Stripe Atlas: 130% More Startups in Q1 2026 — 5 Numbers That Show AI Is Creating Founders, Not Killing Jobs
Stripe Atlas hit 100,000 all-time incorporations with a 130% YoY spike in Q1 2026. The data suggests AI is minting entrepreneurs faster than eliminating roles.
130% More Startups in Q1 2026: What Stripe Atlas’s Numbers Actually Tell Us About the AI Economy
On May 1st, Stripe CEO Patrick Collison posted a number that got buried under the usual AI news cycle. Stripe Atlas — the service that helps founders incorporate new companies — just hit 100,000 all-time incorporations. More specifically, Q1 2026 was up 130% year-over-year. That’s not a rounding error. That’s a structural signal about what AI is actually doing to the labor market, and it cuts directly against the dominant narrative.
You’ve probably heard the story: AI is eliminating jobs, hollowing out entry-level roles, and building toward some version of mass unemployment. It’s a coherent story. It has smart people behind it. And the Stripe Atlas data suggests it might be, at minimum, incomplete.
Derek Thompson — Ezra Klein’s co-author on Abundance — put it plainly: “AI agents are better at creating firms than destroying jobs.” That’s a provocation, not a conclusion. But the data underneath it is worth taking seriously.
Here are five numbers from the current moment that, taken together, tell a different story than the one dominating the discourse.
The Stripe Atlas Number: 130% YoY, 100,000 All-Time
Start with the headline. Stripe Atlas incorporations up 130% year-over-year in Q1 2026. One hundred thousand companies incorporated through the platform since launch, with a significant chunk of that milestone arriving in a single quarter.
Coding agents automate the 5%. Remy runs the 95%.
The bottleneck was never typing the code. It was knowing what to build.
Stripe also shared that AI-sector startups specifically are showing faster revenue growth than historical norms. So it’s not just that more companies are forming — the ones forming in AI are monetizing faster than prior cohorts of startups did.
The obvious question is whether this is a leading indicator or a lagging one. Are these new companies the result of people who got displaced and turned to entrepreneurship out of necessity? Or are they people who saw an opportunity that AI made newly accessible — a business that would have been too expensive, too slow, or too technically demanding to build two years ago?
The honest answer is probably both. Greg Eisenberg, host of the Startup Ideas podcast, made the case that “millions will get laid off or never hired over the next 24 to 36 months” and that many of them “become entrepreneurs out of necessity at first, then out of opportunity.” That’s not a utopian framing. It’s a realistic one that still ends with more companies being built.
The more interesting version of the question is whether AI is actually expanding the total population of people who can be entrepreneurs — not just reshuffling the existing pool. The Stripe data doesn’t answer that definitively. But a 130% YoY spike is hard to explain as mere reshuffling.
Software Engineering Jobs: Up 18% From the May 2025 Inflection Point
The most AI-exposed occupation in the economy, according to most analyses, is software engineering. If AI were eliminating jobs in a straightforward way, you’d expect to see it here first.
Instead, Citadel Securities data shows software engineering job postings up 18% from an inflection point in May 2025. The Federal Reserve’s data puts software engineering employment at its highest level since November 2023. These numbers don’t prove that AI won’t eventually displace software engineers. But they do suggest that the displacement narrative is, at minimum, running ahead of the actual data.
The mechanism here is Jevons’ paradox, which University of Chicago economist Alex Immis applied to AI labor in an essay that’s been circulating widely. The short version: when a resource gets cheaper, consumption of that resource tends to go up, not down. When AI makes coding cheaper, you don’t use less coding — you build what was previously too expensive to justify building. As Merzmik Ahmed from Emergence AI put it: “Code is digital brick. If bricks get much cheaper and easier to lay, you don’t use fewer builders. You build what was previously too expensive, too slow, too bespoke, or too annoying to justify.”
