, publicly traded franchisors typically reduce shares outstanding by 3-5% per year, depending on valuation and the free cash flow return ratio; the remainder is returned via dividends. This delivers 12-15% sustainable EPS / free cash flow per share growth, with minimal disruption risk and a slightly higher total shareholder return when dividends are included.

2025: The Year of Bifurcation

The U.S. travel market in 2025 was characterized by volatility and a stark divergence in performance (as highlighted in June and November). A clear divide emerged between the high-end luxury segment, which continued to thrive, and the lower-end economy segment, which faced significant and persistent headwinds. This created a challenging landscape for hotel companies with more exposure to the lower end, like WH and CHH.

The lower-end of the market has been contracting for most of the period since 2023. The initial weakness stemmed from lapping the exceptionally strong “revenge travel” period that followed the pandemic, compounded by the expiration of COVID-related government stimulus which disproportionately supported lower-income consumers. This environment was briefly interrupted in late 2024 and early 2025, when displacement from hurricanes Helene and Milton created a temporary surge in demand for economy and midscale lodging in the Southeast. However, once this tailwind faded, the underlying softness returned.

In 2025, this divergence became extreme. Luxury hotels posted approximately 5% growth in Revenue Per Available Room (RevPAR), a key industry metric calculated by multiplying a hotel’s average daily rate (ADR) by its occupancy rate. In stark contrast, Economy properties saw RevPAR decline by about 3%, creating an 8-percentage-point performance gap. This spread has recently widened to a staggering 11 points. Underscoring the severity of this downturn, the Economy segment has now endured 35 consecutive weeks of year-over-year RevPAR decline since March 2025.

The bifurcation between higher-end and lower-end consumer has been seen elsewhere in the economy, driven by increasing wealth effect on the high-end and low relative real income growth on the lower-end. This bifurcation has also been reflected in consumer sentiment surveys.

Sources: Morning Consult, Goldman Sachs Global Investment Research

To put a finer point on the duration of the lower-end weakness, the graph below compares year-over-year RevPAR growth from the first negative week during the Great Financial Crisis and from the recent period starting Mar 2023. The GFC was a more severe contraction but the duration now well surpasses even the GFC. The graph shows weekly growth returns in Q4 2024, but as pointed to above, this was mostly due to displacement from hurricanes. Looking at it on a 2-year basis since Sept, 10/11 weeks have been negative—indicating that, excluding the hurricane bump, trends would’ve remained weak last year. This was further impacted by the government shutdown, which is still being felt a month after it concluded.

Sources: STR, Ottavi

The Case for a 2026 Rebound

The investment case for the broader lower-end lodging segment in 2026 is centered on a potential cyclical recovery. The thesis is built on the idea that 2025’s persistent headwinds will become easier comparisons to lap, while emerging economic tailwinds could provide a much-needed boost to the core economy-segment consumer.

Several potential catalysts could drive improved performance in 2026. The year will benefit from lapping the “death by a thousand cuts” environment of 2025, which saw numerous small but cumulative negative impacts (Doge, tariffs, government shutdown, etc.). More concretely, consumers are expected to receive higher tax refunds in the first half of 2026, which could directly stimulate travel demand among lower- and middle-income households. Forecasts from Goldman Sachs suggest a stabilization or improvement in real income growth for the cohorts most critical to Wyndham’s customer base. While the lowest income cohort’s growth is expected to slow, all other cohorts are projected to see stable to accelerating real income growth.

Income ranges estimated based on 2022 BEA data in current dollars.

Sources: US Bureau of Economic Analysis (BEA), Goldman Sachs Global Investment Research

Additionally, 2026 will benefit from several favorable calendar shifts, America’s 250th birthday celebration and the World Cup in June and July. Marriott estimates that the World Cup could increase its global RevPAR by 30-35bps, which could imply a ~40bps growth tailwind in the US.

