HomeArticle

The man who defeated Ke Jie raised 33.3 billion yuan in financing, breaking the record.

铅笔道2026-11-20 18:22
From defeating Ke Jie to discovering new drugs, Demis Hassabis has secured 33.3 billion yuan in financing.

Nine years ago, Ke Jie, a professional 9-dan Go player, lost three consecutive games to the artificial intelligence AlphaGo in Wuzhen.

Nine years later, the man behind AlphaGo has set another record in the capital market.

On November 17, Isomorphic Labs, an AI pharmaceutical company founded by Demis Hassabis, announced the completion of a $5 billion Series C financing, equivalent to about 33.3 billion yuan. This financing has refreshed the record for the largest single round of financing in the AI pharmaceutical sector.

Adding the previous two rounds of financing, the company, which was founded only five years ago, has raised a total of 7.7 billion U.S. dollars, equivalent to about 51.3 billion yuan.

Playing Go does not make money, but drug development does

Hassabis, 50 years old this year, is one of the most legendary figures in the field of artificial intelligence.

He was once a chess prodigy, ranking among the world's top players of his age at the age of 13. At 17, he participated in the development of the game Theme Park, later went to the University of Cambridge to study computer science, and obtained a doctorate in cognitive neuroscience.

In 2010, he co-founded DeepMind with two partners, trying to enable computers to learn and think like humans.

Four years later, Google acquired DeepMind for approximately 400 million pounds.

It was AlphaGo that made Hassabis truly famous.

In 2016, AlphaGo defeated the renowned South Korean Go player Lee Sedol by a score of 4 to 1. The following year, it beat Ke Jie 3-0 in Wuzhen.

This victory made the world intuitively feel the capabilities of artificial intelligence for the first time.

After that match, Hassabis withdrew AlphaGo from official competitions and turned to protein research.

In the past, resolving the 3D structure of a single protein could take months or even years. In 2020, DeepMind launched AlphaFold 2, which significantly improved prediction accuracy. Since then, more than 200 million predicted structures have been made available to researchers.

In 2024, Hassabis and his colleague John Jumper won the Nobel Prize in Chemistry for this achievement.

However, figuring out proteins is only the beginning of drug development.

If we regard disease-causing proteins as locks, AlphaFold helps scientists see the keyholes clearly; to actually develop drugs, we still need to design a key that can not only open the lock but also avoid harming other parts of the body, and then go through animal experiments and human clinical trials.

In 2021, Alphabet established Isomorphic Labs, allowing Hassabis to find such "keys" from a large number of molecules and then send them to the laboratory.

In August this year, 50-year-old Hassabis stepped down as CEO of Google DeepMind, and took up the positions of Chairman and Chief Scientist of Alphabet, while continuing to lead Isomorphic Labs. Three months later, the company secured 33.3 billion yuan in financing.

No drugs have been developed yet, but pharmaceutical companies have already sent capital

Another reason why capital dares to bet is that Isomorphic Labs already has a group of large clients.

Eli Lilly, Novartis, Johnson & Johnson, and Merck & Co. are all global pharmaceutical giants.

In January 2024, Eli Lilly and Novartis signed contracts with it successively. Eli Lilly paid an upfront payment of 45 million U.S. dollars, with follow-up amounts up to 1.7 billion U.S. dollars; Novartis paid an upfront payment of 37.5 million U.S. dollars, with follow-up amounts up to 1.2 billion U.S. dollars, plus subsequent revenue sharing from drug sales.

In 2025, Novartis added up to three additional R&D projects. In January 2026, Johnson & Johnson joined the cooperation list. Just two days after this round of financing, Merck & Co. also reached a cooperation with Isomorphic Labs on oncology drug R&D.

It helps pharmaceutical companies identify and optimize candidate drugs: it charges upfront payment at the start of the project, collects fees again after completing agreed milestones, and continues to share revenue after the drug is approved for marketing.

The most special part of this business is that part of the trial and error process can be moved into the computer.

Traditional new drug R&D requires continuous molecular design, sample synthesis, experiment conduction, and modification based on results, and each round usually takes 1 to 3 months. An internal case disclosed by Isomorphic Labs shows that its system calculates for 2 to 4 days first, then selects a batch of molecules for scientists to synthesize and test.

In this way, a batch of molecules with little hope can be eliminated in advance.

If this process can be achieved repeatedly, the value will not come only from a single drug.

Greenoaks and HSG are willing to lead the investment, mainly because Isomorphic Labs does not only aim to develop one drug. It hopes to advance multiple candidate drugs at the same time. If one fails, there are other projects that can continue.

