AI is no longer exclusive to large enterprises; small and mid-sized businesses (SMBs) are increasingly embracing this technology.
DigitalOcean is enhancing its cloud services for developers, specifically catering to SMBs, with the introduction of support for the industry-leading Nvidia H100 GPU AI accelerator. Following its acquisition of Paperspace in July 2023, DigitalOcean aims to penetrate the AI market. Paperspace has been providing GPU access and hosts a data science platform called Gradient.
The new offering on the Paperspace platform allows developers scalable access to the sought-after H100 GPUs through virtualization, making powerful AI capabilities more accessible. H100 GPUs dramatically reduce training times and enhance inference response times for AI models.
As generative AI gains traction, the demand for Nvidia H100 GPUs surges, presenting challenges for smaller organizations seeking access. "For these companies, when they need very powerful GPUs, it's tough to secure them, which hampers their business growth," said Kanishka Roychoudhury, GM of AI/ML at DigitalOcean.
DigitalOcean is adopting a virtualization strategy for H100 GPU access. While not every SMB requires a full H100, many could benefit from partial access to its AI capabilities. “Smaller startups come in all shapes and sizes,” explained Roychoudhury.
The DigitalOcean Paperspace platform enables tailored GPU access and performance. Each physical Nvidia H100 GPU can be virtualized to provide up to eight separate GPU instances, allowing organizations to provision as needed—whether a single instance or the full power of all eight.
SMBs are already leveraging the DigitalOcean Paperspace platform to create a variety of AI applications. Roychoudhury anticipates that the new Nvidia H100 support will broaden their application development.
Examples of use cases include a startup developing a video augmentation service for multilingual video translation. Roychoudhury expects that while the H100 GPUs will primarily be used for AI model training, some customers are already utilizing the less powerful Nvidia A100 GPUs for inference tasks, with plans to transition to the H100 as demand increases.
The integration of Paperspace into DigitalOcean is ongoing, with expectations for further enhancements in the coming months. Currently, GPU resources, including the Nvidia H100, can be deployed through Paperspace alongside DigitalOcean's computing storage resources.
Future plans include making H100 GPUs available on the Paperspace Gradient data science platform. Gradient offers a notebook-based environment for data scientists and engineers to build, train, test, and deploy AI models. Developers can start projects with basic computing resources and upgrade to more powerful options like GPUs. Gradient also integrates with Hugging Face to support the downloading and fine-tuning of pre-trained models.
DigitalOcean's overarching goal is to lower barriers for SMBs, making high-end AI tools and hardware like the Nvidia H100 easily accessible and affordable. "We aim to make this technology widely available to smaller companies," Roychoudhury stated.
Ranking
