Tag: TPU
Four chips ride in orbit and Google can switch on only one
Google's Suncatcher prototype is talking to the ground and writing verse, but its TPUs draw so much power…
Meta commits large model training to Google chips through a new backend
TorchTPU swaps a brittle bridge for a dispatcher hook, letting PyTorch models run on TPUs with one device…
Google hands PyTorch a native path onto its TPUs
TorchTPU replaces PyTorch/XLA and puts Google's accelerators beside GPUs and CPUs as a first-class device in the framework.
Google puts a TPU in orbit to test space data centers
Google launched its first compute satellite, sending a TPU to space to prove orbital data centers can work.
Alphabet stock rises on report of new Frozen v2 AI chip for Gemini models
Alphabet shares rise on reports of a new Frozen v2 server chip designed to run Gemini models 6-10x…
Google Splits Its TPU Line in Two. Good Luck Keeping Up With Nvidia.
Google's new TPU 8t and TPU 8i split the chip into training and inference roles, betting the agentic…
Google Commits Up to $40 Billion to Anthropic in Cloud Compute Deal
Anthropic's rapid growth, driven by products like Claude Code and controversies at OpenAI, has pushed demand beyond its…
Google Splits Its Next Gen TPU: One Chip for Training, One for Agents
Google bifurcated its eighth generation TPU into a training specific chip and an inference focused chip, claiming twice…