Meta, DeepMind and Isomorphic pay $300 million for first use of Biohub's open cell data

The four commitments behind Biohub's headline add up to $1.8 billion ^[17], give or take the "more than" attached to both federal figures ^[9][11]. One of them, the NIH's, is federal money already spent building repositories that exist today ^[11]. Take it out and new commitments come to roughly $1.3 billion ^[18]. The Department of Energy's portion works out to more than $100 million a year ^[20]. Meta, Google DeepMind and Isomorphic Labs, with $300 million between them ^[2], supply about 17% of the total ^[19].

That 17% buys time ^[19]. "With commercial funders we have embargo periods where there's a period of time where the groups can work on the data, and then it becomes available as a public scientific resource," Alex Rives, Biohub's head of science, said ^[4]. For $300 million the three companies get a lead on data that will eventually be public, and they split the cost of producing it with Biohub and two federal agencies ^[2][9][11]. Biohub did not say how long the embargo runs or how the $300 million divides among the three, and neither report cites a startup financing that would show the effect on venture money ^[4].

"We have always held this as a community asset, not just for one group, so that it can build upon itself over time," Priscilla Chan said ^[7]. With Meta also writing a check, Cryptobriefing noted, the Zuckerberg name appears on both sides of the funding ^[16]. For biotech investors the funder that matters is Isomorphic Labs, which the CNA report calls a drug discovery startup ^[10]. A company in the business of finding drugs gets to work on the shared data before its competitors do ^[4][10].

Scale is still the binding constraint. Current cell datasets run to hundreds of millions of cells, and Rives said an accurate predictive model will need billions and eventually trillions ^[13]. Reaching one billion would take more than eight times the largest set the initiative has produced so far ^[21]. Rives expects a first dataset in about a year and accurate predictive models within five ^[6].

For startups that raise money on proprietary cell data, the embargo length decides most of the outcome ^[4]. If it is long, the three funders have bought a private lead for $300 million ^[2], and startups face better-funded rivals who see a larger set first. If it is short, the public baseline rises each year and a private perturbation screen is worth less as a fundraising asset. And if the data reaches billions of cells late, private sets keep their value for longer ^[13]. I think the short-embargo case is the most likely one for startups whose data resembles Biohub's, which comes from spatial transcriptomics and screens of how cells respond to changes in environment ^[14]. The federally funded work running alongside has no embargo ^[5].

The counter-thesis is that cell-level data was never where these startups kept their value, and that companies holding other kinds of data keep their pricing. Other labs are moving into biology data too: Anthropic has added a wet lab, and the OpenAI Foundation has started a grant program of more than $125 million for biological and medical datasets for AI research ^[15]. If the commercial embargo turns out to run for years, I am wrong, and the advantage goes to whoever can pay for early access ^[4].