Most consensus estimates have the overall US lodging market growing ~1% in 2026 with Luxury outperforming while Economy lags the overall industry. While plausible for the market overall (I think the likely outcome is higher), the Economy segment could outperform materially if stimulus drives demand growth. From the beginning of Q2 through November, Economy RevPAR is down an average of ~5% year-over-year. Typically when lodging segments come out of this type of contraction, they don’t just grow 1%, they overshoot to the upside. In the 52 weeks following the last negative weekly print after the GFC, the overall market and Economy segment grew 7% and 8%, respectively, after declining 13% in the prior 52-week stretch. By no means is this meant to compare the current period to the GFC, but to show a -5% for three-fourths of 2025 is unlikely to produce 1% growth in the following year.

Ultimately, there will be surprises—positive and negative—as there tends to be in any given year, but 2026 is setting up to be a recovery year after 2025’s weakness. This should benefit all lodging chain scales, with total RevPAR likely to return to low-single-digit growth. Forward-looking data remains weak until mid-February when the hurricane comps are lapped and easier comps from Doge and Liberation Day begin. Hotel booking windows tend to be relatively short (~30-45 days) but the data is constructive to a better H1. There’s consensus that Luxury’s strength will continue as long as wealth effects persist from a strong stock market, but is this the year that…Economy outperforms? It seems plausible. If so, it begs the question how this could translate into stock performance for WH and CHH who are more exposed to the low-end of the market.

Implications for Lodging Stocks

Hilton (HLT), Marriott (MAR), Intercontinental (IHG), and Hyatt (H) are all positive on the year—with only HLT beating the market ever so slightly (15.8% vs. S&P 500 at 15.6%). WH and CHH are down 22% and 35%, respectively, trailing HLT and the market by ~36% and ~50%, respectively. The entire group has rallied in the past ~week.

Source: Bloomberg

HLT trades at a premium to the peer set due to its best-in-class net unit growth with guidance at 6-7% for 2025 and for the foreseeable future and higher free cash flow conversion from lower key money relative to peers. While HLT trades at a premium, MAR/H/IHG trade more tightly, while WH and CHH more decoupled from the pack in 2025.

Source: Bloomberg

The further decoupling of WH and CHH in 2025 was due in part to a bifurcation in RevPAR growth, which widened adjusted EPS growth dispersion relative to typical levels (in addition to long-term concerns about HLT and MAR moving more into their core segments). ~50%+ of WH and CHH’s total rooms are from the US lower-end segment (Economy and Upscale). Compared with HLT, MAR, and H, the Economy segment has very little exposure, even if both are expanding more into WH and CHH’s segments. Furthermore, both benefited more significantly from stronger international RevPAR trends. This meant that even in a relatively choppy lodging market, RevPAR for those not named WH or CHH is expected to finish 2025 flat to up ~2%, while WH and CHH declined by 3% and 2%, respectively.

This resulted in WH and CHH revenue lagging peers, likely only finishing up 0-2% with EPS similarly trailing.

*Hyatt is omitted given sales from assets materially impacted adj. EPS.

Making the Case for Wyndham

Going into 2026, much of the peer set is positioned to grow EPS in the double-digits, except for CHH which has the headwind for much more lackluster net unit growth holding it back. If HLT/MAR/H/IHG hold their current multiples, these could produce low-to-mid-teens returns in 2026. Alternatively, one could paint a picture of accelerating growth for WH/CHH, which both trade at a discount to their historical multiples and relative to peers.

However, due to a better outlook for net unit growth (~4%), WH has a much more straightforward case to accelerating adjusted EPS to at least the low-teens, which would be 5-8 percentage points higher than 2025 and more in line with peers, including HLT.

Assuming no multiple expansion, WH could produce double-digit returns. A more optimistic scenario could see WH trading back to the historical discount versus HLT. If we assume a 25-35% discount versus HLT on its NTM EV/EBITDA multiple, the WH’s multiple could expand from ~11x EBITDA to 12.5-14.5x, which would add an additional ~14-30 percentage points to the stock’s return in 2026—making it potentially the most asymmetric stock in the group.

*This is meant to be illustrative, not a recommendation—use your own assumptions and analysis.

