Isomorphic Labs Raises $2.1B: Demis Hassabis Bets AI Cures Disease

The number that matters most in the Isomorphic Labs announcement is not $2.1 billion. It is zero. Zero human patients have been dosed with a drug that Isomorphic designed. The entire raise — the largest AI biotech funding round of 2026 — is a bet that this number changes before the year ends.

Isomorphic Labs closed its Series B on May 12, 2026. Thrive Capital led the round. Alphabet and GV participated as existing investors. New money came from MGX (Abu Dhabi's sovereign wealth fund), Temasek (Singapore's sovereign wealth fund), CapitalG, and the UK Sovereign AI Fund. The investor geography is the story within the story: this round is funded simultaneously by capital from the United States, the United Kingdom, Abu Dhabi, and Singapore. AI drug discovery has become a geopolitical priority, not just a research project.

What Isomorphic Labs Actually Is

Isomorphic Labs was founded in 2021 as a spinout from Google DeepMind — the London-based AI research lab that Demis Hassabis co-founded in 2010. The premise: if the same deep learning that solved protein structure prediction could be applied to drug molecular design, you could compress the drug discovery timeline from decades to years.

Hassabis serves as CEO. The company now operates across London, Cambridge Massachusetts, and Lausanne Switzerland. Its internal pipeline focuses on cancer and immunology. It holds strategic partnerships with Novartis, Lilly, and Johnson & Johnson — three of the largest pharmaceutical companies in the world — who are paying to use Isomorphic's AI on their own discovery programmes.

The name is a nod to the mathematical concept of isomorphism: the idea that two apparently different structures can be fundamentally equivalent under a mapping. The thesis is that the structure of drug molecules and the structure of biological targets are isomorphic in a way that AI can learn to exploit.

AlphaFold: The Nobel Prize That Made This Possible

Isomorphic Labs exists because of AlphaFold. Understanding the lineage matters for understanding why $2.1 billion is credible.

AlphaFold is the DeepMind AI system that solved protein folding — predicting the three-dimensional structure of proteins from their amino acid sequence with near-experimental accuracy. This had been one of biology's hard unsolved problems for 50 years. AlphaFold 2 solved it in 2020. AlphaFold 3 extended the approach to predict the interactions between proteins, DNA, RNA, and small drug-like molecules.

The practical significance for drug discovery: a drug works by binding to a specific protein target in the body. If you can accurately predict how a drug molecule interacts with a protein structure, you can computationally screen billions of potential candidates without synthesising them in a lab. You identify the promising ones computationally and only run wet-lab experiments on the best candidates.

AlphaFold earned Hassabis and DeepMind researcher John Jumper the 2024 Nobel Prize in Chemistry. Isomorphic Labs takes the same protein-structure reasoning and applies it directly to the drug design problem — not just predicting structure, but generating novel molecules that interact with that structure in therapeutically useful ways.

IsoDDE: How the AI Drug Design Engine Works

The company's core product is IsoDDE: the Isomorphic Drug Design Engine. It is a unified AI system built from several breakthrough models that work together across the drug discovery pipeline.

Structure prediction with AlphaFold 3 at the base: IsoDDE starts with protein structure understanding. AlphaFold 3 predicts how proteins, DNA, RNA, and small molecules interact — giving IsoDDE an accurate physical model of what happens when a candidate drug meets its target.

Cryptic binding pocket identification: One of IsoDDE's key capabilities is identifying "cryptic" binding sites — pockets on a protein surface that only become accessible when the protein moves, and which are invisible in static structural data. Isomorphic demonstrated that IsoDDE can identify these sites using only the protein's amino acid sequence as input, without specifying what ligand might bind there. This capability approaches the performance of experimental techniques like fragment-soaking that require weeks and significant cost to run. IsoDDE does it in seconds on a computer.

Concrete example — cereblon: Isomorphic has publicly described how IsoDDE recapitulated the discovery of a novel cryptic site on cereblon, a protein involved in protein degradation pathways that are a target for cancer therapies. The model predicted both a previously known binding site and the novel cryptic site from the protein sequence alone.

Generative molecular design: Rather than screening a fixed library of known molecules, IsoDDE generates new molecular structures. It proposes molecules that have never existed before, then evaluates their predicted binding, selectivity, solubility, and synthesisability computationally.

Multi-objective optimisation: Drug design requires simultaneously optimising for binding affinity (does it hit the target?), selectivity (does it hit only the target, not off-targets?), solubility (can it be formulated?), metabolic stability (does the body break it down too fast?), and synthesisability (can a chemist actually make it?). IsoDDE handles these as a joint problem rather than iterating sequentially — collapsing months of iterative medicinal chemistry into a computational cycle.

The $2.1B Round: Who Invested and Why

Thrive Capital (lead, USA): Josh Kushner's venture fund has become one of the most active investors in frontier AI. Leading Isomorphic's Series B follows Thrive's OpenAI investment. The thesis is the same: AI-native companies — not legacy industries digitising themselves — capture the value creation from AI in each sector.

Alphabet and GV (existing, USA): Alphabet is Isomorphic's founding backer and parent company. Their continued participation signals that Isomorphic is not being absorbed into Google but is being backed as an independent company with its own capital structure. GV's involvement adds healthcare-specific investment expertise.

