14.3 billion yuan, the AI pharmaceutical industry has just received its largest single financing.

On May 12, 2026, in London, Isomorphic Labs, a subsidiary of Google's parent company Alphabet focused on AI drug development, announced the completion of a $2.1 billion Series B financing, equivalent to approximately 14.3 billion yuan.

Thrive Capital led the investment, with Alphabet and GV participating. New investors include Abu Dhabi's MGX, Singapore's Temasek, and the UK's sovereign AI fund. Just a year ago, the Series A financing was only $600 million, but the Series B financing more than tripled, setting a record for single - round financing in AI drug development.

Isomorphic Labs was spun off from Google DeepMind in 2021. Demis Hassabis, the CEO, also heads DeepMind.

The foundation of this company is AlphaFold, a protein structure prediction model that emerged in 2020 and is hailed as having solved a "50 - year biological puzzle" at one stroke. In 2024, Hassabis and his colleague John Jumper won the Nobel Prize in Chemistry for this achievement.

The Nobel Prize is the ultimate endorsement in the scientific community, and this $2.1 billion is the same vote cast by capital, but with a much larger denomination.

This money will be spent in three directions: continue to develop the AI drug design engine IsoDDE, accelerate the internal pre - clinical pipeline, and expand the AI, engineering, and drug design teams globally. Translated into one sentence: Spend money for speed and bet on an unproven future.

01

From Auxiliary Tool to Infrastructure

The entry of AI into drug discovery did not start in 2026.

As early as the 1980s, computer - aided drug design was a common tool for scientists. However, CADD relies on human - preset rules, and its capabilities are always limited. The change brought by AI is not to replace it, but to redefine the role of "computation" in drug discovery.

Around 2010, machine learning began to penetrate drug development, mainly focusing on virtual screening, ADMET prediction, etc. In 2016, deep learning pushed the capabilities of molecular generation, activity prediction, and other aspects to a new level.

The real qualitative change occurred after the emergence of AlphaFold: it proved that deep learning can solve the most difficult structure prediction problem in biology, with an accuracy approaching that of experimental methods. The first hurdle in drug discovery, understanding the three - dimensional structure of target proteins, was overcome by AI at extremely low cost and high speed.

Since then, the industry has evolved at an accelerated pace.

At the beginning of 2026, Eli Lilly and NVIDIA announced an investment of $1 billion to establish a joint AI drug discovery laboratory. On April 16, 2026, OpenAI launched GPT - Rosalind, specifically designed for biology and drug discovery, directly entering the pharmaceutical track.

This is a typical new - generation infrastructure competition in biopharmaceuticals. In 2026, AI drug development has crossed the line from an "optional auxiliary tool" to a "mandatory core productivity."

02

The Ambitions and Blind Spots of IsoDDE

AlphaFold solves the problem of "static structure prediction": given an amino acid sequence, it outputs a three - dimensional structure. However, drug design requires much more. Molecules not only need to bind, but also bind strongly; they need to be effective against the target and harmless to "off - targets."

AlphaFold can handle the first part, but the subsequent parts are beyond its scope. Isomorphic Labs' response is IsoDDE: a drug design engine.

In challenging benchmark tests, this system's accuracy in predicting protein - ligand structures is more than twice that of AlphaFold 3. It can predict binding affinity, that is, the binding strength between the drug and the target, with an accuracy comparable to the previous gold - standard physical simulation method, but at a much faster speed.

It can also identify drug - binding pockets based solely on amino acid sequences, including hidden pockets that the scientific community has not noticed for more than a decade. Isomorphic Labs attempts to complete the first half of the drug discovery process, including "discovering targets, designing molecules, and predicting activity," in a single system.

However, the technical bottlenecks are also obvious.

Currently, AI still lacks generalization ability in new protein systems. Once it encounters a biological system not seen in the training data, the prediction accuracy will decline significantly. The physical laws of protein - ligand interactions are extremely complex, and AI models have so far been unable to fully capture their dynamic characteristics.

More fatally, all AI models are limited by the bias of training data - the known drug - target interaction data is severely skewed, and there is a lack of high - quality annotations for a large number of potential targets. No matter how good the model is, it is fed with the same incomplete textbook.

03

"Drug Discovery = A Computable Engineering Problem"

The $2.1 billion is an extremely expensive bet.

