DigitalOcean's Bare Metal GPUs are now generally available.
Launched on November 21, the offering gives customers a single-tenant infrastructure that is designed for artificial intelligence (AI) and machine learning (ML) workloads.
Bare Metal GPUs differ from DigitalOcean's GPU Droplets in that the infrastructure is not shared, and gives customers full access to the entire GPU. According to DigitalOcean, this is ideal for customers needing direct control over hardware and is "tailored" for projects like large-scale model training and real-time inference.
“Our new GPU Droplets provide customers with simple scalability and quick provisioning, but we recognize that this isn’t what all of our customers require, leading to the introduction of our Bare Metal GPUs,” said Bratin Saha, chief product & technology officer at DigitalOcean. “Bare Metal GPUs give our customers the performance and control that they require to tackle large datasets with ease.”
The Bare Metal GPUs are currently available in New York in the US and Amsterdam in the Netherlands.
DigitalOcean began offering Nvidia H100s via the DigitalOcean Kubernetes platform in Q3 of 2024, and has been offering them via its Paperspace cloud computing platform since January of this year.
During a recent earnings call, CFO Matt Steinfort said that it had Nvidia H200s coming, and "we're keeping an eye on Blackwell to see the timing of that."
The cloud provider saw a revenue of $198 million during Q3, 2024.