Miles Wang spent two years at OpenAI thinking about safety. Now, apparently, he's thinking about raising money.
Reports surfaced in mid-July through the usual tech-news channels—TechCrunch via Techmeme, the kind of sourcing that's maddeningly vague but rarely wrong—suggesting Wang is shopping a new AI drug-discovery venture at a roughly $2 billion valuation, potentially raising around $200 million. Lightspeed Venture Partners is said to be circling as the potential lead. Wang, who describes himself on his personal site as focused on "beneficial and safe AGI" since March 2024, hasn't confirmed anything publicly, nor has Lightspeed—which is how these things usually work until the wire transfer clears.
What makes the story interesting isn't just the number, though $2 billion is a striking price for a founder with no disclosed track record in pharmaceuticals. It's the timing. Wang's reported fundraise lands in the middle of what might charitably be called a capital frenzy. Billions have poured into AI-enabled drug development in 2026 alone—a figure that dwarfs most prior years and suggests the market has decided something fundamental has changed.
Whether that something is real progress or expensive groupthink remains to be seen.
When Everyone Arrives at Once
The sector didn't used to look like this. For years, AI drug discovery lived in a peculiar limbo: respected in principle, underfunded in practice, and perpetually two years away from proving itself. The algorithms were promising. The capital was cautious. Pharma partnerships came with milestones structured to hedge against disappointment.
That changed in 2026, and the change happened fast.
Isomorphic Labs—Alphabet's bet on turning AlphaFold into an actual drug-design engine—closed a $2.1 billion Series B on May 12, 2026. Thrive Capital led. GV, Temasek, CapitalG, and the UK's Sovereign AI Fund all piled in. The round valued Isomorphic at a figure that would have seemed absurd three years ago, before the company had delivered a single approved drug. But Isomorphic had something better than drugs: credibility borrowed from DeepMind, partnerships with Eli Lilly and Novartis worth a combined $3 billion in potential milestone payments, and the kind of technical pedigree that makes investors suspend disbelief.
Generate:Biomedicines went a different route. On February 26, 2026, the company priced a $400 million IPO at $16 per share, becoming one of the first generative-biology platforms willing to face quarterly earnings calls before proving much of anything in the clinic. It was a bold move, perhaps bordering on reckless. Going public this early means every trial readout gets dissected in real time. It also means access to capital without the constraints of venture ownership.
Then there's Chai Discovery, another OpenAI-adjacent startup, with unconfirmed reports suggesting it's raising $400 million at a $3.8 billion valuation around the time Wang's own fundraise leaked. Chai had been profiled by TechCrunch back in January discussing its Lilly relationship. These aren't seed rounds. They're bets sized for platform-level ambition: compute infrastructure, wet-lab automation, partnerships that assume molecule flow at scale.
The valuations suggest investors believe the tools have matured enough to justify capital deployment at pharma-like levels, even in companies with limited—or nonexistent—clinical data. That belief might be correct. It might also be expensive.
Why Now, Exactly?
Several forces converged to create this moment, some technical, some strategic, some harder to classify.
OpenAI launched GPT-Rosalind on April 16, a suite of life-sciences models designed for experimental workflows rather than just literature searches or diagnostics. A month later came the Rosalind Biodefense initiative, targeting biosurveillance and biosecurity. The message was clear: OpenAI intends to commercialize foundation models for drug R&D, not just chat with researchers about it.
Demis Hassabis, speaking to Fortune after Isomorphic's raise, framed the company's ambition as solving "all disease." The timeline and capital required, he acknowledged, would make typical biotech ventures look quaint. It was the kind of statement that sounds grandiose until you remember that Hassabis tends to deliver on grandiose statements.
Meanwhile, the infrastructure layer matured in ways that matter more than they sound. Eli Lilly and NVIDIA announced a Co-Innovation AI Lab in January to "reinvent drug discovery" using in-silico exploration at scale. IQVIA launched IQVIA.ai in March, an agentic platform for clinical operations and real-world evidence—proof that contract research organizations are betting on AI workflows across the entire development lifecycle, not just the sexy early-stage design work.
