Amazon.Com, Inc (NASDAQ:AMZN) revealed a new feature for its cloud computing service, Amazon Elastic Compute Cloud (EC2), called EC2 Capacity Blocks for ML.
This new offering addresses the growing demand for GPU capacity in machine learning (ML) applications. With the rise of ML, there has been a surge in the need for GPUs, which has led to a shortage in supply.
EC2 Capacity Blocks for ML provide a solution by allowing users to reserve GPU instances for specific durations, proving particularly useful for training ML models or running AI experiments.
The feature ensures that users can access GPUs when they need them without the long-term commitment of traditional cloud rentals.
The service offers access to Nvidia Corp (NASDAQ: NVDA) H100 Tensor Core GPUs in EC2 UltraClusters. These clusters are designed for high-performance ML tasks and use Elastic Fabric Adapter (EFA) networking to ensure optimal network performance.
Reserving GPU instances with EC2 Capacity Blocks is likened to booking a hotel room. Users can specify the date, duration, and the number of GPU instances they need. Once the reservation starts, users can launch their P5 instances, and any running instances will automatically terminate at the end of the reserved period.
This feature benefits those who need guaranteed GPU capacity for specific tasks, such as training ML models or handling increased demand for ML applications. It provides predictability and cost assurance, ensuring users know their expenses upfront.
Currently, EC2 Capacity Blocks for ML are available in the AWS US East (Ohio) Region.
Nvidia and its key supplier, Taiwan Semiconductor Manufacturing Company Ltd (NYSE: TSM), are facing a tough time in keeping up with the AI frenzy, leading to higher prices for the GPUs.
The price of an AI accelerator using Nvidia’s GPUs doubled to 80 million won ($59,193) from about 40 million won earlier in 2023. Also, the lead time for Nvidia’s GPU-based accelerators has reached 52 weeks.
Price Action: NVDA shares traded higher by 2.63% at $434.40 premarket on the last check Thursday.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
