Google executed a coordinated three-layer strategy at its Cloud Next conference, announcing new TPU chips, transforming Chrome into an AI coworker, and securing a multi-billion dollar compute deal with Thinking Machine Labs, the company founded by former OpenAI co-founder Miriam Ratti. The moves position Google as the only major player credibly operating across the entire AI stack, from silicon to applications.
The company introduced two new TPU models: the TPU-8T for training and the TPU-8I for inference. Google claims the chips deliver three times faster training performance and 80 percent better performance per dollar compared to NVIDIA alternatives, with the ability to scale more than one million TPUs in a single cluster. The inference-specific silicon addresses what has become the dominant cost in running AI at production scale. Google will also resell NVIDIA's Vera Rubin chips later this year, allowing customers to choose between competing hardware options.
Chrome's new auto-browse feature, powered by Gemini, reads context across open browser tabs and automates workplace tasks including CRM data entry, vendor quote comparisons, and competitor research. The feature requires human approval for each action and allows users to save frequent workflows as skills that can be triggered with a forward slash command. Enterprise prompts will not be used to train Google's models, addressing a key concern for corporate customers deploying AI tools with confidential data.
Key facts
- 01Google. A key thread of reporting in this story.
- 02TPU. A key thread of reporting in this story.
- 03Thinking Machine Labs. A key thread of reporting in this story.
The multi-billion dollar deal with Thinking Machine Labs grants the company access to NVIDIA GB 300 systems on Google Cloud, plus training and deployment services for their first product, Tinker, a tool for building custom frontier models. Thinking Machine Labs is raising capital at a twelve billion dollar valuation and runs compute-intensive reinforcement learning workloads. Anthropic already operates on TPUs, and the addition of Thinking Machine Labs strengthens Google's position as the preferred compute host for frontier AI labs.
“Google is basically the only company really doing the full stack.”— Jaeden Schafer
Google's full-stack approach contrasts sharply with competitors. Microsoft relies heavily on OpenAI at the application layer but lacks proprietary silicon. Amazon has bet heavily on Anthropic. NVIDIA dominates chips but does not touch applications. Google operates across all three layers: silicon through TPUs and resold NVIDIA chips, compute hosting for frontier labs, and the agent layer through Chrome and Workspace integration. The strategy faces potential regulatory scrutiny as the DOJ antitrust case over search remedies remains active, and turning Chrome into an agent layer that pulls from Workspace data may draw additional attention.
The announcements also included news from other AI companies. Stanford spinout 10xScience closed a 4.8 million dollar seed round led by Initialized Capital to build tools that triage drug candidates generated by AI models. Neocognition emerged from stealth with forty million dollars to develop AI agents that self-specialize rather than function as unreliable generalists. Bloomberg reported that an unauthorized group accessed Anthropic's exclusive cybersecurity tool Mythos through compromised contractor credentials and a predictable URL pattern. OpenAI partnered with Infosys to distribute ChatGPT and Codex across more than sixty countries, targeting enterprise accounts where Microsoft does not have distribution.
“I think this is one of the clearest signals from me so far that Google is structurally perhaps ahead of OpenAI and Amazon in the AI stack.”
“Google is basically the only company really doing the full stack.”
“The customer really doesn't care whether Gemini tops the Elo leaderboards. If they can run inference cheaper on Google stack, and if they can serve it to their employees through Chrome and wire it through the workspace data without, you know, having this big, huge system integrator structural position, I think matters way more than just pure benchmark lead in the next 12 months.”
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
