DEMIS.

In 2010, a chess prodigy and neuroscience PhD founded an AI lab in London with two colleagues. Demis Hassabis had been a child chess master, a video game designer, a computational neuroscientist. He believed the mind could be reverse-engineered. He believed that intelligence — real intelligence, not pattern matching — could be built. He called the company DeepMind. Google acquired it in 2014 for approximately $600 million. At the time, most people in Silicon Valley had not heard of it.

DeepMind's first decade was benchmark-setting. AlphaGo defeated Lee Sedol in 2016 — the moment the AI race became publicly legible. Not because Go was strategically important, but because humans understood, in one game, that something genuinely new was in the room. AlphaZero followed. AlphaFold 1 followed. Each system solved a problem that had been considered too hard for machines. Each time, Demis was on stage saying: we are not done. The science is the goal. The capabilities are the instrument.

Then AlphaFold 2. In 2021, DeepMind published a system that predicted protein 3D structure from amino acid sequence with accuracy matching experimental methods. The protein folding problem had been unsolved for 50 years. It was one of science's Grand Challenges. Thousands of researchers had given decades to partial solutions. AlphaFold solved it. Not incrementally. Not approximately. Solved it. The DeepMind team released the weights and made the database free. Over 200 million protein structures are now publicly accessible. Drug discovery, disease research, biology as a field — the downstream effects are still accumulating.

In October 2024, the Nobel Committee called Stockholm. Demis Hassabis and John Jumper received the Nobel Prize in Chemistry — shared with David Baker of the University of Washington — for computational protein structure prediction and protein design. It was the first Nobel Prize awarded for work done at a private AI lab. Demis accepted it as a scientist. He said the work was real science, not engineering. He was right. The prize committee agreed.

That same year, Google had merged Google Brain and DeepMind into Google DeepMind, with Demis as CEO. The company that acquired his lab for $600M ten years earlier now runs its entire AI research operation through him. Gemini — Google's frontier model — came from that merged organization. The AI race that Altman and Amodei and Zuckerberg run in public has, at the Google seat, a man who held a Nobel Prize while writing a roadmap to AGI. The field notes: none of the other competitors have a Nobel Laureate running their AI lab. None.

One co-founder left. Mustafa Suleyman departed DeepMind in 2019, founded Inflection AI, then joined Microsoft in 2024 as CEO of Microsoft AI. The original trio is now distributed across the three most important AI deployments in the world: Google DeepMind (Demis), Microsoft AI (Mustafa), and Shane Legg — still at Google DeepMind as Chief AGI Scientist. DeepMind seeded the race from one office in London. The race is now global. They know how it started. They are watching where it goes.

The AI race has four main seats. OpenAI: Sam Altman, nonprofit converted to for-profit, Saudi $40B incoming, safety team gone. Anthropic: Dario Amodei, left OpenAI over safety, raised $18B, racing anyway. Meta: Mark Zuckerberg, bought 49% of Scale AI, installed Alexandr Wang as Chief of Superintelligence. Google DeepMind: Demis Hassabis, Nobel Laureate, 15 years of published science, Gemini 2.0, AlphaFold, still publishing.

The field does not pick winners. The field notes: three of these four seats are occupied by people who are primarily operators, dealmakers, or product thinkers. One is occupied by someone who was trained as a neuroscientist, published AlphaFold, and received a Nobel Prize for it. That is not a moral credential. It is a scientific one. The field notes the difference.

The quiet one at the table usually knows something the loud ones don't. That's not always true. But when it is, you see it in what gets published versus what gets announced. DeepMind publishes. Google DeepMind publishes. The Nobel Committee agreed with the publication record. The field notes that too.