As of May 2026, no drug designed by AI has truly entered the clinical stage globally, let alone been successfully launched on the market.

Even Isomorphic Labs itself has not advanced any AI - designed molecules to clinical trials.

Hassabis predicted in 2025 that "candidate drugs will enter clinical trials by the end of the year," which was later revised to "pre - clinical research." According to Bloomberg, during the World Economic Forum in Davos in January this year, he postponed the schedule for the first clinical trial from the end of 2025 to the end of 2026.

This is not a failure, but it is indeed calibrating the market's expectations: the last mile of AI drug development is longer than everyone initially imagined.

The situation of other players in the industry is even more disheartening.

Recursion Pharmaceuticals has fallen by nearly 90% from its peak since its listing in 2021, with a market value of approximately $1.9 to $2 billion, almost returning to the valuation of a medium - sized biotech company. NVIDIA completely liquidated its shares in RXRX in February 2026. UK - based BenevolentAI announced its delisting from the Euronext Amsterdam in late 2024 and carried out large - scale layoffs and cost - cutting.

After the high - foam period, no one can answer when the high returns will appear.

The $2.1 billion can do many things. It can cover most of the expenses of an entire clinical development chain or directly acquire several medium - sized biotechs. However, Isomorphic Labs chose to bet all this money on an extremely expensive underlying assumption: drug discovery can be transformed into a computable engineering problem.

The underlying message is: as long as the model is good enough, the data is sufficient, and the computing power is strong enough, designing effective drugs can be as automated as designing chips.

If this assumption holds, $2.1 billion is just an entry ticket. If not, it will be the most expensive philosophical trial - and - error in the AI era.

04

Charge In or Be Left Out

If Isomorphic Labs succeeds, drug discovery will shift from "screening" to "design," which will be a great paradigm revolution.

The current average cycle of drug development is 12 to 15 years, and the average cost exceeds $2 billion. 90% of candidate drugs fail in clinical trials. If AI can improve R & D efficiency by even a few percentage points, it will save hundreds of billions of dollars for the global industry each year.

However, the path chosen by Isomorphic Labs is completely different from that of traditional pharmaceutical companies.

Traditional pharmaceutical companies' drug discovery relies on scientists' experience and a large number of experiments, which are labor - intensive tasks in wet laboratories; Isomorphic Labs' methodology is a purely digital "design - predict - validate" cycle.

AI designs molecules in the silicon - based space, predicts activity, toxicity, and pharmacokinetic properties, selects the optimal candidates, and then verifies them in a small number of wet laboratories. If this path can be successful, the "wet experiment" part of drug discovery will be systematically reduced. This is the essence of what Hassabis calls "AI - first drug design."

The capital market sees the possibility, but history does not easily give answers.

A hundred years ago, chemist Paul Ehrlich proposed the concept of "magic bullets," foreshadowing the era of precision - targeted drugs - this era took three - quarters of a century to truly mature. Thirty years ago, gene therapy attracted a frenzy of capital, but a serious adverse event set the entire industry back by a decade. AI drug development has just passed the initial fanatical period and is still a long way from real large - scale clinical validation.

This $2.1 billion is a bet on an unproven assumption: AI can evolve from "predicting protein structures" to "designing all effective drugs"; it is a bet that the complexity of drug discovery can be reduced to a computable engineering problem.

In the history of science, such bold bets have won. The Human Genome Project ultimately brought about the golden age of the entire biotech industry. They have also lost. Several AI winters buried countless ambitious laboratories. The final outcome is determined not by the scale of funds, but by the real - world efficacy data.

Isomorphic Labs is betting that a paradigm revolution in drug discovery is underway. The $2.1 billion is the entry ticket to this revolution.

For Chinese innovative pharmaceutical companies, what really deserves attention from this financing is not the amount, but a completely different R & D philosophy taking shape: when the world's top AI laboratories start using engineering thinking to solve drug discovery problems, latecomers still using linear thinking to catch up will face not competition, but generational elimination.

The window of the paradigm revolution only opens once. Charge in or be left out. Time waits for no one.

This article is written based on publicly available information and is for information exchange only. It does not constitute any investment advice.

This article is from the WeChat official account "Medical Shine", author: Uncle Yao. It is published by 36Kr with authorization.