Oracle has become the first cloud provider to offer a Zettascale supercomputing cluster, utilizing Nvidia’s latest Blackwell hardware, enabling customers to build, train, and run AI models at an unprecedented scale.
Announced at the company’s Oracle CloudWorld event in Las Vegas, it unveiled the supercomputing cluster, which is made up of some 131,072 Nvidia Blackwell GPUs.
The cluster offers 2.4 zettaFLOPS of peak performance — that’s the equivalent of performing 2,400,000,000,000,000,000,000 (2.4 quintillion) floating-point operations every second.
The maximum scale of the cluster contains three times as many GPUs as Frontier, currently the most powerful supercomputer in the world, surpassing its peak performance of 1.12 exaFLOPS by some distance
Subscribe today for free
“We have one of the broadest AI infrastructure offerings and are supporting customers that are running some of the most demanding AI workloads in the cloud,” said Mahesh Thiagarajan, executive vice president at Oracle Cloud Infrastructure (OCI).
“With Oracle’s distributed cloud, customers have the flexibility to deploy cloud and AI services wherever they choose while preserving the highest levels of data and AI sovereignty.”
The cluster achieved the jump from exascale power to zettascale by utilising Nvidia’s next-gen flagship chips to achieve its incredible speeds. Blackwell was unveiled at GTC 2024 back in March, with tech companies all over the world scrambling to get their hands on a line of chips that supports double the compute compared to the prior H100s.
Nvidia was expected to start shipping towards the end of 2024, but this was pushed back slightly after manufacturer TSMC uncovered a design flaw with the hardware’s processor die, preventing mass production.
Nvidia has since resolved the Blackwell issue, with a ramp-up of production expected to begin at the end of the year.
OCI did not confirm when its Blackwell-powered service will come online considering the chips are currently being produced. Capacity has reached out for clarification.
OCI, however, is already accepting orders for the Blackwell-powered supercomputing cluster, allowing customers to utilise it for their AI workloads.
Videoconferencing platform Zoom is among those already leveraging OCI’s AI infrastructure, using it to provide inference for the company’s AI personal assistant, Zoom AI Companion.
“By harnessing OCI’s AI inference capabilities, Zoom is able to deliver accurate results at low latency, empowering users to collaborate seamlessly, communicate effortlessly, and boost productivity, efficiency, and potential like never before,” said Bo Yan, head of AI at Zoom.
RELATED STORIES
Who is Matt Souders, Oracle's new VP of data centre infrastructure?
Oracle and AT&T partner to enhance cloud applications
Digital Realty, Oracle extend partnership on cloud native and AI enterprise workloads
