From Words to Action — A $1 Billion Joint Lab
Digital biology is an interdisciplinary field that fuses AI, computational science, and bioinformatics with biology to model, simulate, and design biological systems — cells, genes, proteins, and more — on a computer. It marks a shift from observation-centered science toward engineering, in which hypotheses are verified computationally and optimized before experiments begin.
In January 2026, NVIDIA and Eli Lilly announced a joint innovation lab at the JPMorgan Healthcare Conference. The investment was reported at roughly $1 billion, and Huang called it a "blueprint for what is possible in the future of drug discovery." Using NVIDIA's BioNeMo platform, the lab will generate large-scale datasets and develop AI models for biology and chemistry. Its scope reaches beyond drug discovery into clinical development, manufacturing, and commercialization.
Funding and Clinical Milestones
Isomorphic Labs, a spinoff of DeepMind, was reported to have raised a Series B of roughly $2.1 billion led by Thrive Capital in May 2026. Alphabet, GV, MGX, Temasek, and a UK government-backed fund were among those named, and some reports say an additional raise at a valuation above $40 billion is under discussion. Founder Demis Hassabis won the 2024 Nobel Prize in Chemistry for the AlphaFold work.
On the technical front, the company has positioned its Drug Design Engine — which builds on AlphaFold's structure predictions to identify binding pockets in disease-related proteins and generate and evaluate thousands of small molecules — as "the frontier beyond AlphaFold," and preparations are said to be advancing toward human clinical trials of an AI-designed drug. The AlphaFold DB now holds more than 241 million structures, and in 2026 attention has centered on the announcement of a new model widely seen as its de facto successor. At NVIDIA's GTC 2026 as well, the arc was framed as a milestone: from simple structure prediction to simulating complex protein-protein interactions, and on to agentic AI in healthcare.
Two Bottlenecks
At the same time, the factors holding growth back are clear. The first is data. Unlike text, high-quality biological data such as binding and functional assays can only be obtained through experiments, so even though AI drug discovery is the field's most closely watched application, its training material remains in short supply. The NVIDIA and Lilly lab starting with data generation is precisely a head-on answer to this wall.
The second is talent. Researchers who understand how biological data is generated and can also work with LLMs are scarce, and cultivating people who can bridge both is widely regarded as an urgent task.
How to Read It
If Huang's prediction is right, the speed at which biology can be treated as engineering becomes the speed of progress. A $1 billion lab, a $2.1 billion raise, more than 200 million structures — now that the money and the data are moving to the core, what comes next is an engineering question: how to produce the data only experiments can yield, faster and cheaper.
この記事を閉じる3つの問い
- When experimental data generation is the rate-limiting step, will Japanese pharma bet on in-house capabilities or external partners?
- How far will clinical results for AI-designed drugs reinforce the premise that biology equals engineering?
- How will the gap between the haves and have-nots of structural data — those with more than 200 million structures and those without — change the cost of drug discovery?