Source: Bloomberg

Stock returns for MELI and SE have lagged the broader tech indices materially since MELI cut its free shipping threshold in early June from R$79 (~$15) to R$19 (<$4). MELI’s underperformed by 37%, while SE has lagged by 35%. The graph below shows SE’s and MELI’s absolute performance and relative performance to QQQ.

Source: Bloomberg

The divergence between the market was more pronounced initially for MELI and has persisted through November, while SE’s underperformance occurred more recently, following revisions to Shopee’s margin estimates. MELI’s reaction was a direct response to SE’s competitive threat in Brazil, while SE’s more recent selloff is partially due to a more competitive Brazil and to investments SE is making in its core Southeast Asia markets.

MELI and SE’s recent stock underperformance has sparked a debate on X. While most likely aren’t looking for another opinion on the topic, I wanted to add a more nuanced perspective. A few thoughts:

1. MELI remains the scaled #1 player in Brazil across ecommerce and financial services. I suspect this will remain true for the foreseeable future.

2. SE’s Shopee has done an admirable job of scaling in an area of the market that MELI ignored (lower-income consumers, low average order values). It has used its strategy to scale a business that’s #2 to MELI in GMV and revenue, and #1 in metrics, like total active users and orders. SE’s built enough of a business with this focus that it is here to stay long-term and is now trying to move up-market into MELI’s core business.

3. SE’s momentum and scale were sufficient for MELI to fight back, including impairing its unit economics by lowering its free shipping threshold materially.

4. #3 being true does not mean #1 or #2 cannot also be true. Brazilian ecommerce is consolidating into a duopoly, and MELI and SE are both growing profitably. The market still remains under-penetrated with estimates of only a mid-teens percentage of the market conducted online. There will be room for both MELI and SE to grow in the foreseeable future. However, there are legitimate questions about MELI BZ’s profit margins going forward if SE continues to gain overall market share.

Signal vs. Noise: Direction of the Fundamentals

It can be tempting to look at a stock’s underperformance and state that the market is being irrational. However, a stock’s performance needs to be put into the context of how new information has impacted the market’s expectations for future free cash flow generation. Since May, consensus estimates for MELI Brazil’s 2026 contribution profit are down by 34% due to the reduced free shipping thresholds and investments in sales and marketing to drive stronger top-line growth.

Source: Bloomberg

In 2024, Brazil accounted for 46% of MELI’s total contribution profit (before accounting for corporate costs and other overhead). Unsurprisingly, as MELI BZ’s profit estimates got cut, so too did overall estimates. MELI’s 2026/2027 consensus adjusted EPS estimates peaked in early June and have been in a downward trajectory since, coinciding with the free shipping reduction. Since then, MELI’s 2026/2027 adjusted EPS consensus estimates are down ~13% and ~10%, respectively (the graph below overlays MELI’s stock price with changes to 2026/2027 adjusted EPS estimates). Importantly, this is the first time since SE entered the Brazilian market in 2019 that forward-looking estimates have been revised downward due to SE’s risks. When put into this context, it’s unsurprising that MELI’s stock has underperformed over the last several months.

Source: Bloomberg

The incremental growth shows why MELI is taking margins down. From 3Q24-2Q25, Shopee’s incremental growth was approximately at par or above MELI’s. While MELI’s overall market share is materially larger than Shopee’s (estimated at ~2x), the incremental growth painted a potentially weaker outlook. However, 3Q25’s incremental growth reverted in MELI’s favor for the first time in 5 quarters, showing that the actions are helping MELI grow incremental share again.

Sources: MELI and SE public filings, Ottavi.

*Note: GMV is calculated in USD. Shopee Brazil’s GMV are Ottavi estimates given it is not reported by SE.

Additionally, SE’s credit expansion (through its Monee unit) likely raised more concerns at MELI, prompting it to act now. In emerging markets like Brazil, ecommerce and financial services are synergistic. Many Brazilians pay for ecommerce purchases with short-term loans due to lower disposable income and high inflation rates. On the back of securing a Sociedade de Crédito Direto (SCD) license in August 2025, which allows it to lend with the company’s own funds, Monee’s loan growth was 3x year-over-year in 3Q25, after growing 2x year-over-year in 2Q25. It’s still a small fraction of MELI BZ’s loan book, but is incrementally scaling quickly.