Pharmaceutical companies do not pay out tens of billions of dollars in the contract at one time.

The potential total amount of the two cooperation agreements with Eli Lilly and Novartis is close to 2.9 billion U.S. dollars, but the total upfront payment paid at the time of signing is only 82.5 million U.S. dollars, less than 3%. The remaining money will be paid gradually after the candidate drug enters the clinical stage, gets approval, or even generates sales. If the project fails halfway, the subsequent payment will not be received.

Cooperation in the industry basically adopts this payment method. According to statistics from McKinsey, the total value of AI drug R&D cooperations announced by the world's top 20 pharmaceutical companies in 2025 reached 29.5 billion U.S. dollars, but the actual upfront payment was only 600 million U.S. dollars, accounting for about 2%.

Therefore, the cooperation worth tens of billions of dollars in the news does not mean that AI pharmaceutical companies have already earned that much money. The company can only get more payment when the drug moves forward one step.

Isomorphic Labs is currently at this stage. In January this year, Hassabis once said that the company's first candidate drug is expected to enter human clinical trials by the end of the year. But by November, he no longer confirmed this timeline, only stating that the regulatory declaration is still in progress, and more news will not be available until 2027.

The financing has been secured, and pharmaceutical clients have also arrived. The next step is to see when the first drug will actually enter human trials.

Large pharmaceutical companies are short of new drugs, and tech giants enter the market

Another practical reason for capital to increase investment in AI pharmaceuticals is that large pharmaceutical companies need to supplement their new drug pipelines.

IQVIA estimates that by 2030, more than 230 billion U.S. dollars of global biomedical revenue will be affected by patent expirations. After generic drugs enter the market, the price and sales volume of original drugs may decline. As old drugs are about to expire and new drugs come slowly, pharmaceutical companies can only conduct independent R&D, acquire companies, or cooperate with external teams.

However, new drug R&D is inherently a slow business.

a16z listed a set of data in its investment analysis: a successful drug usually takes about 13 years, and the total investment is close to 2 billion U.S. dollars when calculated by the time cost of capital; the success rate from candidate drug to FDA approval in the United States is less than 5%.

This institution believes that if a company can repeatedly identify candidate drugs, the market may no longer only value the company based on a single pipeline, but also pay for its entire R&D methodology. This explains why the valuation of Isomorphic Labs has risen, and also points out the risk: "repeated success" still needs to be verified by clinical results.

It is not only Google that is eyeing this business.

According to media reports, Anew Labs, which was spun off from ByteDance earlier this year, completed a $290 million financing with a post-investment valuation of $1.5 billion. Institutions including GL Ventures, IDG Capital, and 5Y Capital participated in the share subscription, and ByteDance still holds about 56% of the shares.

Mithril Medicine, acquired by Anthropic, has only one full-time employee. The founder uses Claude to organize data, design experiments, and coordinate 11 external laboratories. The team obtained mouse experiment results in less than 18 months, but it is still far away from human trials.

Insilico Medicine, which entered the market earlier, has moved further along the pipeline. After its candidate drug Rentosertib completed the Phase IIa study, it launched the Phase III trial in September this year, planning to recruit 320 patients and observe for 52 weeks. It has not been approved yet, but it has begun to undergo larger-scale human testing.

The entire industry is waiting for this step.

A statistic as of March 2026 shows that a total of 129 AI-discovered candidate drugs from 69 companies around the world have entered human trials, 72 have completed Phase I, 24 have completed Phase II, and only 5 have entered or completed Phase III. So far, no drug entirely discovered by AI has been approved for marketing.

In other words, AI pharmaceutical development has moved from the stage of "whether molecules can be found" to the stage of "whether these molecules can really cure diseases". The former question can be answered by computing speed, experimental cases and cooperation contracts, while the latter can only be verified by patients, time and clinical data.

Nine years ago, AlphaGo played three games of Go with Ke Jie, and the result was revealed quickly.

Drug development has no such fast game. 33.3 billion yuan can buy top scientists, more experiments and more candidate molecules, but it cannot buy regulatory approval, nor can it buy real results in patients.

For Isomorphic Labs, the next truly important news is no longer how much more financing it can raise, but when the first batch of candidate drugs will enter human trials, and whether they can prove their efficacy and safety.

This article does not constitute any investment advice.

This article is from the WeChat Official Account "Pencil News" (ID: pencilnews), written by Xin Xiao, and published with authorization from 36Kr.