AWS to raise GPU cloud prices 20% on July 1

Amazon Amazon.com, Inc. ↗ 0.40% Web Services will increase prices on its EC2 Capacity Block reservations for machine learning GPU instances by approximately 20% effective July 1, 2026, according to an update posted to the company's official pricing documentation page. The increase marks the second hike to the service in six months, following a 15% rise quietly implemented on January 4, and reflects mounting cost pressures from a global memory chip shortage that has upended semiconductor supply chains.aws.amazon+1

The New Rates

The updated hourly rates per accelerator cover AWS's most powerful Nvidia ↘ 1.39%-powered instance families. The P5e, which features eight H200 GPUs, will rise from $4.975 to $5.97 per accelerator in most regions, while the P5en will jump from $5.721 to $6.865 per accelerator in U.S. regions. Newer P6-generation instances equipped with Blackwell B200 and B300 chips will be priced at $12.355 and $14.04 per accelerator, respectively. AWS said on its pricing page that "EC2 Capacity Blocks for ML reservation prices are updated periodically based on supply and demand".finance.yahoo+2

EC2 Capacity Blocks allow companies to reserve guaranteed GPU capacity for defined time windows, ranging from one day to several weeks, up to eight weeks in advance. The service is used for large-scale AI training and inference workloads. AWS has emphasized that on-demand, reserved instance, and spot pricing for the same GPU instances remain unchanged, meaning the increases affect a specific slice of its cloud portfolio rather than its broader pricing structure.itpro+2

A Structural Shift in Cloud Economics

The January increase broke what industry observers called a two-decade pattern of declining cloud compute costs. The Register, which first reported that earlier hike, noted that AWS had rarely applied "a straight increase to a line item". An AWS spokesperson told ITPro at the time that the adjustment "reflects the supply/demand patterns we expect this quarter".datacenterdynamics+3

Behind both price hikes is a memory chip shortage that analysts say could persist through 2027 or 2028. Memory manufacturers including Samsung, SK Hynix, and Micron ↘ 5.49% have shifted production capacity toward High Bandwidth Memory used in AI accelerators, creating scarcity in conventional DRAM and NAND flash. DRAM costs surged more than 170% year over year by late 2025, and Synopsys CEO Sassine Ghazi told CNBC in January that the chip "crunch" would likely extend into 2027.cnbc+2

Ripple Effects Across the Industry

As the world's largest cloud provider, AWS's pricing decisions carry weight across the technology sector. The company reported that its cloud business grew at its fastest pace in nearly four years during the first quarter of 2026, with Amazon posting total revenue of $181.5 billion. Capital expenditure guidance for 2026 was raised to approximately $200 billion as the company expands AI infrastructure.crn+1

Industry analysts had already been warning enterprises to budget for 10 to 25% cloud cost increases through 2027. One analysis noted that for teams running continuous GPU workloads, even the January increase translated to more than $3,700 per month in additional costs per instance.devzero+2