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)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 inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaDealsCorvex 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)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 inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to Nvidia
EveryNeocloud

Chipmaker clouds

Cerebras, Groq and SambaNova: chip companies that sell their chips as a service. 3 providers on EveryNeocloud so far.

A few AI chip companies sell their chips mostly as a service rather than as boxes. Cerebras, Groq and SambaNova each found that customers would rather pay for answers than learn to run unfamiliar hardware, so each built data centres and an API cloud on its own chips. They compete with Nvidia GPUs mainly on inference speed, measured in tokens per second for each user.

Three designs

Cerebras builds one chip from a whole silicon wafer, the WSE-3, with 900,000 cores and memory spread across it. Groq's language processing unit (LPU) keeps model weights in fast on-chip memory and runs in a fixed, planned order, which makes answers quick and predictable. SambaNova's reconfigurable dataflow unit (RDU) maps a model onto the chip as a flow of data. All three avoid the separate high-bandwidth memory that GPUs need.

Their fortunes

Cerebras listed on the Nasdaq in May 2026 after earlier plans were withdrawn, with G42 long its largest customer. Groq opened data centres in the US, Canada, Saudi Arabia and Finland; in December 2025 Nvidia licensed its technology and hired its founder, while GroqCloud kept running. Cloud revenue gives these firms steady income and shows their chips at work, but it also means building data centres, a costly business of its own.

Famous chipmaker cloud providers

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.