AWS raises GPU instance prices 20% as memory shortage drives AI infrastructure costs higher

AWS GPU prices jump 20% as memory crunch bites

Amazon has raised the price of reserving AI chips on AWS by about 20%, its second hike this year. The rising AWS GPU prices show the memory shortage has reached the cloud, where customers have nowhere else to go. Renting an AI chip is starting to feel like booking a hotel in a sold-out city. You pay to hold the room, and the rate keeps climbing. On AWS, it just climbed again. Amazon Web Services has raised prices for EC2 Capacity Blocks for ML by roughly 20%, starting in July. Business Insider first reported the change, and AWS confirmed it. The service lets companies reserve Nvidia GPUs in advance, so a long training run keeps going instead of stalling halfway. This is the second increase in six months. AWS had already lifted the same prices by about 15% in January. Stacked together, the cost of locking in this compute has jumped sharply since the new year. AWS said the prices change "periodically based on supply and demand." The rise is narrow, not blanket. It hits one purchasing option: the reserved blocks favoured by serious AI teams training or fine-tuning large models. Other options keep fixed prices, AWS said, and the company says it will hold them there. The increase spared Trainium, Amazon's in-house AI chip, according to The Information. The scope still matters, because of how much sits on top. AWS is the world's largest cloud provider, and a sprawl of AI services runs on its servers. When the priciest tier of its compute goes up, the cost ripples out to the start-ups and enterprises renting it. AWS, for its part, framed the move as proof of how strong demand for GPUs has become. Why the cloud could not stay cheap The shift worth watching is where the squeeze has reached. For two years, the limit on AI was software and know-how. Now it is physical. The bottleneck is high-bandwidth memory, the chips stacked beside AI processors, and there is only so much of it to go around. The constraint on AI has moved from code to silicon, as Business Insider put it, and silicon takes years to build. The chain is short and unforgiving. Less memory means fewer GPUs. Fewer GPUs means fewer data centres. "As there is a limit to how much memory can be produced, then there is a limit to how many GPUs can be produced, which means that there's a limit to how many data centers can be built," Peter Berezin, chief economist at BCA Research, wrote on X. That scarcity hands the cloud giants a lever. When GPU capacity is tight, customers have few alternatives, so AWS, Microsoft, Google, and Oracle can pass higher costs straight through. The shortage raises their own bills, Berezin noted, yet it also keeps demand above supply, which gives them pricing power over who gets to compute at all. A price rise that is now everywhere Amazon is not alone, and that is the point. Apple raised prices across its Macs and iPads this week, blaming memory. Xbox did the same. Elon Musk called the jump in memory the biggest price increase he has seen in anything. Now the same memory prices are turning up in cloud bills. The other side of the squeeze is a windfall. The shortages lifting AWS GPU prices have pushed memory makers Micron and SK Hynix to record valuations. Investors are betting the high-bandwidth memory crunch keeps the market tight, and prices high, for years. For AWS customers, the message is plain. The cheapest AI compute is behind them, and the reserve button now costs more to press. The open question is how far this travels. If a 15% rise became another 20% in six months, the teams building the most ambitious models are left guessing what the next reservation will cost.

