llms.blog
Google DeepMind

Why DeepMind Researchers Are Leaving: Limited TPU Access, Bureaucracy, and a Chips Conflict

As senior AI researchers continue to leave Google DeepMind, reporting is surfacing reasons behind the departures. Two recurring themes: limited access to Google's TPU chips, and a conflict of interest where Google Cloud sells those same chips to rivals the researchers are expected to beat. Semafor reported that CEO Demis Hassabis stepped back from daily operations about a year ago, turning those duties over to DeepMind chief Koray Kavukcuoglu. Hassabis was not forced out, but reportedly found m

axc
axc
2 min
Why DeepMind Researchers Are Leaving: Limited TPU Access, Bureaucracy, and a Chips Conflict

As senior AI researchers continue to leave Google DeepMind, reporting is surfacing reasons behind the departures. Two recurring themes: limited access to Google's TPU chips, and a conflict of interest where Google Cloud sells those same chips to rivals the researchers are expected to beat.

Semafor reported that CEO Demis Hassabis stepped back from daily operations about a year ago, turning those duties over to DeepMind chief Koray Kavukcuoglu. Hassabis was not forced out, but reportedly found management unfulfilling and sees himself more as a visionary scientist than an executive.

Mid-century industrial design concept of scarce compute resources allocated unevenly between two recipients

CNBC reported that many researchers are frustrated by limited access to Google's TPU infrastructure. Google allocates computing capacity years in advance across research teams, product operations, and cloud customers, but priorities can shift with little notice, a source told CNBC. The reporting also points to Google's internal bureaucracy as a reason younger companies look more appealing, alongside the fact that Google Cloud sells TPUs to competitors such as Anthropic.

External customers ahead of internal teams

The timing is awkward for Google. The same week it announced the departure of Jeff Dean after 27 years and Hassabis stepping back, it said frontier AI lab Mirendil will use over $100 million worth of TPUs and Nvidia GPUs through a Google Cloud partnership. The arrangement puts external customers ahead of some internal researchers in line for the chips those researchers helped build the demand for.

Google has not publicly disputed the reporting. The departures add to a stretch where DeepMind has lost both a CEO and a chief scientist, leaving its talent pipeline and chip allocation under scrutiny as demand for compute outpaces supply.

Sources

DeepMind's talent drain likely comes down to chip shortages, a conflict of interest, and Google's bureaucracy - The Decoder: https://the-decoder.com/deepminds-talent-drain-likely-comes-down-to-chip-shortages-a-conflict-of-interest-and-googles-bureaucracy/

Demis Hassabis was shifting away from DeepMind CEO duties for a year - Semafor: https://www.semafor.com/article/08/05/2026/demis-hassabis-was-shifting-away-from-deepmind-ceo-duties-for-a-year

Google is expanding its AI empire and losing the people who built it - CNBC: https://www.cnbc.com/2026/08/05/google-is-expanding-its-ai-empire-and-losing-the-people-who-built-it.html

Mirendil selects AI Hypercomputer - Google Cloud: https://cloud.google.com/blog/topics/startups/mirendil-selects-ai-hypercomputer?hl=en

Google DeepMindAI LabsAI Hardware
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min

  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min

  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min