72,000 H100 GPU equivalents a month, as OpenAI says that scaling laws show no sign of slowing down

Currently at just over $15k - but shipping is not included

Company says it has 35,000 Nvidia H100 GPU equivalents in operation

"We would probably build out bigger clusters than we currently can if we could get the energy to do it"

This model 2 hasn't been canceled

Troubled generative AI business owes millions to cloud companies

Enormous supercomputer could launch in 2028

Alongside the senior director of generative AI and another researcher

And takes most of Inflection AI startup with him, putting GPU cluster at risk

Researching AI and quantum at a Digital Realty data center

Available through hyperscale clouds

Using classical computing to help develop the next big thing

World's most powerful AI supercomputer will be used by Nvidia for its own R&D

Spatial computing meets accelerated computing

DGX B200, DGX GB200, and the DGX GB200 SuperPOD

Quantum-X800 InfiniBand and Spectrum-X800 Ethernet

720 petaflops of training performance, 1.4 exaflops of inference

Here's what's coming to a data center near you

Using AI to make chips better at AI

On-premise capacity can't keep up with NIST needs

Inference chip developer hopes to ship 1.5m LPUs

Newly-launched state-backed AI fund MGX in talks with Sam Altman

And is building its third supercomputer, alongside plans for Qualcomm inference chip deployments

As it raises investments and debt to build a GPU cloud