The Signal — March 31, 2026
Eli Lilly puts $2.75B behind AI-discovered drugs. A Quinnipiac poll finds most Americans use AI but few trust it. Mistral borrows $830M for European compute.
Pharma is putting real money behind machine-discovered drugs. A pollster finds nobody trusts the tools they're already using. And banks are treating GPU clusters like real estate.
Eli Lilly Bets $2.75 Billion on Machine-Discovered Drugs
Eli Lilly has signed a deal worth up to $2.75 billion with Hong Kong-based Insilico Medicine for exclusive worldwide rights to develop, manufacture, and commercialize preclinical oral drug candidates discovered through Insilico's research platform. The upfront payment is $115 million, with the remainder structured as development, regulatory, and commercial milestone payments plus tiered royalties.
This expands an existing relationship that began with a software licensing agreement. Lilly is no longer just licensing Insilico's tools; it's buying the output. The compounds haven't entered human trials, so the real proof is years away. Clinical failure rates for machine-discovered drugs remain an open question.
But the scale of the commitment is the story. A top-five pharma company putting milestone-backed billions behind these candidates sends a demand signal. As STAT News noted, the biobucks structure is standard pharma risk management, but the total number is not.
Sources: Reuters · CNBC · STAT News · FierceBiotech
51% Use AI for Research. 21% Trust It.
A Quinnipiac University poll found that more than half of Americans have used generative AI tools for research, but only one in five trust the results. The gap between adoption and trust is the widest measured in any major national survey to date.
The poll also found sharp generational divides: younger users adopt faster but aren't necessarily more trusting. Respondents who reported using AI tools daily were only marginally more confident in accuracy than occasional users. The tools are spreading through habit, not conviction.
Sources: Quinnipiac University Poll
Mistral Borrows $830 Million to Build European Compute
French lab Mistral has secured $830 million in debt financing from a seven-bank consortium to purchase 13,800 Nvidia chips and build a data center near Paris. The financing is debt, not equity, and that distinction carries weight. Banks are now underwriting compute hardware as bankable collateral, treating GPU clusters the way they've traditionally treated real estate or industrial equipment.
Bloomberg reported this is Mistral's first-ever debt raise. The company is targeting 200 megawatts of compute capacity across Europe by the end of 2027, with expansion planned into Sweden. Owning infrastructure gives Mistral independence from hyperscaler pricing and the ability to offer European data residency.
Sources: Reuters · CNBC · Bloomberg · TechCrunch
On the Editor's Desk
Three stories that didn't make it today: a Rolling Stone investigation on undisclosed AI use across the recording industry, a Stanford paper on models auto-optimizing their own scaffolding, and the LiteLLM/Delve breakup following last week's credential theft (covered three times already). A multi-front governance battle between states and the White House was too complex for this format.
