FT Tech Tonic AI Labs: Google DeepMind plans its comeback
21 snips
May 13, 2026 
Stephen Morris, San Francisco bureau chief covering Silicon Valley, and Madhumita Murgia, FT AI editor and tech analyst, discuss DeepMind and Google’s AI strategy. They explore DeepMind’s early doubts about transformers, Google’s scramble after ChatGPT, talent losses and startup competition. They debate research publication pauses, commercialization pressure, ethics around military contracts, and whether DeepMind can steer a Google comeback.
AI Snips
Chapters
Transcript
Episode notes
DeepMind Misread The LLM Moment
- DeepMind initially resisted transformer-driven LLMs but later found large language models much better at mimicking intelligence than expected.
- Madhumita Murgia and Stephen Morris highlight Demis Hassabis calling ChatGPT's arrival “war” after OpenAI's unexpected lead, showing strategic shock and pivoting at DeepMind.
Google's Stack Is A Strategic Moat
- Google leverages an ecosystem advantage: custom TPUs, vast cloud infrastructure and deep pockets to integrate and scale AI models.
- Stephen Morris argues these assets give Google both cost and technology edges for deploying Gemini across consumers and enterprises.
AlphaGo And AlphaFold Proved DeepMind's Power
- DeepMind produced AlphaGo and AlphaFold, flagship projects that demonstrated superiority in Go and protein folding.
- Madhumita Murgia says AlphaGo beat champion Lee Sedol and AlphaFold transformed protein prediction, earning Nobel-level acclaim.