ExpertsFMStripe’s 2025 annual letter
Cheeky Pint · February 24, 2026
5 Selected Lessons:
case · 3 framework · 1 priority · 1
What CEOs of mature businesses should prioritize for growth
Source
Cheeky Pint — Stripe’s 2025 annual letter (February 24, 2026) · at 14:55 ▸
When this applies
when any leader of an established organization is allocating scarce attention/budget between flashy strategic bets and unglamorous operational fixes
The takeaway
Rank near-guaranteed operational fixes above high-prestige, high-variance strategic bets when both are available as growth levers.
Why it works
Marketing campaigns carry high variance and uncertain payoff even when critically acclaimed, while payments/conversion optimization recovers quantifiable, already-existing demand with near-certain payoff.
One level up
Prioritize certainty of return over prestige of initiative when allocating scarce growth resources.
Where it stops holding
This stops working when the operational layer is already optimized and no further gains are available, making speculative bets the only remaining growth lever.
Detail
hierarchy:
- 1. Fix payments/checkout leaks (auth rates, local payment methods, fraud prevention) — near-guaranteed ROI
- 2. Pursue speculative demand-generation bets (advertising campaigns) only after the operational layer is optimized
rationale: Splashy campaigns can win awards and still fail to move revenue; payments optimization has repeatedly, measurably moved revenue across multiple named companies.
In their words
So CEOs, splashy ad campaigns are good fun, but don't overlook the revenue growth right under your nose. Your heads of payments almost definitely deserve more recognition.
Five levels of agentic commerce
Source
Cheeky Pint — Stripe’s 2025 annual letter (February 24, 2026) · at 18:18 ▸
When this applies
when trying to stage the maturity of any AI-agent automation of a complex human task, from assistance to full autonomy
The takeaway
Break an 'autonomous AI' vision into discrete levels defined by which specific human burden is removed at each stage, rather than assuming a single leap to full autonomy.
Why it works
Each level removes a specific category of friction or decision-making burden from the human (form-filling, then keyword search, then re-explaining context, then choice-making, then even prompting), requiring progressively more trust and system capability.
One level up
Overhyped technology visions should be decomposed into incremental capability stages to track real progress against hype.
Where it stops holding
This stops working when the task lacks natural sub-stages of delegation (e.g., single-step purchases) or when trust/liability concerns prevent users from ever ceding higher-level decision rights to an agent.
Detail
procedure:
- Level 1: Eliminate web forms — agent fills in payment/shipping details on command, human still decides what to buy.
- Level 2: Descriptive search — human describes a situation/need rather than keywords, agent reasons across many variables to surface options.
- Level 3: Persistence — agent retains preferences/history so the human doesn't reintroduce context each time; human still confirms each purchase.
- Level 4: Delegation — human sets a goal/budget constraint; agent searches, evaluates, and purchases on the human's behalf.
- Level 5: Anticipation — agent proactively completes tasks without prompting, notifying the human after the fact.
benchmarks: {"current_industry_position":"hovering around levels 1-2"}
rationale: Staging the vision prevents mistaking hype (level 4-5 utopian narratives) for current reality (level 1-2 capability).
In their words
But today, the industry is hovering right on the edge of levels maybe one and two.
New company cohorts starting global and growing faster
Source
Cheeky Pint — Stripe’s 2025 annual letter (February 24, 2026) · at 1:46 ▸
When this applies
when evaluating whether a new generation of founders/tools is producing structurally faster-growing companies
The takeaway
When the cost and time of building and distributing a product drops sharply, the resulting cohort of new entrants will show structurally different growth curves than prior cohorts, not just more of them.
Why it works
Lower friction to build (AI coding tools, instant incorporation, instant global payments infra) compresses the time from idea to revenue and to international scale.
One level up
Infrastructure improvements don't just add incremental users — they can shift an entire cohort's growth distribution.
