Analysis: Google DeepMind has lost frontier-model momentum amid leadership and RL talent departures, with Google Cloud emerging as a winner with more compute
THE SO WHAT
If frontier-model leadership is fragmenting while Google Cloud’s AI revenue grows >100% YoY, the center of gravity is shifting from pure research to monetized infra and services. For buyers, that means more leverage to negotiate integrated compute + model + tooling bundles — but also more lock-in pressure as cloud becomes the AI control plane.
READ THE SOURCE
MORE FROM THE WIRE
Applied AI'Great music is made by people' — Suno, the biggest AI music company, is finally trying to solve a problem its own success helped create
When the largest AI music platform starts talking about putting the brakes on its own output, it’s a signal that pure volume has outrun both user experience and rights frameworks. Any AI content platform at scale will need active curation, provenance, and economic rules — not just better models — to stay viable with creators and regulators.
Google’s AI shake-up puts Demis Hassabis where the company needs him most, insiders say
Moving Demis Hassabis into a role focused on big leaps is a bet that frontier research and product integration can be decoupled and still move fast. For operators building on this ecosystem, expect more visible model roadmaps and tighter integration with core Google products as the company tries to turn research into durable platform surface.
I'm a co-creator of Alexa and an AI expert. If AI singularity ever arrives, everyone will notice — here’s how.
When an Alexa co-creator publicly downplays near-term singularity, it lowers the temperature on existential narratives and recenters on concrete capability curves. Use that air cover to focus leadership conversations on 6–24 month deployment risks and opportunities instead of speculative AGI scenarios.
Applied AIWhy AI infrastructure planning must happen now
The argument for early, balanced, open AI infrastructure is really about avoiding three-year technical debt — proprietary traps, bandwidth ceilings, and GPU bottlenecks. This week, map where your data, models, and orchestration actually live and which vendors control each layer before you greenlight the next AI initiative.