Overview
- DeepMind has moved most original AlphaFold contributors into projects tied to Google’s Gemini LLM and agent research while a subset have left for Anthropic and Isomorphic Labs, reflecting a deliberate pivot from single‑purpose science work to platform and agent development.
- Big cloud providers are expanding capacity fast but face a core choice about whether to reserve new GPUs for in‑house model training or sell excess compute to customers, a trade‑off companies say will shape short‑term revenue and long‑term competitive position.
- Amazon raised planned 2026 capital expenditures to about $220 billion and warned that even that increase will not fully meet demand for AI compute, underscoring persistent supply pressure on servers and GPUs.
- WIPO data shows a surge in generative‑AI patent filings through 2025 with SoftBank leading global filings and multiple Chinese firms among top filers, highlighting divergent corporate IP strategies between patenting and protecting work as trade secrets.
- China’s NDRC announced acceleration of a national Artificial Intelligence Law and large compute investments that aim to scale domestic data centers and link regulation with industrial policy, a move that could reshape where training and sensitive workloads run.