Isomorphic Labs, the Alphabet spin-off founded by Google DeepMind chief Demis Hassabis, has raised $2.1 billion to push its AI drug design platform into the next phase. The message from investors is blunt: if this works, pharma could get its own OpenAI moment, only with fewer chatbots and a lot more lab coats.
The new money lifts total outside funding for the company to about $2.6 billion. That is an enormous vote of confidence for a business that has not yet brought a single AI-designed drug into clinical trials, but it also reflects a broader bet on AI moving from text generation into hard science, where the payoff can be measured in years saved and molecules discovered.
IsoDDE is the next step after AlphaFold
Isomorphic’s roots are in AlphaFold, the protein-structure system that helped Hassabis and John Jumper win the Nobel Prize in chemistry in 2024. But the company is no longer just predicting what already exists. Its newer platform, IsoDDE, is designed to create drugs, and internal benchmarks suggest it is pushing well past AlphaFold 3 on the hardest molecular tasks.
On the Runs N’ Poses benchmark, which tests how well a model handles new protein pockets and ligands, IsoDDE more than doubled AlphaFold 3 on the toughest cases. In antibody-antigen modeling, one of the hardest problems in drug discovery, it reportedly reached about 76% accuracy versus 48% for AlphaFold 3. That is the kind of gap that gets drug companies interested fast.
A closed system, not a science project
The bigger shift is strategic. AlphaFold 1 and AlphaFold 2 were open; AlphaFold 3 at least came with a scientific paper. IsoDDE is fully closed, which tells you exactly where the value is now: not in prestige, but in proprietary infrastructure. The open-source ecosystem has been racing to catch up with tools such as Boltz-2, Chai-1, OpenFold3, and RoseTTAFold All-Atom, but Isomorphic has again moved the goalposts.
That secrecy also mirrors what has happened across the AI industry more broadly. The best-known model makers have turned their flagship systems into tightly controlled products, and scientific AI is following the same route. For researchers, that means less visibility into how these models work; for investors, it means a stronger moat.
The real test is still the clinic
For all the hype, Isomorphic has not yet delivered a drug to patients. Hassabis said at the World Economic Forum in Davos that first clinical trials would move to the end of 2026, later than his earlier target of the end of 2025. The company says it is already running 17 drug-development programs across oncology, immunology, and cardiovascular disease, with help from partners including Novartis, Eli Lilly, and Johnson & Johnson.
- Total outside funding: about $2.6 billion
- New round: $2.1 billion
- Programs underway: 17
- Clinical trials target: end of 2026
The new capital will go into IsoDDE, internal therapeutic programs, and team expansion in London, Cambridge, and Lausanne. If the platform keeps outperforming older systems, the pressure on rivals will only grow. If it stalls before the clinic, the sector gets another expensive reminder that predicting biology is one thing; changing medicine is another.