Anthropic told IPO investors that its Claude Mythos Preview model wrote or ported nearly all the low-level kernel code needed to run an Anthropic AI model on AMD's MI355X GPUs in about three days with minimal human help. Co-founder Tom Brown said the same task previously took six to 12 months using teams of dozens of engineers. Anthropic already runs its models across Nvidia, AMD, Google TPUs and AWS Trainium chips, and has a deal with AMD for up to 2 gigawatts of capacity.
tomshardware.com
· 2026-11-16
Google and Meta jointly released a PyTorch-native version of the SGLang inference engine, called TorchTPU, letting Google's tensor processing units serve AI workloads without requiring a switch to Google's JAX framework. Meta's PyTorch-TPU lead Zongwei Zhou said the engine is already running in production at Meta, built with help from RadixArk, the startup founded by SGLang's creators. A matching PyTorch-native version of the vLLM inference engine was also released as open source.
tomshardware.com
· 2026-10-22
Alphabet is sending its tensor processing units to space on a SpaceX Falcon 9 rocket as part of the Transporter-18 mission launching from Vandenberg Air Force Base. The flight carries Planet Labs satellites, including a solar-powered prototype fitted with Google's AI chips, marking the first in-orbit test of Google's Project Suncatcher initiative first announced in November 2025.
cnbc.com
· 2026-10-01