There's an old saying about the Gold Rush that every venture investor has heard so many times they've stopped hearing it: the people who got rich weren't the ones panning for gold, they were the ones selling picks and shovels. It's a tidy story, and like most tidy stories it's only half true. The other half, the half nobody quotes, is that eventually the smartest miners stopped buying shovels from the hardware store and started buying the hardware store.
That is more or less what biopharma dealmaking has been doing over the last couple of weeks, and it is worth a Friday to sit with, because it marks a quiet turn in how this industry decides to spend its money, and a turn whose center of gravity sits in Seoul, Shanghai, and Hong Kong as much as anywhere west of them.
For a while now, "AI in drug discovery" has been the thing everyone agrees is important and few have been sure how to bet on. The pitch decks promised the world. The deals, when they came, tended to be tentative: small bets that staked a claim without sinking a shaft, a research collaboration here, a target-ID partnership there. The technology was real but the commitment was provisional. Companies wanted the upside of AI discovery without having to believe in this is it hard enough to write a real check.
Lately, several of them have written the check, and the structure of how they wrote it tells you they have stopped surveying the claim and started digging.
The Trial Period, With an Exit Clause
Start with LG Chem, the Seoul-headquartered chemicals-and-pharma group, which on June 17 signed a multi-year deal with a UK firm called LabGenius to use its machine-learning platform, EVA, to design next-generation multispecific antibodies aimed at a solid-tumor antigen that turns up across several hard-to-treat cancers. The structure shows exactly how much trust is and isn't being extended. LabGenius runs the program through early preclinical work, the in vitro efficacy studies. Then LG Chem takes the baton for the expensive, dangerous part, the in vivo studies, and only then decides whether to in-license the asset at all. LG Chem funds everything along the way.
This is not a company buying a magic box. It is a company renting the box for one clearly defined job, paying to watch it work on a single problem, and keeping the right to walk away if the demo disappoints. The triple-digit-million milestones everyone will headline only matter if the option gets exercised, and the option only gets exercised if the AI actually delivers a molecule worth owning. It's the difference between buying the mine outright and leasing the right to dig one test shaft before committing to the seam. LG Chem, sensibly, has leased the test shaft.