Kyunghyun Cho launches Ortet, a frontier AI health lab, with $500M
A co-author of the attention mechanism spent five years building AI for drug discovery inside Genentech. Today he launched his own lab — with a half-billion dollars behind it, one disclosed backer, and no priced round.
Ortet launched on September 29 as a frontier AI lab for health, with a $500 million commitment from Thoreau and a founding team led by Kyunghyun Cho. Cho co-developed the attention mechanism and the gated recurrent unit — the groundwork under every transformer in production today — and holds the Glen de Vries Professorship of Health Statistics at New York University. He previously co-founded Prescient Design, which Genentech acquired, and then built that company's frontier research team. The founding bench is deliberately vertical: Keunwoo Choi (Upstage, Spotify, ByteDance, Genentech) as chief AI officer; Henri Dwyer, who built and managed the largest GPU cluster in life sciences at Genentech, as chief technology officer; Cloudera founder Jeff Hammerbacher as chairman; neural text-to-image pioneer Elman Mansimov as chief science officer; and Claudia Shi from Meta Superintelligence Labs leading frontier research.
The product is the Ortet Health Model — one model expected to reason across a patient's biology, clinical record, care operations and financial history at the same time. Ortet's argument is that healthcare AI today is task-specific and narrow, so a model that reads a scan never learns how the same patient's treatment, outcomes and billing connect. Its alternative is a "closed learning loop" that keeps improving as real care generates new data, with the individual patient as the unit of analysis rather than the population. The company says it has already secured its initial GPU cluster, and that Goldman Sachs advised on the transaction.
Read the structure before the ambition: this is a commitment from one named investor, not a priced round, and Ortet is described as "a founding Thoreau investment." Thoreau builds, invests in and acquires health businesses across provider, payer, employer, life sciences and government markets; its other named platform partnerships are Ensemble, which manages more than $55 billion in net patient revenue, and Penelope Health, a health policy intelligence product. That is where the data is presumably meant to come from — and it is also the catch, because a closed learning loop needs patient data from health systems willing to feed it, and the only pipeline anyone has named belongs to the investor. Cho has the résumé to attempt this: he comes from exactly the world we wrote about this morning, where Roche is building labs where the AI runs the experiments. But what decides whether Ortet becomes a platform or stays a pilot is a customer outside Thoreau's own house.
What to watch: whether Ortet names a health system outside Thoreau's portfolio, and whether the commitment converts into a priced round with disclosed terms — a single strategic backer funding an "initial" GPU cluster is a starting position, not a balance sheet.
If the founder helped invent the architecture and the investor already owns the health platforms, which is the harder part — the model, or getting the patient data? Tell us in the comments.
