DigitalOcean is now offering access to Nvidia H100 GPUs via its Paperspace cloud computing platform.

The cloud provider dedicated to small-to-medium businesses and startups, has added to its GPU-accelerated cloud computing platform with Nvidia chips.

Nvidia H100 tensor core GPU
– Nvidia

This infrastructure upgrade will increase DigitalOcean's ability to offer cloud-based artificial intelligence, machine learning applications, and high-performance computing (HPC).

The GPUs can be accessed on-demand or reserved for a specific period. The GPUs are also available as individual machines or in clusters.

DigitalOcean has also upgraded its network infrastructure to offer interconnect speeds of 3.2TBps.

“While many vendors are optimizing their offerings to serve large enterprises, DigitalOcean is proud to offer SMBs and startups with reliable and flexible access to Nvidia H100 GPUs,” said Kanishka Roychoudhury, GM of AI/ML at DigitalOcean.

“As we look ahead, DigitalOcean remains committed to providing solutions that offer superior performance and keep costs in check, lowering the barriers of entry for smaller businesses looking to leverage AI to improve profitability and efficiency.”

“Businesses of all sizes are increasingly looking to AI to boost efficiency and better serve customers,” added Dave Salvator, director of Accelerated Computing Products at Nvidia. “The availability of Nvidia H100 GPUs through DigitalOcean’s Paperspace platform will help empower the next generation of startups, providing greater access to the tools they need to develop cutting-edge AI products and solutions.”

DigitalOcean acquired Paperspace in July 2023 for $111m. According to DigitalOcean's website, the company has 15 globally distributed data centers, including in Amsterdam in the Netherlands, two in New York, one in San Francisco, and other facilities in Singapore; London, United Kingdom; Frankfurt, Germany; Toronto, Canada; Sydney, Australia; and Bangalore, India. The company's Regional Availability Matrix only covers 11 of these sites.

Earlier this week it was reported that cloud providers may be underutilizing their GPU deployments. The estimated revenue by TechInsights suggested that in 2023 878,000 accelerators were responsible for turning out seven million GPU-hours of work, for an estimated $5.8 billion in revenue spending, a number that would be much higher were cloud providers using all their resources.

Other cloud providers offering Nvidia H100 GPUs include Yotta Data Services, Google Cloud Platform, AWS, Gcore, Oracle Cloud Infrastructure, and Voltage Park.