Amazon quietly raises price tag on EC2 Capacity Blocks for ML

Amazon (AMZN) is telling investors that the artificial intelligence boom comes with a very real price tag. And Wall Street needs to take note. Amazon Web Services quietly posted an update to the pricing for Amazon EC2 Capacity Blocks for ML, a reservation product that allows customers to lock in accelerator capacity for machine learning workloads. The move impacts some of the most sought-after AI hardware in the cloud market, including Nvidia Blackwell, H100, and H200 systems. The move does not represent a broad price increase for every Amazon product. But it is in one of the most crucial corners of the AI economy: dedicated compute capacity for companies that need serious GPU power and can't risk not having it there when model training or fine-tuning begins. AWS said the new rates take effect July 1, 2026. Some of the key accelerator reservations are moving about 20% higher, according to AWS pricing data and market reports. It's a small but telling glimpse of the AI arms race in action. More compute is still in demand. Cloud providers still possess pricing power. And the cost of turning AI ambition into actual products continues to climb. "Amazon EC2 Capacity Blocks for ML reservation prices are updated periodically based on supply and demand," AWS said. Amazon AWS raises key AI accelerator reservation rates EC2 Capacity Blocks are for customers who want to reserve GPU-based accelerated computing instances for a future date. Capacity Block instances are co-located on Amazon EC2 UltraClusters, which are designed for low-latency, high-throughput networking, AWS documentation says. That makes them particularly relevant for AI teams planning costly, short-term machine-learning workloads. The July update affects high-end accelerator families. In available non-GovCloud Regions, the P6-B300 capacity will move to $14.04 per accelerator hour and the P6-B200 capacity will move to $12.355 per accelerator hour, AWS said. Those two lines are important because they indicate Nvidia's upcoming Blackwell-based machines, the same kind of hardware that major AI builders and cloud platforms have been scrambling to get their hands on. The bump is also seen in Nvidia's older AI workhorses. According to AWS, P5 capacity, linked to H100 accelerators, will cost $5.191 per accelerator hour in U.S. regions where it's available and $4.72 in non-U.S. regions where it's accessible. P5e capacity, connected to H200 accelerators, will be $5.97 per accelerator hour in all accessible regions. The figures are technical, but the message is not. Amazon is showing that premium AI compute is still scarce enough to fetch higher prices, as investors argue whether the AI trade is too crowded. Nvidia hardware sits at center of Amazon price reset The AWS change is also a reminder that the AI boom doesn't just involve chatbots, software features, and model releases. It also involves physical capability. Capacity Blocks enable AWS instances powered by Nvidia Blackwell GPUs, Nvidia H200 Tensor Core GPUs, Nvidia H100 Tensor Core GPUs and Nvidia A100 Tensor Core GPUs and AWS Trainium instances. Key AWS AI pricing changes * P6-B300: July rate moves to $14.04 per accelerator hour in available non-GovCloud regions. * P6-B200: July rate moves to $12.355 per accelerator hour in available non-GovCloud regions. * P5: July rate moves to $5.191 per accelerator hour in available U.S. Regions and $4.72 in available non-U.S. regions. * P5e: July rate moves to $5.97 per accelerator hour in all available regions. * P5en: July rate moves to $6.865 per accelerator hour in available U.S. regions and $6.241 in available non-U.S. regions. * P4de: July rate moves to $2.214 per accelerator hour in available U.S. regions. Source: AWS That makes the price change more than just an Amazon thing. It also reaffirms why Nvidia (NVDA) is still such a huge part of the AI investment story. Cloud clients are still willing to pay a premium to reserve access to Nvidia-powered infrastructure, which means demand is still strong at the layer where AI investing turns real. Business Insider reported that the AWS change applies to one purchasing option, not all AI cloud services, and follows a previous price increase earlier in 2026. It also noted that the broader AI buildout continues to face pressure from constrained GPU and high-bandwidth memory supply. Michael M. Santiago / Getty Images Amazon customers face tougher AI budget test The cost hike could hamper AI planning for AWS clients. Capacity Blocks are paid differently than on-demand cloud consumption. AWS states the reservation fee is paid upfront when the reservation is booked, and consumers pay the going rate when they purchase, even if the Capacity Block starts after a price change later on. That squeezes enterprises trying to lock in computing ahead of the July reset. For Amazon investors, the message is more nuanced but possibly bullish. Higher prices on restricted AI capacity imply AWS still has clout in one of the fastest-growing markets in cloud computing. The AI boom might boost AWS revenue growth and cement Amazon's position as one of the top infrastructure winners if customers continue to take on greater prices. But there is a danger, too. As AI compute becomes more costly, it gets harder for customers to prove that their AI projects can provide a return. That could eventually split well-funded organizations from smaller teams and make some customers reconsider how aggressively they train, fine-tune, or deploy massive models. That is why the pricing update is important. This isn't just a line-item adjustment on an AWS web page. It's another indicator that the AI economy is shifting from hype into hard costs. Amazon isn't claiming demand for AI is waning. It's stating the opposite: Buyers still want the computing power strongly enough to drive the price higher. The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc. This story was originally published June 27, 2026 at 6:37 PM.

