Isomorphic Labs in Talks to Raise $2B on AlphaFold Gains

Isomorphic Labs, the Alphabet unit built around AI-driven drug discovery, is in advanced talks to raise more than $2 billion in new capital. SiliconAngle reports that Thrive Capital is expected to lead the round, with Alphabet also participating directly. That combination of a top-tier growth fund and a strategic parent is not routine: it signals that investors now believe AI-native drug design is approaching commercial maturity.

The company has a credible scientific foundation. Isomorphic Labs was spun off from Google DeepMind in 2021 and is led by Demis Hassabis, who co-founded DeepMind and shared half the 2024 Nobel Prize in Chemistry for work on AlphaFold 2. That model automated protein structure prediction, a task historically requiring years of experimental crystallography. When the Nobel Committee singles out a computational tool, it tells you something about how thoroughly the underlying science has been stress-tested.

The AlphaFold 3 baseline

In 2024, DeepMind and Isomorphic Labs jointly released AlphaFold 3. Per SiliconAngle, the model delivers accuracy improvements exceeding 50% across certain research categories compared to its predecessor. Its particular strength is small-molecule analysis, which matters because most pharmaceuticals are small molecules that must bind to protein targets. For computational biologists, this shifts the modeling problem from isolated protein chains to full ligand-protein complexes.

IsoDDE and the commercial layer

Isomorphic Labs is not positioning AlphaFold 3 as its product. Its commercial system, the Drug Design Engine or IsoDDE, is designed to go further still. On the Runs N' Poses benchmark, which evaluates binding pocket identification and related molecular docking tasks, IsoDDE reportedly more than doubled AlphaFold 3's score across the most difficult task categories.

Binding pocket prediction is the entry point of early-stage drug discovery. Researchers scan disease-relevant proteins for structural pockets through which drug molecules can be delivered to neutralize a target cell. Automating that identification reliably could compress months, sometimes years, from preclinical timelines. IsoDDE's benchmark result, if it holds in live programs rather than retrospective test sets, is the core commercial claim behind the fundraise and the valuation it implies.

What the round signals

Model tracking services like llm-stats have logged multiple flagship model releases per month in 2026, a pace that reflects how intensely capital and engineering talent are now concentrated in artificial intelligence. The Isomorphic Labs raise is a domain-specific counterpart to that general pattern. Rather than betting on a single foundation model platform, Thrive Capital is apparently placing positions across general-purpose and specialized AI simultaneously, given its existing stake in OpenAI. That portfolio logic suggests the fund sees structural biology as a distinct category, not merely a downstream application of language models.

What distinguishes this particular bet from many artificial intelligence investments is the verifiability of the underlying science. Protein structure prediction has ground truth: predicted structures can be checked against experimental crystallography or cryo-electron microscopy data. That falsifiability is part of what made AlphaFold 2 credible enough to earn a Nobel, and it provides a harder evaluative standard for IsoDDE's claims than typical language model benchmarks. Runs N' Poses is, however, an internal evaluation. Independent replication in peer-reviewed research has not yet been reported publicly.

The deeper uncertainty is translation. AI models that excel on retrospective holdout sets do not automatically perform well when confronted with a live drug program, where target conformations may diverge from the training distribution. As CNBC noted recently, the AI sector broadly is experiencing a breakneck pace of successive releases, but in drug development, speed without prospective clinical validation is noise. Pharmaceutical partners evaluating any licensing arrangement will be asking exactly that question.

If the round closes at the reported figure, Isomorphic Labs will have the resources to run the large-scale prospective studies needed to turn benchmark wins into signed pharma partnerships. The bet embedded in the $2 billion is that artificial intelligence in medicine has crossed from proof-of-concept to industrial tool. Whether clinical data agrees is the question the next 18 months will answer.

Frequently asked questions

What is Isomorphic Labs and who owns it?
Isomorphic Labs is an Alphabet subsidiary spun off from Google DeepMind in 2021. It focuses on AI-accelerated drug discovery. Alphabet is its primary backer, and Thrive Capital is an existing outside investor expected to lead the current $2 billion-plus raise.

What improvement does AlphaFold 3 offer over AlphaFold 2?
AlphaFold 3 delivers accuracy gains exceeding 50% across certain research categories relative to its predecessor. It extends prediction beyond proteins alone to model interactions with small molecules, a critical addition for drug design. AlphaFold 2 focused primarily on predicting standalone protein chain structure.

What is the Runs N' Poses benchmark and how reliable is it?
Runs N' Poses is a benchmark that evaluates AI models on binding pocket identification and molecular docking tasks. Isomorphic Labs reports IsoDDE scores more than twice as high as AlphaFold 3 on the hardest task categories. It is an internal evaluation and has not yet been independently replicated in published research.

Why would Thrive Capital invest in both Isomorphic Labs and OpenAI?
Thrive Capital appears to be pursuing a strategy that covers both general-purpose AI platforms and domain-specific AI applications. Structural biology represents a field where specialized models can claim verifiable advantages over general-purpose systems, making it a logical complement to a position in a foundation model company.