# Simile raises $200M at a $2B valuation to simulate human behavior > Source: > Published: 2026-07-30 17:10:41+00:00 [Joon Sung Park (@joon_s_pk)](https://x.com/joon_s_pk), CEO and co-founder of [Simile](https://www.simile.com/blog/series-b), has raised a $200 million Series B at a $2 billion post-money valuation for an AI platform that lets enterprises simulate how people may respond to products, policies and business decisions. [https://x.com/simile_ai/status/2082873889407827980?s=46](https://x.com/simile_ai/status/2082873889407827980?s=46) [Simile announced the financing in a post on X](https://x.com/simile_ai/status/2082873889407827980?s=46) on July 30th. [Greenoaks](https://greenoaks.com/) led the round, with Index Ventures, Hanabi, A*, [Bain Capital Ventures](https://baincapitalventures.com/), CVS Health Ventures and Definition participating. Simile's funding now totals at least $300 million, including the $100 million Series A that Index led in February. The pace is the point. Investors have assigned Simile a $2 billion valuation about five months after its public launch, before the company has disclosed the revenue scale or customer count that would normally provide context for that price. Simile says revenue grew fivefold during that period, but its announcement does not provide a starting figure. The round instead prices Park's research pedigree, the founding team's Stanford ties and early adoption by large enterprises. Park's work predates the current rush to sell synthetic customers to corporate buyers. He earned a Ph.D. in computer science at Stanford after studying at Swarthmore College and the University of Illinois Urbana-Champaign. In April 2023, Park and his collaborators published the ["Generative Agents" paper](https://arxiv.org/abs/2304.03442), which placed 25 language-model agents in a virtual town. The agents used memory, reflection and planning to form relationships, exchange information and coordinate a Valentine's Day party. Simile turns that research into a commercial proposition: companies should test decisions against populations of AI agents before putting those decisions in front of real customers. The agents are grounded in interviews, self-reported information and past choices from real people, rather than assembled solely from prompts describing generic customer personas. Park founded Simile with [Michael Bernstein](https://profiles.stanford.edu/michael-bernstein), a Stanford computer science professor whose work focuses on social and interactive systems; [Percy Liang](https://profiles.stanford.edu/percy-liang), a Stanford professor who directs the university's Center for Research on Foundation Models; and [Lainie Yallen](https://www.linkedin.com/in/lainie-yallen-619156127), who leads go-to-market work. Park describes his broader project as building societal simulations that can test the assumptions underneath high-stakes institutional and commercial decisions. ### From research paper to enterprise product Simile says its foundation model for human behavior has run tens of millions of simulations for Fortune 100 enterprises. The company names CVS Health, Wealthfront, Deloitte and Gallup as customers using the system to evaluate product launches, customer experiences and market-entry plans. Those figures and use cases are company-reported. The clearest public example comes from CVS Health, which is both a customer and an investor in the Series B. In a [case study published by Simile](https://www.simile.com/blog/simile-cvs-health), the companies said CVS's agents were built using 2.9 million consented responses from more than 400,000 participants across over 200 behavioral scenarios. CVS used the simulations to screen research questions, examine customer journeys and test approaches to medication adherence before running real-world studies or pilots. That customer-investor overlap gives Simile capital from an enterprise positioned to judge the product in production. It also makes independent validation central to the company's next phase. The value of simulated populations depends on whether their predictions continue to match real behavior when conditions change, especially in health care, litigation, policy and other settings where an incorrect forecast can redirect significant capital or affect real people. Simile says it has trained a confidence model designed to predict the accuracy of each simulation. The company has not published performance data for that model in its funding announcement. Its research lineage gives Simile a stronger technical foundation than vendors that generate synthetic personas from short prompts, while enterprise buyers will still need repeatable comparisons between simulated forecasts and observed outcomes. ### A $200M bid to define the category The Series B gives Simile enough capital to hire ahead of revenue and establish simulation as a separate layer in the enterprise AI stack. Simile says it has expanded from a small Palo Alto operation to a global workforce of more than 50 people since launch. The company's timing rests on a shift in corporate AI spending. Generative models have reduced the cost of producing product concepts, marketing campaigns and strategic options. Simile is betting that enterprises will pay for a system that helps select among those options before committing engineering time, marketing budgets or operational resources. The $2 billion valuation makes that bet unusually expensive only months into commercialization. Simile must now prove that simulations grounded in real people can outperform conventional surveys, customer interviews, panels and live experiments often enough to become recurring decision infrastructure. Greenoaks and returning backers Index, Hanabi, A* and Bain Capital Ventures are financing that test at scale.