GOOGL
META
Big Tech’s latest side quest: biology
Meta Platforms and Alphabet are teaming up on a $300 million investment in Biohub’s Virtual Biology Initiative, a bid to build AI models that can predict how living cells behave. If that sounds a little sci-fi, it is — but the goal is practical: test drug ideas and genetic interventions in software before burning time and money in the lab.
Why this matters to your portfolio
This isn’t just “AI, but make it weird.” It’s another sign the biggest tech companies are using their cash piles to buy optionality in adjacent markets. For Meta, this adds another showcase for its AI research chops. For Alphabet, it gives Isomorphic Labs more data and more fuel for its drug-discovery ambitions.
- Biohub’s goal is a “universal virtual cell” that can predict how cells respond to interventions.
- Commercial partners get one year of exclusive access to the data they help generate.
- Biohub expects its first large dataset in about a year and says useful predictive models could take roughly five years.
The catch: biology is not chatbots
Here’s the wrinkle: AI models trained on biology don’t get to feast on the internet like LLMs do. The hard part is generating the experimental data from scratch, which is slower, pricier, and way less glamorous than scraping the web.
And even then, there’s no guarantee the scaling laws behave nicely. A recent Nature Methods study found that performance on 400 single-cell foundation models often plateaued before all the available data was used. Translation: more data may help, but biology might refuse to play by the same rules as language models.
Big picture
This is Meta and Alphabet trying to turn AI from a software story into a science story. If it works, the upside is enormous. If not, well, it’s still a very expensive reminder that the road from “promising model” to “real-world breakthrough” is full of potholes.