AWS raising GPU instance prices 20% on July 1 By Investing.com

Investing.com -- Amazon's (NASDAQ:AMZN) AWS will raise prices on its EC2 Capacity Block reservations for machine-learning GPU instances by approximately 20% effective July 1, 2026, citing supply and demand dynamics in a posting to its official documentation page. The new hourly rates per accelerator span AWS's most powerful Nvidia-powered instance families: the P6-B300 will be billed at $14.04, the P6-B200 at $12.355, the P5 (US regions) at $5.191, the P5 (non-US) at $4.72, the P5e at $5.97, the P5en (US) at $6.865, the P5en (non-US) at $6.241, and the P4de (US) at $2.214. All other EC2 prices remain unchanged, according to the AWS documentation. "Amazon EC2 Capacity Blocks for ML reservation prices are updated periodically based on supply and demand," the company said on its pricing page. The hike lands at a moment of sustained, surging enterprise appetite for GPU compute. AWS revenue climbed 28% year-over-year to $37.6 billion in the first quarter of 2026, the cloud unit's fastest growth rate in more than three years, a pace that gave AWS considerable pricing leverage with customers locked into AI training and inference workloads. Amazon has committed roughly $200 billion in capital expenditure in 2026 to AI infrastructure, and Reuters reported in March 2026 that Amazon is set to receive 1 million Nvidia GPU chips by end-2027 under a cloud supply agreement -- a deal that underscores just how supply-constrained the high-end GPU market remains. Capacity Blocks for ML are a reserved-capacity product that lets enterprises secure scarce GPU instances on a future date for time-bound workloads, typically large-scale model training. Because the product is reservation-based, customers have been willing to pay a premium over spot-market rates for the guarantee of availability; the new rates represent a significant step-up in that premium. For context, P6-B200 on-demand rates for an eight-GPU node were already running at roughly $14.24 per hour for the full node ahead of this change, per pricing analysis from Spheron Network published on June 20. For Nvidia, the pricing action is a dual-edged signal. The tight supply of P5 and P6 instances, which are built on Nvidia's Blackwell (B200, B300) and Hopper (H100) GPU architectures, confirms robust end-market demand for Nvidia silicon. Yet rising reservation costs could prompt some AWS customers to evaluate alternatives, including Nvidia-powered offerings on rival clouds or Google Cloud's TPU-based instances, which Alphabet has been actively marketing as a cost-competitive option. Whether Microsoft Azure or Google Cloud follow AWS with comparable GPU reservation price increases will be closely watched. Azure is AWS's nearest rival in enterprise cloud infrastructure, and a unilateral AWS hike could either spur competitive repricing or give Azure and Google Cloud an opening to attract cost-sensitive AI workloads. It also remains unclear whether existing Capacity Block reservations placed before July 1 will be honored at prior rates or billed at the new schedule from that date forward. With the increases taking effect in less than a week, enterprise buyers face an immediate decision: lock in any remaining capacity at current rates before July 1 or absorb the higher costs as a structural feature of the AI infrastructure landscape AWS has helped define.

Amazon Web Services to Raise Prices for Some EC2 AI Capacity Blocks

Amazon.com's Amazon Web Services said it would prices for certain Amazon EC2 Capacity Blocks for machine learning beginning July 1 as part of a periodic pricing update based on supply and demand. The company said hourly reservation rates would increase for several instance types powered by graphics-processing units, or GPUs, including P6-B300, P6-B200, P5, P5e, P5en and P4de Capacity Blocks. Other prices would remain unchanged. The updated rates apply across most AWS regions, with some regional differences. EC2 Capacity Blocks let customers reserve GPU-powered-computing capacity in advance for artificial-intelligence-model training and other machine-learning workloads, ensuring access to high-demand chips without long-term commitments. The service supports Nvidia Blackwell, H200, H100 and A100 GPUs, as well as AWS's Trainium chips, and allows reservations for up to six months. AWS's Capacity Blocks require customers to pay an upfront reservation fee, with operating-system charges billed separately while instances are running. Demand for AI-computing capacity has surged as companies race to build and deploy generative AI models. The Wall Street Journal previously reported that GPU-capacity shortages have pushed up rental prices for advanced Nvidia chips and forced some AI companies to ration computing resources amid strong demand.