News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersWhat Does Hewlett Packard Enterprise (HPE) Gaining A Top 14 AI Supercomputer Mean?ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersWhat Does Hewlett Packard Enterprise (HPE) Gaining A Top 14 AI Supercomputer Mean?ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inference
EveryNeocloud

Google TPU

Google's Tensor Processing Units. 2 providers on EveryNeocloud so far.

The Tensor Processing Unit (TPU) is Google's own AI chip. Google began using TPUs inside its data centres in 2015, announced them in 2016 and has rented them to others through Google Cloud since 2018. Each generation is sold as pods, thousands of chips linked by Google's own interconnect. Google trains its Gemini models on TPUs, and Anthropic is one of the largest outside users.

Generations

Early TPUs ran only inference; from the second generation they trained models too. TPU v4 (2021) linked 4,096 chips with optical switches that can rewire the network. v5e and v5p came in 2023, Trillium (v6) in 2024 and Ironwood (v7), aimed at inference, was announced in April 2025. TPUs work best with Google's JAX and TensorFlow software, though PyTorch support has grown.

Outside Google

For most of their life TPUs were available only inside Google Cloud. That is changing: in October 2025 Anthropic agreed to use up to one million TPUs, and reports in 2025 said Google was offering TPUs to other clouds and to large customers for their own data centres. Supporters see TPUs as the strongest alternative to Nvidia; critics note that using them ties a customer to Google.

Famous google TPU providers

Still to be written: Google Cloud Platform.

Further reading

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.