MGX (new, Abu Dhabi): Abu Dhabi's AI-focused sovereign investment vehicle is building a portfolio across frontier AI sectors. Healthcare AI has global demand regardless of geography — MGX's participation mirrors its pattern of backing infrastructure-level AI bets.

Temasek (new, Singapore): Singapore's sovereign wealth fund has made AI healthcare a declared strategic priority. Temasek's investment reflects Singapore's long-term commitment to biomedical technology — it built Biopolis, its biotech research hub, over 20 years. A stake in the AI company attempting to transform drug discovery is a natural extension of that strategy.

UK Sovereign AI Fund (new): Direct UK government participation signals British industrial policy conviction that Isomorphic — a UK-founded, London-headquartered company — should remain a UK AI champion. The UK Sovereign AI Fund was established to make exactly this class of investment.

CapitalG (new): Alphabet's independent growth fund, which operates separately from Alphabet's direct corporate investments. CapitalG participating alongside Alphabet in the same round suggests the conviction about Isomorphic's upside is broad within the Alphabet orbit.

Why No Clinical Trials Yet

MedCity News captured the situation accurately: Isomorphic raised $2.1 billion without disclosing a single detail about its drug pipeline. No compound names, no target diseases, no Phase I timeline beyond "by end of 2026."

The opacity is partly competitive: announcing which diseases you are targeting tells rival drug companies which proteins you are working on and which pharma partnerships you are likely pursuing. Drug discovery is an enormously competitive field and that information has direct commercial value.

But the more important context is the missed deadline. Hassabis had previously committed to getting an AI-designed drug into human clinical trials by end of 2025. That deadline passed. The $2.1 billion is partly a response to that missed milestone — it provides the capital to run the preclinical development required before an Investigational New Drug application can be filed with the FDA.

To enter a Phase I human trial, Isomorphic must complete animal toxicology studies, formulation work, manufacturing scale-up to GMP standards, and regulatory submission. For a first compound from a novel AI-designed pipeline, that process is not fast. The new commitment — first patient dosed before end of 2026 — is the revised public accountability marker.

The "Solve All Disease" Ambition

Hassabis has used the phrase "solve all disease" in multiple public statements. This reads as hyperbole but has a specific technical argument behind it.

Most diseases involve proteins malfunctioning — overactive, underactive, misfolded, binding to the wrong partner. If you can computationally design a molecule that precisely corrects the specific protein malfunction underlying any disease, you have a general solution to the class of problems that constitutes most of medicine.

AlphaFold gave the field an accurate map of protein structure. IsoDDE is the attempt to build the general tool for intervening in that map. The 20-year vision is a platform where any disease with a known protein target can be approached computationally with the same speed and scale that software engineering applies to code. The 2026 milestone is dosing one patient with one compound. The gap between those two endpoints is where the $2.1 billion lives.

What Developers and Engineers Should Know

For software engineers building in healthcare: Isomorphic's clinical trial result — when it comes — will be the clearest test of whether generative AI can produce validated scientific discoveries rather than pattern-matching on existing knowledge. A Phase I pass means AI-designed molecules work in human biology. The implications for every AI-for-science project extend far beyond drug discovery.

For AI engineers interested in the domain: IsoDDE's capabilities — structure prediction, generative molecular design, multi-objective optimisation, cryptic pocket identification — are combinations of techniques that are well-documented in research literature. AlphaFold 3's architecture is published. This is an engineering and data problem as much as a pure research problem.

For engineers in Singapore, the UK, or Abu Dhabi: Temasek, the UK Sovereign AI Fund, and MGX participating in this round reflects government-level commitment to AI healthcare infrastructure in those geographies. Isomorphic is hiring in London, Cambridge Massachusetts, and Lausanne. The investor geography predicts where the talent demand follows.

Key Takeaways

• $2.1B Series B closed May 12, 2026: Led by Thrive Capital; Alphabet, GV (existing); MGX (Abu Dhabi), Temasek (Singapore), CapitalG, UK Sovereign AI Fund (new); largest AI biotech round of 2026
• Demis Hassabis: DeepMind co-founder, 2024 Nobel Prize in Chemistry for AlphaFold; CEO of Isomorphic Labs since 2021 founding; goal: "solve all disease"
• IsoDDE capabilities: Cryptic binding pocket identification from protein sequence alone; generative molecular design (new molecules, not library screening); multi-objective optimisation across binding, selectivity, solubility, synthesisability; built on AlphaFold 3 structure prediction
• Pipeline: Cancer and immunology; pharma partnerships with Novartis, Lilly, Johnson and Johnson; zero patients dosed as of May 2026
• Clinical trial target: First human patient dosing before end of 2026; previous end-of-2025 deadline was missed; IND filing and preclinical completion required before first dosing
• Geopolitical signal: US, UK, Abu Dhabi, and Singapore sovereign and VC capital all in the same round — AI drug discovery is government-level strategic priority across four major economies simultaneously

For the TSMC $1.5 trillion chip market forecast that provides the compute infrastructure powering AI drug discovery, read TSMC: $1.5 Trillion Chip Market by 2030, AI at 55%. For the broader AI infrastructure story and what Anthropic's $950B valuation signals about where AI compute is going, read Anthropic Mythos: macOS Exploit in 5 Days, $950B Valuation.