AWS Raises EC2 Prices as GPU Demand Surges

Key points

• Amazon Web Services (AWS) has increased the pricing of its Elastic Compute Cloud (Amazon EC2) Capacity Blocks for ML offerings by approximately 15%, affecting enterprises planning large-scale machine learning workloads.
• The price hike is likely due to high demand for Nvidia H100 and H200 GPUs, with AWS applying a scarcity premium to guaranteed inventory to recover higher infrastructure and capital costs.
• The move may lead to changes in how enterprises plan and budget for machine learning workloads, with other cloud providers like Google Cloud and Microsoft Azure offering similar guaranteed capacity options, but with different pricing and reservation models.

According to sources, Amazon Web Services (AWS) has updated the pricing structure for some of its Elastic Compute Cloud (Amazon EC2) Capacity Blocks for machine learning (ML) offerings. The price increase, which is approximately 15%, is likely to affect enterprises planning large-scale machine learning workloads. This move is significant, as it highlights the growing demand for Nvidia H100 and H200 GPUs, which are used to power Amazon EC2 instances.

The Amazon Capacity Blocks allow customers to reserve access to accelerated compute resources for a future start date, up to six months in advance. However, the EC2 Capacity Blocks can only be reserved up to eight weeks in advance. This means that customers need to plan ahead and reserve their resources early to ensure they have access to the compute power they need. AWS has not responded to requests for comment on the price increase, but experts believe it is due to the high demand for Nvidia H100 and H200 GPUs.

The price increase affects the P5 Capacity Blocks, which include EC2 P6 instances accelerated by the latest Nvidia Blackwell GPUs, P5 instances complemented by Nvidia H100 and H200 Tensor Core GPUs, and P4 instances powered by Nvidia A100 Tensor Core GPUs. For example, the price of the p5e.48xlarge instance, featuring eight Nvidia H200 accelerators, has increased from $34.608 to $39.799 per hour in the US East (Ohio) region. Similarly, the price of the p5en.48xlarge instance has jumped from $36.184 to $41.612 per hour in the same region.

Experts say that the price increase is likely due to the high demand for Nvidia H100 and H200 GPUs, which has led to a shortage of these chips. Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting, explained that AWS is applying a scarcity premium to guaranteed inventory to recover higher infrastructure and capital costs. This means that customers who need guaranteed access to GPU clusters will have to pay a premium for this service.

The move by AWS is part of a larger trend in the cloud computing industry, where guaranteed access to GPU capacity is becoming a key differentiator. Other cloud providers, such as Google Cloud and Microsoft Azure, are also offering similar guaranteed capacity options, but with different pricing and reservation models. For example, Google Cloud has introduced a calendar-based scheduling tool that lets customers reserve GPU capacity in fixed blocks ahead of time. Microsoft Azure, on the other hand, leans more on regional capacity reservations, which allow customers to hold specific VM types in specific zones.

While the price increase may lead to some changes in how enterprises plan and budget for machine learning workloads, experts do not expect an immediate fallout. The EC2 Capacity Blocks typically account for a small portion of total GPU spend, and strong data gravity, entrenched MLOps stacks, compliance controls, and skills continue to anchor workloads on AWS. However, the impact will be felt more in new workloads, as enterprises may need to reassess their cloud strategy and consider alternative options, such as Google Cloud or Microsoft Azure. As the demand for GPU capacity continues to grow, it will be interesting to see how AWS and other cloud providers respond to this trend and adapt their pricing and reservation models accordingly.

Read the rest: Source Link

Don’t forget to check our list of Cheap Windows VPS Hosting providers, How to get Windows Server 2022, Try Windows 11 Pro for Workstations & browse Windows Azure content.

Remember to like our facebook and follow us on twitter @WindowsMode.