HomeFull speed ahead. Keep track of advancements in every frontier
Genie world modelsGoogle DeepMind—VeoGoogle DeepMind—Google DeepMind—GeminiGoogle DeepMind—Tensor Processing Unit (TPU)Google DeepMind—AlphaFoldGoogle DeepMind—Genie world modelsGoogle DeepMind—VeoGoogle DeepMind—GeminiGoogle DeepMind—Tensor Processing Unit (TPU)Google DeepMind—GeminiGoogle DeepMind—

Google DeepMind

Gemini, AlphaFold, Genie, Veo, TPU chips

GB/USAlphabet-ownedest. 2010deepmind.google (opens deepmind.google)Checked 25 Sep

The brief

Analyst view
  • Gemini 3 (Nov 2025) led many benchmarks. Gemini 3.5 Pro, announced in May 2026, is delayed.
  • AlphaFold predicted structures for over 200 million proteins and won a share of the 2024 Chemistry Nobel.
  • Google's own Ironwood TPU chip is used in-house and sold to outside labs, including Anthropic.

Technical approach

As reported
Multimodal

Built for all media

Gemini learns from text, images, audio, video and code together from the start, and can take in very long inputs.

RL + search

Practice and planning

From AlphaGo on, DeepMind pairs learned models with search and self-play. Deep Think tries several lines of reasoning at once.

Science

Models for science

AlphaFold (proteins), GNoME (crystals) and GraphCast (weather) build real-world rules in and check results against experiments.

Hardware

Its own chips

Gemini trains on Google's TPUs, linked by the thousand, so chips and models are designed together.

Programs

Google DeepMind, five programs

Updated 29 Jul 2026
Research · Proto · Pilot · Deploy · Scaling

Research & demos

5 items
  • May 2025
    paperAlphaEvolve: a coding agent for scientific and algorithmic discovery

    A Gemini-driven search found a faster way to multiply 4x4 matrices and improved Google's data centers.

  • Nov 2023
    paperScaling deep learning for materials discovery (GNoME)

    Predicted 2.2 million new crystals, about 380,000 of them stable.

  • Jul 2021
    paperHighly accurate protein structure prediction with AlphaFoldAlphaFold

    Predicted protein shapes almost as well as lab experiments. One of the most cited papers of the decade.

  • Jun 2017
    paperAttention Is All You Need (Google Brain)

    Introduced the transformer, the design behind modern language models.

  • Jan 2016
    paperMastering the game of Go with deep neural networks and tree search

    Beat top Go professionals by combining neural networks with search.

Open technical problems

Open · Active · Partial
  • From prediction to discoveryActive research

    Protein shapes are largely solved. Predicting how proteins move and bind, well enough to skip experiments, is not.

  • Shipping on scheduleOpen

    Gemini 3.5 Pro missed several internal targets in 2026. Staying ahead each release is hard.

  • World modelsActive research

    Genie 3 creates worlds you can explore, but keeping them consistent for long stretches is unsolved.

Peers

Frontier models · 5

Company & funding

Parent
Alphabet
HQ
GB/US
Founded
2010

Partners & customers

1 acquirer · 2 partners · 1 customer
  • Acquirer
    Alphabet

    Google acquired DeepMind in 2014.

  • Partner

    Alphabet drug-discovery sibling; co-developed AlphaFold 3.

  • Customer

    Google Cloud TPU customer, up to one million chips.

  • Partner
    Broadcom

    Long-time co-design partner on TPU silicon.

People

Key people
  • Demis HassabisCEO and co-founder
  • Shane LeggCo-founder and Chief AGI Scientist
  • John JumperAlphaFold lead; 2024 Nobel laureate
Primer