Regulators joined the party, tentatively. The UK's MHRA expanded its AI Sandbox on June 9 to accelerate regulatory engagement for AI-assisted medicines development. The FDA finalized ICH M15 guidance in June, creating a clearer path for AI-generated modeling and simulation to inform trial design and dose selection. These aren't game-changers by themselves, but they're signals that regulatory agencies are trying to catch up rather than standing in the way.
Perhaps most important: the first prospective evidence is arriving. IQVIA's Global R&D Trends 2026 report, published in April and May, flagged early signals of stronger success rates among AI-enabled programs, particularly in molecule discovery. Insilico Medicine advanced multiple AI-designed assets into Phase 2a trials during late 2025 and early 2026, including ISM5411 for inflammatory bowel disease. These aren't retrospective validations or benchmarking exercises. They're first-in-human data, the kind that will either validate or deflate the sector over the next year and a half.
The Devil Lives in Phase 2
Isomorphic's massive raise illustrates how willing investors have become to fund at scale before widespread clinical proof, provided the technical pedigree checks out. The company's foundation—AlphaFold 3, published in Nature in May 2024—anchors its credibility, even as peer-reviewed analyses in 2026 have raised reliability concerns about newer binding-affinity models like Boltz-2. Isomorphic's play seems to be closing the loop from prediction to validated design through partnerships that provide therapeutic-area expertise and clinical infrastructure it doesn't yet have in-house.
Insilico's Lilly deal, announced March 29, front-loaded $115 million and structured the rest as milestones and royalties—pharma's traditional hedge against platform risk. The collaboration reflects Lilly's broader strategy, which appears to be placing multiple bets across the emerging AI drug-discovery ecosystem rather than committing exclusively to any single platform. Lilly also partnered with NVIDIA and Chai Discovery. It's the pharmaceutical equivalent of portfolio diversification.
Not every story has an upward trajectory. BenevolentAI delisted from Euronext Amsterdam in March 2025 as part of a restructuring—a sobering reminder that early-stage AI platforms can struggle to translate computational elegance into sustainable business models. The gap between BenevolentAI's retrenchment and Isomorphic's $2.1 billion raise is striking. The market has segmented winners from laggards with brutal efficiency, based on technical depth, partnership traction, and capital discipline.
What Comes Next, Maybe
The capital flooding into AI drug discovery in 2026 is neither irrational exuberance nor a sure bet. Call it calculated gambling. The wager is that foundation models, combined with closed-loop automation—design, synthesize, test, retrain—and pharma partnerships can compress the decade-long, billion-dollar timeline that has defined drug development since the 1990s.
McKinsey estimated in early 2025 that generative AI could unlock $60 billion to $110 billion annually across the pharma and medical-products value chain. Those figures are projections, not realized gains, the kind that look brilliant in hindsight or ridiculous depending on what happens in Phase 2.
Analysts at PitchBook characterized 2026 as an inflection year where agentic AI and hyperscaler partnerships would separate credible platforms from undercapitalized experiments. They might be right. The question is whether the sector can deliver clinical proof points fast enough to justify these valuations. Insilico's Phase 2a data and Generate's upcoming milestones will offer early signals. Positive readouts could trigger another wave of capital in 2027. Disappointing ones will cause the market to reprice quickly, and late-stage rounds will become scarce.
For founders and pharma executives, the window to establish platform credibility feels like it's narrowing. Sam Altman suggested in a February Bloomberg interview that OpenAI might dramatically subsidize firms using its AI for drug discovery, potentially taking royalties instead of upfront fees—a model that could democratize access to frontier models but also intensify competition. Regulatory pathways are clarifying, slowly: the FDA has reviewed over 500 drug submissions with AI components since 2016. The EU's AI Act, which entered force in August 2024, will impose staggered obligations on general-purpose AI models and high-risk applications through 2027.
Miles Wang's reported raise, if it closes, will add another data point to an already crowded landscape. The $2 billion valuation suggests confidence. But confidence without clinical validation is just expensive optimism dressed up in a term sheet.
The next twelve to eighteen months will reveal whether 2026 was the year AI drug discovery became real or the year it became a bubble. Either way, it'll be interesting to watch.
