Miles Wang dropped out of Harvard to join OpenAI. Now he's reportedly raising $200 million to bet that AI can find new lives for drugs pharma already gave up on.
Wang is in advanced talks to launch an AI drug discovery startup at a $2 billion valuation, according to a TechCrunch report published July 14. That is a very rich price for a company without a public name. It also tells you exactly where venture money is moving now: out of broad AI demos and into models that might change the economics of drug development.
Lightspeed Venture Partners is reportedly leading the round, and Wang is expected to bring several other OpenAI researchers with him when he leaves. He dropped out of his computer science degree at Harvard. He joined OpenAI in 2024. Since then he's worked on applying transformer architecture, the same technology underlying ChatGPT, to biological data.
Wang disputed TechCrunch's reporting on the funding figures and the description of the company. He didn't offer his own numbers. He didn't say what was wrong, either. Sit with that. When a founder pushes back on a story without correcting it, the deal is often still moving. Misreported stories usually get corrected.
The startup's apparent focus is drug repurposing: training models to find new indications for drugs that are already FDA-approved, or that failed clinical trials for their original purpose but might work for something else entirely. That's a real shortcut, not a marketing line. A drug that's already cleared safety trials doesn't need to go through that gauntlet again for a new use, which can cut years off the path to revenue and spare a founder the cost of a phase one trial from scratch.
Chai Discovery gives investors the comparison they want
Wang isn't the first OpenAI researcher to walk this path, and the timing here is not a coincidence. Josh Meier, a former OpenAI researcher, co-founded Chai Discovery in 2024. On July 14, the same day the Wang story broke, Chai Discovery closed a $400 million Series C at a $3.8 billion valuation, according to SiliconANGLE. That's a near-tripling of its valuation in under eight months, up from the $1.3 billion mark it hit with its Series B in December.
Chai's flagship model, Chai-2, designs novel antibodies from scratch. Its hit rate on viable candidates runs around 20%, versus roughly 0.1% for older computational methods, and Pfizer and Eli Lilly are both paying customers. OpenAI itself is listed among Chai's backers. If you're an investor staring at Wang's pitch deck this week, that is the comp you keep circling.
That's the comparison every investor looking at Wang's deal is making right now.
The valuation is the bet
Most AI biotech startups launch somewhere between $200 million and $500 million in valuation, even with strong founding teams behind them. Wang's reported $2 billion starting point sits well outside that range, and he doesn't have a product yet, or even a public company name. What he has is a resume: OpenAI pedigree, a specific technical bet on applying language-model architecture to biology, and a market that has decided frontier-model builders should be priced before they prove much in wet labs.
Frankly, that's the real story here, more than any single startup. Lightspeed and firms like it aren't underwriting a product. They're underwriting a belief: that whoever builds the best model for biological data will look a lot like whoever built the best model for language, and that the people who did the second thing are the safest bet on the first.
That belief has some evidence behind it. Chai Discovery's valuation trajectory gives investors a name and a number to point at. Chai-2 gives them a model result that sounds meaningfully better than older computational methods. Pfizer and Eli Lilly give the story commercial weight. You can see why the round is happening.
But biology doesn't care where a researcher used to work. Does a model trained to predict the next word in a sentence actually transfer to predicting which molecule binds which protein? Chai's reported hit rate suggests something real is happening in that direction. Wang's team will still need to show comparable results against real targets, not just credentials, before a $2 billion valuation means anything beyond investor appetite.
For now, the round hasn't closed. The company doesn't have a public name. And Wang still hasn't said what TechCrunch got wrong.
Also read: Meta workers say its AI quietly decided who lost a job in mass layoffs • CoreWeave's Long-Term Contracts Never Hedged Against Memory Chip Prices • Hinge's Founder Just Bet $18 Million That Swiping Is Broken
