Navigating cloud constraints: Limited power, capacity, and skills

Cloud providers love to advertise cloud as having “almost unlimited” resources, storage, or compute power. But with the rise of AI, hyperscalers have shifted their messaging from speed, scale, and innovation to words like efficiency and optimization. 

That shift means something: The cloud is not unlimited, and organizations are starting to feel the strain.

From limited power, data centers, and chips to diminishing ROI, learn what cloud constraints mean for your organization’s cloud computing strategy.

There isn’t a single cause constraining cloud computing; there are multiple factors working together.

The biggest constraint right now is pure lack of power. AI demand is only increasing, and the power grid doesn’t have the capacity to keep up. Already, data center development is slowing down due to limited energy—and it’s likely going to get worse.

According to MIT, the power going to AI-specific purposes will rise to 165 - 326 terawatt-hours per year by 2028. That’s more than all electricity currently used by US data centers for all purposes (and enough to power 22% of US households each year).

As AI-driven energy demand continues to grow faster than the infrastructure that supports it, organizations must prepare for delayed projects or turn to other options like private energy producers.

The Pluralsight 2023 State of Cloud Report found that more than 70% of organizations struggle to drive customer value with the cloud.

Today, getting value from the cloud is still a major concern: According to a VMWare report, nearly half of organizations (49%) believe more than a quarter of their public cloud spend is wasted.

And according to the 2026 State of FinOps Report, practitioners have been seeing “diminishing returns in cloud optimization.”

Optimization can only do so much. For organizations that haven’t planned properly, the cost of cloud is becoming more expensive than the value it’s delivering.

In theory, cloud’s elasticity allows organizations to scale computing resources up or down based on changing workloads and real-time demand. In reality, this elasticity is constrained by chip supply, data centers, and other factors.

Cloud providers have even started new services, like Amazon EC2 capacity blocks, that allow you to reserve compute for future start dates. The rise of these capacity reservation services signals that GPU capacity isn’t truly elastic, putting yet another constraint on cloud computing.

In addition to technical constraints, skills are another limiting factor for organizations looking to drive value with cloud. 

Cloud is one of the top tech skills gaps, and 48% of IT professionals have had to abandon projects partway through due to a lack of tech skills.

What's more, organizations are increasingly looking for cloud professionals with specific skills and experience tailored to their tech stack and cloud architecture. These experts are harder to find, and without them, projects never make it to production.

Design systems with constraint in mind, rather than trying to optimize or reconfigure down the line. This will save time in the long run and help you get the most value with the resources you have.

Before doing anything, assess your workloads. What do they need to be able to do? What real-world constraints will impact them? This might be latency, cost, power consumption, or something else.

You can’t build for everything—identifying your constraints will determine which trade-offs to make. For example, lower latency will increase costs.

One constraint that should be on everyone’s list is cost. In his frugal architecture framework, Dr. Werner Vogels, Chief Technology Officer at Amazon.com, says to treat cost as a nonfunctional requirement. If it’s not part of the design process from the beginning, costs will quickly outpace revenue. 

Learn more about frugal architecture.

You need a way to measure your systems to understand costs and, by extension, ROI. 

Metrics like time to market and feature velocity are helpful, but you need more information to determine if you’re getting the most value from your cloud investments and infrastructure.

Efficiency metrics like cost per transaction, resource utilization, and revenue per resource unit can help you optimize your systems over time and strike the right balance between budget, performance, and business needs. The exact metrics you use should align with broader business outcomes.

Provider limitations can narrow down your options and influence your architecture decisions. Consider data center availability, chip supply, and regional capacity so you can build backups into your strategy instead of scrambling for alternate solutions later.

Power availability and data center connections may be out of your hands, but there’s one thing you can control: Upskilling your people. 

Ensure they have foundational cloud skills like networking, identity, and cost management. From there, boost their platform-specific knowledge.

Training your teams in AWS, Azure, or GCP (or all three)? Uncover tips to build their multicloud mastery.

As demand grows, organizations that build with constraints will be better equipped to navigate future needs and tech changes. On the other hand, organizations that provision whatever they need will likely see increased costs, more frequent adjustments, and delayed or abandoned projects.

It’s easy to see cloud constraints as barriers that impede your organization’s growth. And while that may be true, constraints also encourage creativity and innovation. You may be surprised at what your people can do. 

Need to build your team’s cloud skills? Try implementing one of these upskilling strategies.