Anthropic surpasses DeepMind: a silent overthrow in the race for AI labs for health

In early 2025, if someone said that Google DeepMind, the cradle that created AlphaFold and was once considered the most powerful AI science lab in the world, would have to follow Anthropic in the field of drug discovery, I think most people in the industry would laugh. But actually, that's what's happening in mid-2026.

Pricing and money say it all

Anthropic has just raised capital at a valuation of $965 billion, surpassing OpenAI, and is considering an initial public listing (IPO) as early as this fall. The company's run-rate revenue surpassed $30 billion, up from about $9 billion at the end of 2025. In other words, in less than a year, Anthropic's revenue more than tripled. This is not the growth of a startup still finding a business model, but the growth of a company that has found a formula and is expanding at full capacity.

Brain drain at Google DeepMind

But the most impressive number lies in the personnel. In recent months, DeepMind has continuously lost people to competitors, and Anthropic is the most popular destination. John Jumper, who won a Nobel Prize for his work on AlphaFold and held the role of director at DeepMind, announced his departure to join Anthropic. Soon after, two other people who also contributed to AlphaFold, Jonas Adler and Alexander Pritzel, were also said to be preparing to leave Google to join Anthropic. Broadly speaking, DeepMind's effort to pivot to programming models caused the company to lose a total of six researchers, going to Meta, OpenAI and Anthropic, while Noam Shazeer, another veteran name, chose OpenAI.

Stepping into DeepMind's home turf and Washington is also choosing sides

What makes this story more interesting is that Anthropic is not only competing in the language modeling field, but is quietly building tools for drug discovery and scientific research processes, exactly the territory that DeepMind has monopolized since AlphaFold was born. When the lab that was once the absolute leader in AI applications for science now has to play catch-up, it's a reversal that almost no one predicted a few years ago.

Another piece of the puzzle just appeared at the right time, the US government recently quietly reversed one of the most important AI technology export restrictions of the year, and the main beneficiary is Anthropic. This move is seen as a green light for Anthropic to expand globally, right at a time when China's leading AI models are narrowing the technology gap. The most reasonable reading is that Washington is choosing to block the door to China, but selectively opening the door to American labs that have proven their ability to ensure safety, and Anthropic is standing in that position.

So who is really winning?

If only looking at the strongest model on paper, Google's Gemini is still Claude's formidable opponent. But the race for AI labs in 2026 is not only measured by benchmark scores, it is also measured by capital raising speed, ability to retain people, and relations with the government, which are three things that Anthropic is clearly winning over DeepMind. A company that spun out of OpenAI in 2021 is now putting Google's oldest and most resource-rich AI lab on the defensive, which is something that's worth making for those who follow the AI ​​industry to stop and think seriously, rather than just seeing this as a passing personnel news. The open question is whether, when Anthropic actually IPOs, pressure from public shareholders will slow the company in its trademark approach to AI safety, that will be the real test for the next phase.