News
DealsNscale’s ByteDance GPU deal draws scrutiny over US chip curbsSovereignNexusIgnite Plans Sovereign AI Cloud in Jordan, Starting With Its UniversitiesSovereignNexGen Cloud Appoints Dan Thomas as Chief Infrastructure Officer to Lead Next Phase of European AI Infrastructure Build-OutSovereignNexusIgnite Plans Sovereign AI Cloud in Jordan, Starting With Its UniversitiesSovereignSovereign AI: Africa Is Buying The Chips, Not The RulesMinersBitcoin miner that sold its last BTC warns it may run short of cash for $1.2 billion AI dealInferenceIndia Startup Sells DeepSeek Access While Pursuing Sovereign AIClustersLumen’s Rob Ward: 'The AI factory has no walls'NeocloudsAWS Lambda now supports GPU instances for Event Functions on Lambda Managed InstancesMinersBitcoin Miners Are Selling Their Stacks to Fund the AI BoomDealsNscale’s ByteDance GPU deal draws scrutiny over US chip curbsSovereignNexusIgnite Plans Sovereign AI Cloud in Jordan, Starting With Its UniversitiesSovereignNexGen Cloud Appoints Dan Thomas as Chief Infrastructure Officer to Lead Next Phase of European AI Infrastructure Build-OutSovereignNexusIgnite Plans Sovereign AI Cloud in Jordan, Starting With Its UniversitiesSovereignSovereign AI: Africa Is Buying The Chips, Not The RulesMinersBitcoin miner that sold its last BTC warns it may run short of cash for $1.2 billion AI dealInferenceIndia Startup Sells DeepSeek Access While Pursuing Sovereign AIClustersLumen’s Rob Ward: 'The AI factory has no walls'NeocloudsAWS Lambda now supports GPU instances for Event Functions on Lambda Managed InstancesMinersBitcoin Miners Are Selling Their Stacks to Fund the AI Boom
EveryNeocloud

Groq LPU

Groq's language processing units. 1 provider on EveryNeocloud so far.

The language processing unit (LPU) is the chip made by Groq, a company founded in 2016 by Jonathan Ross, who earlier helped start Google's TPU project. Groq sells it almost entirely as a service, GroqCloud, which serves open models through an API. Its pitch is speed: answers that stream out at hundreds of tokens per second for each user.

How it works

An LPU holds model data in SRAM, very fast memory on the chip itself, rather than in separate high-bandwidth memory. The compiler plans every step in advance, so the chip runs in a fixed, predictable order with no waiting. Because each chip holds little memory, a large model is spread across hundreds of LPUs, which makes Groq systems big but fast.

Where it runs

Groq runs data centres in the US, with Bell Canada in Canada, with Humain in Saudi Arabia and, from July 2025, in Helsinki with Equinix. In December 2025 Nvidia signed a non-exclusive licence for Groq's technology and hired its founder and other leaders, while Groq said GroqCloud would keep running as an independent company.

Famous groq LPU providers

Still to be written: Bell Canada.

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.