On Wednesday, August 5th, Google announced a complete overhaul of DeepMind leadership. A quick recap:
Demis Hassabis, DeepMind co-founder and former CEO, is no longer involved in day-to-day operations.
Jeff Dean, former Google Chief Scientist and Gemini co-lead is leaving to start a neolab called Discovery Loop. Jeff is the undisputed GOAT of Google engineering, co-founded Google Brain, and started the TPU program.
Joining Jeff are Sanjay Ghemawat, Quoc Le, and Oriol Vinyals. Sanjay and Quoc were both Google Fellows, which is a title reserved for the company’s top dozen or so technical contributors.1 Oriol was a Gemini co-lead.
Koray Kavukcuoglu, former DeepMind CTO and the one remaining Gemini co-lead, is replacing Demis as the leader of DeepMind/Gemini.
Obviously these are not the actions of people excited about Gemini 4 Pro.
For all intents and purposes, we believe DeepMind is no longer a frontier lab. We said as much a few months ago to our Tokenomics clients due to large numbers of departures from their reinforcement learning teams and poor compute allocation. Google will continue meandering on and releasing models, but their odds of reaching SOTA again have dropped to zero.
Furthermore, the biggest beneficiary of today’s news is neither Anthropic nor OpenAI—it’s Google Cloud. Whereas Gemini and GCP used to desperately fight for compute allocation, it’s now clear that Thomas Kurian won. We expect GCP revenue growth to meaningfully accelerate as a result. …
There’s no substitute for conviction
We believe Gemini’s core issue has always been a fundamental lack of conviction. Compute is the lifeblood of AI progress, and all the AGI-pilled labs are desperately trying to acquire as much as possible.
Of course, pre-paying multiple GWs is also very scary and expensive, especially if you don’t have a rapidly growing, high margin API business. However, as we’ve explained in previousnewsletters, there are creative ways to overcome this limitation and still monetize excess capacity at very attractive rates while maintaining the optionality to claw everything back for your research team. Explicitly, this is what we see happening at Meta and SpaceX.
Google, on the other hand, decided it was totally worth it to sell enormous amounts of compute to Gemini’s fiercest competitors on long term contracts without any hope of ever returning it to DeepMind.
More than 20% of total TPU shipments from 3Q26 to 4Q27 are being sold directly to Anthropic.2 This is excluding the hundreds of thousands of TPUs GCP already rents to Anthropic today, and the many hundreds of thousands more they’ve committed to rent to Anthropic and Meta over the next 6 quarters. …
If you’ve ever listened to an interview of Google Cloud CEO Thomas Kurian, you know he is not AGI pilled. In one podcast, for example, he argued that it’s great for TPUs to become “general purpose infrastructure” that supports customers like Citadel, the Department of Energy, and generic high performance computing. And when asked why he was selling compute to Anthropic despite them competing with Gemini, he said this was the natural consequence of Google being a “platform company.”
This same man likely just won a major internal political battle to have even more control over Google’s compute allocation.
Attempting to steelman DeepMind’s frontier ambitions
We’ve obviously been quite bearish on DeepMind thus far, and if we had to steelman the case for why they’ll still be able to train a true SOTA model in the future, it would go something like the following:
The current setup clearly wasn’t working. With the existing leadership team, their odds of catching up to Anthropic/OpenAI looked extremely slim.
Now that they’ve cleaned house, the new guys can start from a blank slate. Maybe they’ll even acqui-hire a neolab like SSI or Thinking Machines.
With this new team, their odds of catching up to the frontier actually increase.
Perhaps there’s some world in which this happens, but we think the odds are basically zero. The issue with Google was not Jeff Dean nor Noam Shazeer, but rather their extremely bureaucratic, painfully slow, and strategically timid culture. Remember that DeepMind had an AI chatbot 1 year before ChatGPT but was not allowed to release it due to fears of disrupting their core business.
Tibo, funny enough, is another great example of brain drain from Google. A 9 year Google/DeepMind veteran, he left in July 2024 to join OpenAI, co-led the original coding agent team, and is now the head of Codex. Meanwhile, Google “acquired” Windsurf to build its own agentic coding platform in July 2025, and they don’t exactly have much wind in their sails thus far.
Let’s also not forget that DeepMind was the original AI acqui-hire! And with the benefit of 12 years’ hindsight, we now know how that played out. Adding another neolab to the roster would likely just result in more of the same.
SemiAnalysis commentary on GDM no longer being a frontier lab from their POV:
Selling shovels :-) I wonder if they’ll take the next step and try to commoditize their complement by supporting open models.