This is also the context in which the GStack framework from Y Combinator’s Gary Tan makes sense — it’s a tool designed for solo developers to operate with the leverage of a full team, which is exactly what Jevons would predict: cheaper capability leads to more ambitious building, not less building. The same logic applies to the broader wave of AI coding tools reshaping what a two-person team can ship in a quarter. For more on how these tools are evolving, the Claude Code source code leak analysis surfaces eight specific capabilities that are already changing how developers work.
New College Grad Hiring: Up 5.6%, Youth Unemployment Down Nearly Half
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
The job apocalypse narrative has a specific demographic target: young workers, especially recent college graduates. The argument is that AI will eliminate the entry-level roles that have historically served as on-ramps into professional careers.
The current data doesn’t support that story. New college grad hiring is up 5.6% year-over-year. Unemployment for people aged 20 to 24 with a college degree has fallen from nearly 9% to almost 5%. That’s a significant move in the wrong direction for the doom narrative.
Anthony Pompliano, who explicitly changed his mind on this, wrote: “Previously, I believed AI would replace many entry-level roles typically filled by young employees. The data is saying something different, so when I get new information, I’m willing to change my mind.”
That’s the right epistemic posture. The data is what it is. And right now, the data says that the cohort most vulnerable to AI displacement is actually doing better, not worse.
Ezra Klein made a related point in his New York Times op-ed, citing ASU professor Eldar Maximov: “In every major occupation group that adopted computers heavily, employment grew faster than in groups that did not. Computers eliminated specific tasks within jobs, but the resulting cost reductions created so much new demand that the occupations expanded overall.” Klein’s own experience tracks with this. When he started his podcast, he was its only researcher. Now he has a team — not because AI replaced his researchers, but because the capacity to do more created demand for more. The AI agents for marketing teams use case follows the same pattern: agents handle the repetitive analytical work, which frees human team members to take on more strategic and creative roles rather than disappearing from the org chart entirely.
640,000 Jobs Created by AI Between 2023 and 2025
The Wall Street Journal published an analysis of LinkedIn job posting data that found AI created 640,000 jobs in the United States between 2023 and 2025. These are net new positions — head of AI, AI engineer, AI product manager, and a long tail of roles that didn’t exist in their current form before the current wave.
This number deserves some scrutiny. LinkedIn job posting data has known limitations. Not every posting represents a hire. And 640,000 jobs over two years is a relatively modest number against the scale of the US labor market, where roughly 5 million people are hired and 5 million leave or lose jobs every single month.
But the direction matters. The story that AI is a pure job-destroyer doesn’t survive contact with a number like this. The more accurate story is that AI is simultaneously eliminating certain tasks, creating new roles, and — through the Jevons mechanism — expanding demand in adjacent areas.
Other agents ship a demo. Remy ships an app.
Real backend. Real database. Real auth. Real plumbing. Remy has it all.
The harder question, which Ezra Klein raised and which doesn’t get enough attention, is about distribution. “A world where AI displaces 8 million workers might be harder to handle than a world where it displaces 80 million workers,” he wrote. Mass displacement forces systemic response. Partial displacement gets absorbed without triggering the policy interventions that would actually help the people affected. The China trade shock displaced roughly 2 million workers — small relative to the total economy, devastating for specific communities, and met with almost no meaningful policy response. That’s the model to worry about.
The AI agents for financial services sector illustrates the distribution problem concretely: back-office automation is accelerating in ways that affect lower-wage clerical roles disproportionately, even as front-office demand for AI-fluent analysts grows. The net job count may be neutral or positive; the distribution of who gains and who loses is not.
Anthropic ARR: From $9B to $44B in One Year
This one isn’t about jobs directly. It’s about what the underlying economics of the AI boom actually look like — and whether the entrepreneurship surge is landing in a real market or a speculative one.
SemiAnalysis, which is considered extremely well-sourced, published that Anthropic’s ARR has gone from $9 billion to over $44 billion in the current year. Analyst Ming Li did the math: Anthropic is adding roughly $96 million in ARR per day. For context, AWS took 13 years to reach $35 billion in annual revenue. Salesforce took over 20 years to pass $20 billion.
