Over the past weekend, AWS made a quiet yet significant move by raising the prices of its EC2 machine learning capacity blocks by roughly 15%.

This move completely breaks away from the two-decade-long industry norm where cloud service prices have consistently trended downward, never upward. For tech firms and research institutions relying on these services for large-scale AI training, this price increase not only means higher costs but also signals a profound transformation in cloud computing pricing strategies.

The price hike targets AWS's EC2 capacity block services, specifically focusing on two instance types: p5e.48xlarge and p5en.48xlarge. Both instances are outfitted with eight NVIDIA H200 GPUs and are tailored for large-scale machine learning training.

Price variations differ across regions. Users in the Western United States have been hit harder, with p5e instance prices soaring from $43.26 to $49.75 per hour.

AWS has long leveraged price reductions as a key selling point to attract customers, thereby cementing its strong market position.

AWS designed capacity block services to meet the demand for GPU capacity. Users can pre-book specific GPU instances with time frames ranging from one day to several weeks and prepay at fixed prices.

An AWS spokesperson clarified, "The pricing of EC2 machine learning capacity blocks fluctuates based on supply and demand dynamics. This price adjustment reflects our projected supply and demand patterns for the current quarter."

For companies with enterprise discount plans, this price hike poses complex challenges. Enterprise discount plans typically offer a fixed percentage discount off the public pricing. A rise in public pricing means that even if the discount percentage stays the same, the absolute amount paid at the end will be higher.

The timing of the price hike is strategic, presenting a rare market opportunity for Microsoft Azure and Google Cloud Platform, both of which have been aggressively pursuing machine learning workloads.

Industry experts have noted that once a cloud service provider successfully raises prices on a particular service without experiencing a significant loss of customers, subsequent price increases become more feasible. Once this behavioral pattern is established, it will fundamentally alter the bargaining dynamics between enterprises and cloud service providers.

This price hike is merely a snapshot of the profound changes occurring in the global cloud computing industry. AI is reshaping the entire cloud services landscape, giving rise to new competitive dynamics and pricing strategies.

As various sectors worldwide transition to generative AI, GPU resources are becoming increasingly scarce. This supply-demand imbalance is not only driving up prices but also compelling cloud service providers to reevaluate their business models.

By 2026, hyperscale cloud service providers are projected to accelerate the construction of AI-centric data centers, with capital expenditures potentially surpassing $600 billion. Meanwhile, power and equipment supply chains have emerged as bottlenecks, with global projects worth $162 billion already delayed or canceled.

The cloud computing industry's focus is shifting from general-purpose computing to AI-optimized infrastructure. Enterprises are realizing that critical AI workloads need to be deployed close to data sources, whether in on-premises or hybrid cloud environments, to meet stringent performance, compliance, and data sovereignty requirements.

Hybrid AI is emerging as the new standard, with enterprises constructing intelligent 'AI factories' that seamlessly integrate AI pipelines into existing systems rather than relying solely on public cloud services. This model ensures organizations can responsibly scale AI applications while maintaining control over sensitive information and operational efficiency.

A growing number of enterprises are adopting multi-cloud strategies to avoid vendor lock-in and secure discounts through bulk service purchases. By 2026, multi-cloud and hybrid environments will become essential architectural components, with enterprises strategically deploying critical components in public clouds for scalability and leveraging private clouds for data security.

In the second quarter of 2025, the market size for full-stack AI cloud services in China reached 25.9 billion yuan, while the entire public cloud market exceeded 120 billion yuan. Although AI-related services are experiencing rapid growth, they still constitute a small portion of the overall market, indicating substantial future growth potential.

Behind the fluctuating numbers on cloud service pricing boards, industry analysts are closely watching whether RAM-intensive services will become the next target for price hikes. If the global RAM supply for data centers remains constrained, more enterprises will face cost impacts.

References:

https://m.sohu.com/a/973155956_114765?scm=10001.325_13-325_13.0.0-0-0-0-0.5_1334

https://www.theregister.com/2026/01/05/aws_price_increase

https://tech.it168.com/a2026/0104/6912/000006912696.shtml