Where it stops holding
This stops working when the underlying infrastructure improvements are only accessible to a narrow number of founders, or if the data reflects one unusually strong年 rather than a persistent trend (the letter itself flags this uncertainty).
Detail
situation: In 2025, more new companies joined Stripe than ever, 57% based outside the US.
decision: Stripe measured growth speed and revenue ramp of this cohort against prior years.
outcome: This cohort grew ~50% faster on average than 2024's, and twice as many companies hit $10M ARR within three months of launch.
lesson: A sharp drop in building/distribution friction (AI tools, instant incorporation, global payments) can shift the growth curve of an entire startup cohort, though whether this is a one-year anomaly or a new regime remains uncertain, per Stripe's own hedge.
In their words
This new cohort is by far the highest performing and fastest moving we've ever seen, growing around 50% faster than the 2024 cohort on average.
Fintechs ironically slowest to go global themselves
Source
Cheeky Pint — Stripe’s 2025 annual letter (February 24, 2026) · at 5:42 ▸
When this applies
when a company building infrastructure for others is itself constrained by the same frictions it's trying to remove for its customers
The takeaway
The company closest to a structural bottleneck is often the slowest to escape it, because it must build the very infrastructure others merely consume.
Why it works
Regulatory licensing, banking rails, and compliance burdens that are specific to financial services create a higher bar for financial companies to expand internationally than for software companies riding on top of existing financial infrastructure.
One level up
Infrastructure providers frequently inherit the frictions they exist to remove, and transcending them is harder from inside the stack.
Where it stops holding
This stops working when regulatory and licensing regimes genuinely converge across jurisdictions (e.g., via stablecoins or global accounts), which the letter notes is beginning to happen.
Detail
situation: Even the most successful fintechs (Chime, NuBank) expanded internationally far slower than consumer tech companies.
decision: Stripe itself admits its issuing product took 7 years to reach 22 countries.
outcome: A new cohort of fintechs (Sling Money, Dollar App, Felix, Cast) and Stripe's own Financial Accounts product are now launching in 100+ countries on day one, enabled by stablecoin-based borderless infrastructure.
lesson: Regulatory-heavy infrastructure businesses are structurally slower to globalize than the products built on top of them — until a new layer (e.g., stablecoins) removes the underlying friction.
In their words
Ironically, the tech companies most constrained by barriers of financial geography have tended to be fintechs themselves.
Shared payment tokens remove credential exposure risk for agentic payments
Source
Cheeky Pint — Stripe’s 2025 annual letter (February 24, 2026) · at 22:00 ▸
When this applies
when building any system where an automated intermediary needs to transact on a principal's behalf without holding or exposing the principal's sensitive credentials
The takeaway
When delegating transactional authority to an automated agent, use revocable, scoped tokens instead of sharing the underlying raw credential, to contain risk.
Why it works
A tokenized credential that can be passed to any processor (not just the issuing platform) lets an agent initiate a payment without ever holding or transmitting the raw credential, limiting the blast radius of compromise.
One level up
Design access/authority delegation around minimal, revocable scope rather than handing over root credentials.
Where it stops holding
This stops working when the receiving party cannot support token-based settlement and requires raw credential data, forcing a less secure fallback.
Detail
situation: AI agents need to make payments on behalf of users across many businesses and platforms, which otherwise would require either storing/exposing credit card credentials or building bespoke integrations per platform.
decision: Stripe introduced shared payment tokens, a portable credential businesses can forward to their own vaults or processors.
outcome: Agentic Commerce Suite and tokens are being adopted by major retail brands (Anthropologie, Urban Outfitters, Etsy, Coach, Kate Spade) and integrated into ChatGPT shopping and Microsoft Copilot.
lesson: Tokenized, scoped credentials are the right primitive for enabling automated agents to transact safely at scale.
In their words
We introduced shared payment tokens, a new payment primitive that lets agents initiate payments without exposing credentials.