AWS Unveils EC2 Capacity Blocks for ML: A Game-Changer in GPU Access

Flexible GPU Access Model Aims to Boost Diversity and Efficiency

Amazon Web Services (AWS), a renowned cloud computing service favored by developers seeking high-performance hardware for AI workloads, has introduced a groundbreaking solution for more flexible, short-term requirements.

This innovation, known as Amazon Elastic Compute Cloud (EC2) Capacity Blocks for ML, is being touted as an industry-first by Amazon. It empowers customers to access GPUs through a consumption-based model, providing an affordable alternative.

The Seattle-based cloud giant envisions that these more cost-effective options will open doors for smaller organizations, fostering a more diverse landscape within the AI ecosystem.

In an official statement, Amazon explained, “With EC2 Capacity Blocks, customers can reserve hundreds of Nvidia GPUs colocated in Amazon EC2 UltraClusters designed for high-performance ML workloads.”

 

 

Customers can now tap into the latest Nvidia H100 Tensor Core GPUs, perfectly suited for training foundation models and large language models. They have the flexibility to specify the cluster size and duration, ensuring that they only pay for the resources they need.

Amazon highlighted the growing demand for GPUs in tandem with the proliferation of generative AI. Many businesses grapple with the dilemma of either overpaying for excess service or having GPUs idling when not in use, which can lead to inefficiencies.

AWS users now have the option to reserve EC2 UltraClusters of P5 instances for periods ranging from 1 to 14 days and up to eight weeks in advance. They can choose from various cluster size options, ranging from 1 to 64 instances, accommodating a maximum of 512 GPUs.

David Brown, AWS Compute and Networking VP, emphasized the significance of this development, stating, “With Amazon EC2 Capacity Blocks, we are introducing a new way for enterprises and startups to predictably acquire Nvidia GPU capacity to build, train, and deploy their generative AI applications—without committing to long-term capital investments. It’s the latest example of AWS’s commitment to expanding access to generative AI capabilities.”

For pricing details and to access this service, prospective users can visit the AWS website and sign up for this short-term, cost-effective GPU solution. This innovation promises to revolutionize GPU access for AI applications, offering a more efficient and accessible path for businesses of all sizes.