The archive · AI & Models · Product decision · 2025–2026
Simile's synthetic-user bet: $200M Series B at $2B five months after Series A
Simile simulates how customers will behave before companies commit budget; Greenoaks led a $200M Series B at $2B, 5 months after a $100M Series A.
Simile
What the business is
Simile builds AI foundation models that simulate human behavior, letting companies test how customers, patients, or employees will respond to a product, campaign, or policy before committing budget.
How it started
Joon Sung Park, a Stanford PhD graduate advised by Michael Bernstein and Percy Liang, built Generative Agents, a project where 25 AI agents lived simulated lives in a virtual town called Smallville - the paper won Best Paper at UIST 2023. In late 2025 he co-founded Simile with Bernstein, Liang, and Lainie Yallen, spun it out of Stanford, and launched publicly in February 2026 with a $100M Series A led by Index Ventures.
What happened
Simile raised $200M in a Series B led by Greenoaks at a $2B valuation in July 2026, five months after the Series A, with Index doubling down and Hanabi, Bain Capital Ventures, A*, Factory, and CVS Health Ventures participating. Revenue grew fivefold between the February launch and July, headcount passed 50, and customers CVS Health, Wealthfront, Deloitte, and Gallup ran tens of millions of simulations. Simile also shipped a confidence model that scores each simulation's reliability, a direct answer to the calibration critique of synthetic-user research.
How it ended up
Scaling: total funding passed $300M in under six months; the new capital funds foundation-model training, simulation compute infrastructure, and commercial teams in healthcare, financial services, and media.
Background
Simile is a Stanford spinout that builds AI foundation models simulating human behavior. Its pitch: with AI, anyone can create a product, campaign, or script, so the hard question is no longer whether you can build something but what to build, for whom, and how it will land. Companies run tens of millions of simulations - customers, patients, or employees - before committing real budget.
The company came out of Joon Sung Park's doctoral research, in which 25 AI agents lived simulated lives in a virtual town called Smallville, a project that won Best Paper at UIST 2023. Park co-founded Simile in late 2025 with Stanford professors Michael Bernstein and Percy Liang and operator Lainie Yallen, launched publicly in February 2026 with a $100M Series A led by Index Ventures, and closed a $200M Series B at a $2 billion valuation in July 2026.
Customers include CVS Health, Wealthfront, Deloitte, and Gallup, and the startup's revenue grew fivefold between its February 2026 launch and the Series B. A recently shipped confidence model scores each simulation's reliability, an explicit answer to the standard critique that simulated humans are only as good as their calibration.
The Series B, led by Greenoaks with Index, Hanabi, Bain Capital Ventures, A*, Factory, Definition, and CVS Health Ventures participating, brought total funding past $300 million in under six months - one of the steepest early funding trajectories among enterprise AI startups of 2026.
What has to be true
- Simile targeted the bottleneck that generative AI created: making things got cheap, but deciding what to make and whether it will work did not.
- The Smallville research gave the founders a credible technical claim - simulated agents carried on conversations and routines without scripting - backed by a top-conference Best Paper award.
- Named enterprise customers (CVS Health, Wealthfront, Deloitte, Gallup) and tens of millions of simulations give the bet verifiable traction rather than demo-stage promise.
- A confidence model directly addresses synthetic users' biggest weakness, calibration, turning the obvious objection into a product feature.
- Capital followed the evidence: a $100M Series A in February and a $200M Series B at $2B in July 2026, five months apart.
What can be applied
The value shift in AI is not generating more content but knowing in advance whether content will work - if you can predict that before anyone commits budget, you own a category, not a feature.
Aftermath
As of September 2, 2026, Simile is scaling with more than $300 million raised in under six months. It is using the Series B to train its core foundation models, expand simulation compute infrastructure, and grow commercial engineering teams in healthcare, financial services, and media. The company has over 50 employees, counts CVS Health, Wealthfront, Deloitte, and Gallup among customers, and has launched a product that lets organizations act on predictions directly rather than treating them as background research.
Sources
- Synthetic-user startup Simile raises $200M at $2B valuation 5 months after $100M Series A
- Simile bags $200M at $2B, five months after $100M Series A, to predict what humans will do before AI gets it wrong
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card