Perhaps the most telling graphic is the ~2-year downward trend for MELI BZ’s contribution margins after peaking at 28% in 3Q23. In 3Q25, MELI Brazil’s margins contracted further to 12%, down more than 370bps quarter-over-quarter, 550bps year-over-year, and nearly 1,600bps since the 3Q23 peak. There are multiple factors at play but it seems clear that MELI is responding to the growing threat from Shopee.

Sources: MELI public filings, Ottavi.

Closing Remarks

Are 3Q25 margins the floor and do contribution margins stabilize from here? Will SE successfully move up-market and put additional pressures on MELI’s margins? Will MELI’s actions use the might of the scale it’s built and force SE to choose between growth and profitability? Or do MELI and SE grow rationally with both generating profitable growth with margin expansion? Ultimately, MELI’s stock performance over the next 1-2 years will at least partially depend on answers to these questions and how they impact the direction of the lines above.

As written in June, travel demand showed signs of softness heading into the summer. It strengthened steadily through July and August and accelerated after Labor Day as the U.S. entered the peak corporate travel season. Recently, this momentum has reversed, with the trailing 28-day average turning negative year-over-year. This dip is likely tied to flight-capacity reductions stemming from the government shutdown rather than a true demand deterioration. If that’s the case, TSA throughput should rebound in the coming weeks as the holiday travel season ramps up.

Assuming this recent weakness proves transitory, the rebound since July has been far from broad-based. The U.S. economy continues to display a “K-shaped” pattern: higher-income consumers are spending freely, buoyed by rising wages and wealth effects from strong equity markets, while lower-income households remain pressured by persistent inflation and stagnant real wage growth. This divide is clearly visible in the hotel sector.

Across the industry, U.S. revenue per available room (RevPAR) has hovered within a narrow –1% to +1% range for much of the year. But performance at the ends of the price spectrum has sharply diverged. Luxury hotels have posted roughly 5% RevPAR growth in 2025, while Economy properties are down about 3%—an 8-point gap. That spread has widened to 11 points more recently as Economy demand has continued to deteriorate. For context, the Luxury–Economy gap averaged just 2 points from 2010 through February 2020. Economy hotels have not generated year-over-year growth since March—33 consecutive weeks of contraction.

International travel has also weighed on overall demand in 2025, though its impact is smaller given foreign visitors represent only a mid-single-digit share of total trips. Total foreign arrivals are down 4% year over year, driven largely by a sharp pullback from Canadians. Statistics Canada reports that return trips to the U.S. from February to October fell 21% by air and 34% by car. Canadians accounted for roughly a quarter of all foreign visitors to the U.S. in 2024, making them responsible for much of the inbound decline. Notably, foreign travel to the U.S. rarely contracts outside of economic downturns.

Most publicly traded travel companies, including airlines and hotel chains, have struck an optimistic tone about the 2025 summer season after the initial pause in the spring from tariff-induced uncertainty. Much of this commentary focuses on stronger trends in April, following a weaker March, and general optimism that the shorter booking windows they’re seeing will not ultimately translate into materially weaker revenue. However, early indicators point to a slower start to summer travel.

The WSJ recently published an article citing that foreign arrivals at the US’s main airports were down 6% YoY (as of the end of May). These figures included the tail-end of tailwinds from Easter occurring later this year versus 2024. Since Easter, international arrivals have worsened. Non-US passport holders entering 12 of the country’s largest airports is -10% on a trailing 14-day rolling average, with the trailing 7-day average at -13% (as of June 6). For May, non-US passport holders entering 12 of the country’s largest airports were down 6.6% year-over-year, in line with the WSJ figure.

Year-to-date, international travelers to the US are down more than 3% year-over-year, which would be the first decline, excluding COVID-impacted 2020, since 2016 and the worst drop since 2009. 2021-2023 growth benefited from the post-Covid recovery, but international travel demand to the US typically grows unless there's a demand shock / recession (GFC, 9/11, dotcom bubble, Covid), which the current data does not support.

