Gemini is Cooked but GCP is Cooking

SubscribeSign in

Gemini is Cooked but GCP is Cooking<br>GCP YoY rev growth >100%, DeepMind's long term failure is Google Cloud's short term gain<br>Max Kan, Joey Brookhart, Doug O'Laughlin, and Dylan Patel<br>Aug 07, 2026<br>∙ Paid

61

Share

On Wednesday, August 5th, Google announced a complete overhaul of DeepMind leadership. A quick recap:<br>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.<br>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.<br>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 .<br>Gemini 3 Pro was the peak

At the end of 2025, Anthropic, OpenAI, and Google were the clear AI big 3. Some still held this opinion as recent as last week, but the downfall of Gemini in 2026 was clear to those paying attention. Here’s an excerpt from an institutional note we published to our Tokenomics Model subscribers on July 9th :<br>In November 2025, Gemini 3 Pro was arguably the best model in the world, forcing Sam Altman to issue a “code red” at OpenAI.

Since then, the gap between Deepmind and OpenAI/Anthropic has widened dramatically. Gemini 3.5 Flash was a total flop, and industry chatter suggests the upcoming Gemini 3.5 Pro is roughly Opus 4.5 level . This is quite literally worse than GLM 5.2, much less Fable 5 and GPT 5.6.

We believe things will only get worse for Gemini, as Google lacks the religious conviction necessary for building RSI. We think it’s likely both MSL and SpaceXAI surpass them in model quality by the end of this year.

Along with Noam Shazeer and John Jumper, most of the best RL people at Gemini recently left the company.

OpenAI has overcome their pre-training issues, and a much larger model code named “Doug” is actively in the works.

The second to last point is worth emphasizing. Jeff, Sanjay, Quoc, and Oriol are just the latest in a long string of high profile departures from DeepMind. Despite discovering most of the foundational breakthroughs behind the current LLM paradigm, Google is now simply unable to retain top AI talent.<br>Notably, Jeff and co. have followed a playbook to create their neolab, Discovery Loop, that was established by David Silver of Ineffable Intelligence in November last year. To pursue their mission, top AI researchers are now choosing to leave Google, raise billions from outside investors and Google Ventures, and then spend that money on Nvidia GPUs in GCP.<br>Since our institutional note, Google has silently canceled Gemini 3.5 Pro and is now coping by hyping up Gemini 4 instead. As a bridge model, they released Gemini 3.6 Flash, but it’s generally worse than Muse Spark 1.2, Grok 4.5, and the tier 1 Chinese open-source models. Depending on how you count, Gemini is currently in 8th or 9th place, and we don’t see Gemini 4 reversing their downfall.

Source: Artificial Analysis<br>At the same time, Gemini first-party API token growth has famously decelerated. Whereas they grew 60% in 1Q26, going from 10B to 16B tokens per minute, 2Q26 saw just 38% growth to 22B. This has caused a corresponding decline in Gemini 1P API revenue growth.

Source: SemiAnalysis Tokenomics Model<br>However, Gemini Enterprise Agent Platform (formerly Vertex) has been meaningfully stronger overall thanks to third-party models like Claude. To see a full breakdown of Amazon Bedrock, Microsoft Foundry, and Google Vertex revenue separated by model provider, as well as all the implications for OpenAI and Anthropic, see our Tokenomics Model.<br>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.<br>Of course, pre-paying multiple GWs is also very scary and expensive, especially if you...