Even the biggest cloud providers may be feeling the effects of the artificial intelligence boom. Amazon Web Services is reportedly putting stricter limits on how easily its engineers can spin up EC2 computing instances.
The development highlights a less visible pressure point in AI infrastructure. While much of the public conversation has focused on high-demand Graphics Processing Units (GPUs), Central Processing Units (CPUs) remain essential to keeping modern data centers and AI systems running smoothly.
What happened?
Tom's Hardware, citing The Information, reported that AWS used a May meeting with engineers to ask for less wasted CPU use so more capacity could be reserved for customers.
Some of those internal requests are reportedly taking much longer. In some cases, engineers are waiting days instead of hours for EC2 instances.
A core AWS service, EC2 underpins a wide range of online applications and private workloads. Inside the company, engineers have long had room to launch instances for their own use, partly because older web workloads often did not consume all available CPU capacity.
Tom's Hardware reported that AI agents are changing that balance. Because these workloads require more CPU-heavy orchestration, tool calls, and coordination around inference tasks, the older pattern of four or even eight GPUs for each CPU is said to be moving closer to parity.
TCD Picks » Quince Spotlight
💡These best-sellers from Quince deliver affordable, sustainable luxury for all
AWS rejected the suggestion that these changes point to a new capacity crunch.
"Demand for AWS services, including EC2, is incredibly strong and growing," an AWS spokesperson said. The spokesperson also disputed the broader framing, saying, "This narrative on EC2 is inaccurate" and "encouraging efficient use of resources isn't unique to Amazon."
The spokesperson also said AWS routinely works with teams to reclaim idle instances, right-size workloads, and scale resources efficiently.
Why does it matter?
If cloud providers have to manage internal compute resources more carefully, it suggests demand is rising quickly enough to affect how software gets built behind the scenes.
TCD Picks » Quince Spotlight
💡These best-sellers from Quince deliver affordable, sustainable luxury for all
That could eventually influence pricing, availability, and how quickly new AI-powered services reach the market.
AI's infrastructure demands extend well beyond flashy chatbot tools. CPUs play a central role in the background work that makes AI agents function, and when supply tightens, companies may find themselves competing for capacity from Intel, AMD, Amazon's Graviton5 chip, and newer products such as Nvidia's Vera CPU.
There is also a direct link between AI growth and strain on the energy grid. Larger data center footprints can drive up electricity demand, increase water use for cooling, and add pressure on local utilities, potentially contributing to higher energy costs or delays in grid planning.
Tom's Hardware also reported that any shortages appear to be more concentrated in spot instances, while a consultant told The Information that contracted capacity has not experienced the same issue.
What's being done?
AWS says the move is an efficiency effort, not an emergency response.
Reclaiming unused instances, matching workloads to the right size, and scaling resources up or down as needs change are all part of standard operating procedure, the company said.
Chipmakers are responding to the broader demand as well. AMD has introduced its new Zen 6 "Venice" data center CPUs, while Amazon continues deploying its Graviton5 chip.
Those launches reflect a market no longer focused solely on GPUs, but on the full stack needed to support more advanced AI systems. AI services depend on massive physical infrastructure, and their convenience comes with real costs.
Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.