The mechanism driving this is the shift from seats to tokens. In 2025, the AI revenue question was about how many people would pay $20 or $30 a month for a subscription. The skeptics were right that this math didn’t justify the infrastructure buildout. What changed is that the unit of consumption is no longer a seat — it’s a token. A single developer using Claude Code or Codex isn’t worth $200 a year. They’re potentially worth hundreds or thousands of dollars a month to the companies selling the tokens.
This matters for the startup formation story because it means the market that those 130%-more startups are entering is real. The revenue is there. The demand is there. Morgan Stanley raised its five-hyperscaler CapEx forecast to $805 billion for 2026 and $1.1 trillion for 2027. The backlog of demand — roughly $1.3 trillion against roughly $400 billion in Q1 CapEx spend — is diverging upward. Larry Fink said it plainly at Milkin: “There is not an AI bubble. There is the opposite. We’re short power. We’re short compute. We’re short chips.”
The founders incorporating through Stripe Atlas aren’t building into a speculative void. They’re building into a market where demand is structurally outrunning supply.
The Infrastructure Behind the Founders
One thing the Stripe Atlas number doesn’t capture is what it now takes to actually build a company. The tooling available to a solo founder or a two-person team in 2026 is categorically different from what existed even two years ago.
The AI agents for product managers use case is a good example — tasks that used to require dedicated headcount (user research synthesis, competitive analysis, spec writing) can now be handled by agents that run continuously. That’s not just productivity improvement; it’s a change in the minimum viable team size for certain types of companies.
Remy is new. The platform isn't.
Remy is the latest expression of years of platform work. Not a hastily wrapped LLM.
On the development side, tools like Remy take a different approach to the build layer entirely: you write a spec — annotated markdown where prose carries intent and annotations carry precision — and the full-stack application gets compiled from it. Backend, database, auth, deployment, all of it. The spec is the source of truth; the code is derived output. For a founder who wants to ship a production app without a dedicated engineering team, that’s a meaningful change in what’s possible.
The agent infrastructure layer is also maturing fast. MindStudio handles the orchestration problem that stops a lot of AI projects from reaching production: 200+ models, 1,000+ integrations, and a visual builder for chaining agents and workflows. For a founder building an AI-native product, that’s the difference between spending months on infrastructure and spending weeks on the actual product.
These tools are part of why the Stripe Atlas number is plausible. The barrier to incorporation has always been low. The barrier to building something that actually works has historically been much higher. That gap is closing.
What the Numbers Add Up To
The five numbers here — 130% startup formation growth, 18% increase in software engineering job postings, 5.6% rise in new grad hiring, 640,000 AI-created jobs, and Anthropic’s ARR trajectory — don’t add up to “everything is fine.” They add up to something more complicated.
The displacement narrative isn’t wrong. It’s incomplete. Certain tasks are being automated. Certain roles will shrink. The transition will be uneven, and the communities that get hit hardest are unlikely to receive the policy support they’d need. Ezra Klein is right that partial displacement is often crueler than mass displacement, precisely because it doesn’t force a systemic response.
But the creation narrative is also real. AI is minting entrepreneurs. It’s expanding the demand for software engineers even as it automates software engineering tasks. It’s creating new job categories faster than the old ones are disappearing, at least so far.
The Stripe Atlas data is the most interesting piece of this because it suggests something that’s easy to miss in the doom-and-creation debate: the unit of economic participation might be changing. The question isn’t just “will there be enough jobs?” It’s “will the job be the right unit?” If AI makes it genuinely easier to build a company than to find a job — and the 130% number suggests we might be moving in that direction — then the labor market conversation needs a different frame entirely.
For now, the data says AI is creating founders faster than it’s eliminating jobs. That could change. But it’s where we are.
