TPU (Product): news timeline & CVEs

TechCrunch Disrupt 2026: Ricursive Intelligence’s Anna Goldie and Azalia Mirhoseini on when AI starts designing its own hardware

Ricursive Intelligence, valued at $4 billion, is building AI that designs and improves chips.

TechCrunch previews a Disrupt 2026 session with Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini on AI that designs chips. Before the startup, the pair co-led Google’s AlphaChip, which generated layouts used in multiple generations of Google TPUs, and both later worked at Anthropic and Google DeepMind. Launched in late 2025, Ricursive raised $335 million within four months at a $4 billion valuation, including a $300 million Series A, with Nvidia among the investors. The company aims to automate placement through verification and shorten chip-design cycles from two to three years toward weeks.

TechCrunch · AISep 25, 15:00 UTC · 1d agoAI industry

Google's first Suncatcher orbital data center test launches October 1

Google's first Suncatcher orbital data-center test launches October 1 to run Gemini on TPUs in space.

Google plans to launch its first Suncatcher orbital data-center test on October 1, 2026, after conducting testing only on the ground. The satellite will carry unmodified TPUs and run Gemini models, but radiators can cool the accelerators for only about 15 minutes before they must shut down. Google still plans 2027 launches and says it will be years before Suncatcher becomes a product.

Ars Technica · AIupdated Sep 24, 17:45 UTC · 2d agofirst Sep 24, 16:16 UTC · 2d agoAI industry 3 sources

MaxKernel: Agentic Kernel Generation for TPUs

Researchers open-source MaxKernel, a multi-agent LLM system that generates and optimizes TPU kernels matching expert hand-tuned baselines on JaxBench.

MaxKernel is a multi-agent system offering three paradigms for TPU kernel development: human-in-the-loop collaborative design, a fully autonomous metric/trace-driven optimization loop, and graph-based autonomous search for global exploration. All paradigms draw on a shared pool of specialized sub-agents for planning, implementation, self-debugging, testing, and hardware profiling. Evaluated on JaxBench's 50 diverse TPU kernel tasks and real-world workloads from open-source models, it consistently matches expert hand-tuned baselines. The system is open-sourced via the AI-Hypercomputer GitHub repository.

Hugging Face daily papersSep 2, 20:00 UTC · 24d agoAI tools & infra1

TPU

Timeline3 articles

Appears with

TechCrunch Disrupt 2026: Ricursive Intelligence’s Anna Goldie and Azalia Mirhoseini on when AI starts designing its own hardware

Google's first Suncatcher orbital data center test launches October 1

MaxKernel: Agentic Kernel Generation for TPUs