Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage.
AI isn’t just for big enterprises anymore; small and mid-sized businesses (SMBs) are increasingly using it too. DigitalOcean, known for its cloud services focused on developers catering to SMBs, is expanding its offerings with the Nvidia H100 GPU AI accelerator. This powerful GPU will be available on the Paperspace platform, which DigitalOcean acquired in July 2023 to enhance its AI capabilities. Paperspace already offers GPU access and has a data science platform called Gradient.
The new service provides access to the high-performance H100 GPUs on Paperspace, making them more accessible to developers through virtualization. These GPUs can drastically reduce training times and improve inference response times for AI models. However, due to the current demand for generative AI, getting access to Nvidia H100 GPUs has been challenging for smaller organizations.
“For these companies, when they need very powerful GPUs, it’s tough to get access to them, which restricts their growth. That’s where we come in,” said Kanishka Roychoudhury, GM of AI/ML at DigitalOcean.
The Nvidia H100 GPU is extremely powerful, but not every small business or developer needs the full capability. The DigitalOcean Paperspace platform uses virtualization to provide the appropriate level of access and performance. Each physical H100 GPU can be split into up to eight virtual instances, allowing organizations to use a fraction of the GPU or the entire unit depending on their needs.
SMBs are already leveraging the DigitalOcean Paperspace platform to create various AI applications. Roychoudhury expects the new H100 support to further extend and expand these applications. He mentioned that customers from small businesses and startups are using the platform for diverse AI projects, including a video augmentation service for translating video between languages.
The H100 GPUs will primarily be used for AI model training due to their high performance, although some customers are already using the less powerful Nvidia A100 GPUs for inference tasks. The H100 is expected to support these inference workloads as demand grows.
The integration of Paperspace into DigitalOcean is still ongoing, but the platforms already work together. GPU resources, including the Nvidia H100, can be deployed via Paperspace with storage resources running in DigitalOcean compute. Future plans include making the H100 GPUs available through the Paperspace Gradient data science platform, a notebook-based environment for building, training, testing, and deploying AI models. Gradient integrates with Hugging Face for downloading and fine-tuning pre-trained models.
DigitalOcean aims to reduce barriers to accessing high-end AI tools and hardware like the Nvidia H100, making them easily available and affordable for smaller companies. “We want to make this very easily available and affordable to smaller companies,” Roychoudhury said.
Stay informed! Subscribe to get the latest news delivered to your inbox daily. By subscribing, you agree to VentureBeat’s Terms of Service.