Weakness in international demand is unsurprising. It’s been well-documented that Canadians are foregoing US summer travel in favor of domestic or other foreign markets. Data from Statistics Canada, cited by the WSJ, shows that in April, the number of Canadians traveling to the US by air and car was down 20% and 35% year-over-year, respectively. Furthermore, Canadian summer flight bookings to the US are down by 22%. The US Travel Association estimates Canadians accounted for ~25% of all visitors to the US in 2024 so declines of 20%+ are likely having a material impact on international inbound demand.

In normal times, international travel provides a tailwind to total US travel, especially in the summer months when travel over-indexes to leisure, but it still only accounts for a mid-single-digit percentage of total trips in the US annually. A 50bps impact to US travel demand (assuming 5% of total trips and the current -10% YoY trends hold throughout the summer) would generally be offset by domestic travel growth. However, domestic travel is also showing weakness. Average daily US travelers going through TSA checkpoints – a proxy for travel demand – were down 1.8% in May – after ranging from 0-2% in four prior months to start the year. If we assume international accounted for 5% of total demand, then the domestic growth was also negative in May.

The June data thus far is showing continued softness, with the 14-day and 7-day averages both down 4%+ as of June 10.

Similar to international travel, domestic travel demand typically grows absent economic contraction. Revenue passenger miles (RPMs) – a proxy for air travel demand – have only gone negative three times in the past 25 years (9/11 / dotcom bubble, GFC, and Covid). Unlike the TSA data which only dates back to 2020, RPM data can be analyzed over decades. The TSA and RPM data are correlated, and can be used relatively interchangeably as an air travel demand proxy. The TSA data informs us that RPMs growth likely went negative in May, with continued declines so far in June.

Hotel data is also unsurprisingly weak. Revenue per available room – or “RevPAR” – which calculates the amount of revenue a hotel generates each night, is down 2.5% for the first two “unofficial” weeks of summer, starting with Memorial Day and has been flat year-over-year since April. Like air travel, the US hotel industry typically grows absent economic contractions, usually in line with or better than nominal GDP. Lodging has trailed nominal GDP since early 2023, with growth diverging more in Q2 (my Q2 estimates assume the last 3 weeks of the quarter resemble the current trends over the past month).

Trends started to weaken leading up to Liberation Day, with Delta pointing to softening trends in mid-March. Then there was a more protracted pause post-Liberation Day. Most travel companies have called out stabilizing-to-improving trends since April 2nd, which was reflected in the data – helped by the Easter calendar shift. What is more surprising is that this recovery hasn’t sustained, with data worsening over the last 2-4 weeks and at least some forward-looking data inform us that this may continue into the summer. STR (owned by CoStar), which tracks data for 87k+ hotels globally – and most of the hotels in the US – shows room nights on the books for July and August down in the low single digits year-over-year. The graph below is occupancy on the books for future dates as of the end of May. The data can be volatile from week to week, but July and August consistently show a negative trend.

A word of caution before over-extrapolating this data: for July and August, most hotels and vacation rentals typically have only 35-50% of their summer stays booked. And if the forward-looking data is mostly from shorter booking windows, sales could strengthen as the summer progresses. Shorter booking windows can occur when travelers push out booking travel until closer to arrival, usually from a drop in confidence about economic conditions. This last occurred in late summer 2024, but ultimately proved to be a blip after growth recovered in the fall, setting the stage for a strong Q4. The trends shown above may improve as we approach arrival. And the relatively better on-the-books June trends shown in the graphic could indicate demand will firm up closer to the arrival date in July and August.

Objectively, the current economic data favors a rebound later in the summer or early fall. Wage growth – while slowing – is still supportive of healthy PCE growth at ~4% year-over-year. Additionally, Q1 corporate earnings remained healthy. However, if growth trends don’t improve over the coming weeks, we’ll likely see travel companies start talking down Q2 numbers by July 4^th. More importantly, the longer the weak travel data persists, the louder the chatter of a more prolonged travel downcycle becomes.