<h1>AWS Introduces Price Cuts and New GPU Instances for Enhanced AI Capabilities</h1>
In a strategic move to tackle the growing demand for Artificial Intelligence (AI) workloads amid industry-wide GPU shortages, Amazon Web Services (AWS) has announced significant price reductions for its NVIDIA GPU-accelerated EC2 instances. These cuts, which apply to both On-Demand and Savings Plan pricing across all available regions, offer up to a 45% cost decrease, particularly benefiting users of the P5 series with reductions as high as 44-45%.
Moreover, AWS is enhancing its service offerings by expanding at-scale On-Demand capacity availability in additional regions. This expansion is designed to ensure that more customers can leverage the robust performance of AWS instances powered by NVIDIA GPUs without delays or limitations in availability.
Complementing financial adjustments is the introduction of the P6-B200 instances for Amazon SageMaker AI Training Jobs, now generally available. These cutting-edge instances, fueled by NVIDIA B200 GPUs, promise a substantial performance boost, delivering up to twice the output compared to the prior P5en instances. Each P6-B200 instance comprises 8 Blackwell GPUs with a massive 1440 GB GPU memory and a 60% increase in memory bandwidth, facilitating complex AI workloads. Additionally, the network capabilities have been enhanced to handle up to 3.2 terabits per second, ensuring seamless data flow and high-efficiency training processes.
These P6-B200 instances are currently accessible in the US West (Oregon) region under SageMaker HyperPod Flexible Training Plans, showcasing the potential to scale to tens of thousands of GPUs utilizing the AWS Nitro System. This scalability and the technological advancements embedded in the new instances are set to revolutionize large-scale AI deployments, catering to the escalating computational needs of modern industries.
By blending cost-effective solutions with powerhouse processing capabilities, AWS continues to evolve its cloud computing services, supporting the burgeoning AI ecosystem with infrastructure that caters to both economic and technological demands.

Collections

AWS announced significant price cuts up to 45% for NVIDIA GPU-accelerated EC2 instances and introduced new P6-B200 instances powered by NVIDIA B200 GPUs. The new instances offer double the performance of previous P5en instances with 8 Blackwell GPUs, 1440 GB GPU memory, and enhanced network capabilities up to 3.2 terabits per second. These improvements aim to address GPU shortages and support growing AI workload demands with better cost-effectiveness and scalability.