DigitalOcean Holdings Inc., the cloud infrastructure platform provider for small developer teams, said its latest artificial intelligence compute services powered by Nvidia Corp.’s H100 graphics processing units are now generally available.
The DigitalOcean GPU Droplets are H100 GPU-powered virtual machines available in single- and multi-node configurations, designed for AI use cases including experimentation, training and inference on machine learning models, generative AI applications and more.
In addition to the new VMs, DigitalOcean is also launching a new managed Kubernetes service that supports H100 GPUs, bringing the power of Nvidia’s most advanced chips to containerized application environments.
With the launch, the company says it’s giving small and medium-sized businesses and startups the opportunity to access the critical infrastructure required to develop next-generation applications powered by AI, which require access to extremely powerful computing hardware.
Many smaller companies have been left behind in the AI arms race due to their inability to access GPU resources. Nvidia’s GPUs are in high demand because every company wants to press ahead with its AI initiatives, and there simply aren’t enough of the chips to go around. That has resulted in extremely high costs to access GPUs on cloud infrastructure platforms such as Amazon Web Services, Microsoft Azure and Google Cloud, beyond the ability of what many smaller firms are able to pay.
The GPUs are viewed as an essential resource by AI practitioners, because they’re unique in their ability to reduce the time it takes to train AI models and speed up model inference.
DigitalOcean says its GPU Droplets can help to lower the barrier to AI development, providing fast, easy and affordable access to high-performance GPUs without any upfront investment in expensive hardware, nor any long-term commitments to rent them.
Developer-friendly
Unlike traditional cloud platforms, which require extensive setup and maintenance, DigitalOcean offers a much more user-friendly experience that’s focused on speed of deployment.
The company is a competitor to Amazon Web Services Inc. and Microsoft Corp. in the public cloud infrastructure market. However, rather than take on those giants directly, it has carved a niche for itself serving small businesses with its “developer cloud” that makes it easy for small teams of developers to create modern applications.
With the DigitalOcean App Platform, developers can deploy application code in production with a few clicks, in line with the company’s stated aim of keeping cloud computing simple. Its pitch is that it takes care of the cloud infrastructure and deployment side of things, so developers can maintain a focus on their code.
As an example of its ease of use, the company said each GPU Droplet includes two high-performance local disks, with one for storing the operating system, applications and AI framework, and the other reserved for staging data. They come pre-integrated, so there’s no need to manage storage and networking resources separately.
The company said its GPU Droplets can be used for a range of workloads, including deep learning, high-performance computing, data analytics, graphics and video rendering, and generative AI applications.
DigitalOcean Chief Executive Paddy Srinivasan said smaller developer teams have waited too long for dedicated AI infrastructure that’s tailored to their requirements.
“We’re making it easier and more affordable for developers, startups and other innovators to build and deploy generative AI applications and move them into production,” he said. “To do that, they need access to advanced AI infrastructure without the cost and complexity. Our GPUs as a service open this opportunity to a much broader market.”
In the near future, DigitalOcean is planning to launch a new GenAI platform for developers that aims to simplify the process of creating generative AI models. It will come with prebuilt components such as hosted large language models, data ingestion pipelines and knowledge bases, and everything else developers need to get started quickly.