From Constraint to Capacity Maximizing the Value of Existing Infrastructure as AI Reshapes the Enterprise

Perhaps the only thing getting as much attention as AI itself is the related shortage in computing power, as popular tools continue to gobble up data center capacity, GPUs, memory and other infrastructure.

That’s not just a problem for AI development, but across global technology contexts. Your next iPhone (or any other modern smartphone, for that matter) will almost certainly cost more as a result of skyrocketing memory costs, Apple CEO recently told The Wall Street Journal.

More urgently, the compute shortage is already impacting enterprise IT. 

CIOs and their teams aren’t just driving their organizations’ AI strategies – and delivering all of the hardware required to execute those strategies – but must continue to run their traditional workloads as well. But they’re running out of hardware, and buying more capacity is becoming prohibitively expensive, and that’s presuming they can buy more infrastructure to match their own increasing consumption.

Even if they can, it’s almost certainly not the best solution, especially not as prices continue to rise.

What to do instead? Get more out of the hardware you already have, run on a wider range of chips and offload to outside capacity when you need it, all without re-platforming.

It used to be the case that when you ran out of computing resources, you could just purchase more servers, processing power, memory and storage. That may be an oversimplification; budgets were still budgets, of course, but it was generally true.

As the global AI arms race has escalated at breakneck speed, it has shifted hardware manufacturers’ (such as chipmakers) focus toward AI-centric requirements such as high-bandwidth memory and specialized GPUs, leaving much of the traditional enterprise market to fight over whatever’s left in AI’s wake.

Even for the mythical enterprise with a virtually unlimited budget, this poses considerable operational problems. Memory prices are up nearly 95% quarter-over-quarter in Q1 2026, and data center vacancy rates are at a record low of 1.4%. If you want to purchase more hardware of your own, be prepared to wait multiple quarters (a lifetime in today’s business world) to take delivery.

No IT leader has time to wait. In addition to an enterprise’s own AI strategies, IT teams are still running their traditional workloads. Good luck telling your CEO or board “we just don’t have the capacity to accomplish all of our goals.”

In this context, another stat becomes particularly eye-popping: the typical enterprise server runs at 40-50% capacity. Enterprises are essentially leaving as much as half of their existing multi-million-dollar infrastructure investments unused, at the same time that costs for new hardware are surging.

As SUSE CMO Margaret Dawson put it recently, “the most valuable server in your fleet isn’t the one on backorder, it’s the one you already own that is running at half-throttle. Resilience isn’t about more iron; it’s about better software.”

It’s time for enterprises to rethink how they approach capacity. Stop approaching it as a hardware problem.

Focus on capacity as a software and management effort. Don’t look at capacity as something you simply buy; instead, approach it as something you recapture, broaden and optimize. In all likelihood, you’ve got plenty of hardware already, whether bare metal or virtual. You just need to unlock its full capabilities.

This approach helps enterprises reduce the manual work, modernize without ripping anything out and build on a platform that keeps you open, agile and ready for whatever comes next.

Think of it this way: if you had an employee who was only operating at 40-50% capacity, would you hire another full-time employee to do the other half of their job, or would you coach and lead them to better performance?

Here are some of the key strategies to maximize your existing infrastructure recapture, expand and optimize capacity, and how a proven partner like SUSE can help in 2026 and beyond.

Get more out of your hardware

Again, capacity isn’t necessarily a hardware problem, especially if you’re leaving half of it sitting idle. Do this instead:

• Converge VM and container clusters onto one platform so that legacy databases and new AI workloads can share the same machines. This is what moves ~40% utilization toward ~80%.

• Use fast NVMe as a second memory tier to delay expensive DRAM purchases. Memory is now more than half the cost of a new server, and is what is driving recent or anticipated price increases for virtually all kinds of hardware, from bare metal servers in your data center to iPhones and everything in between.

• Share GPUs across teams with SUSE AI multi-tenancy. This ensures AI workloads don't run into “noisy neighbor” processing problems that lead many organizations to decide they simply need to buy more (expensive and backordered) GPUs.

• Run on the silicon you can actually get. SUSE enables a wide range of accelerators — including NVIDIA, Huawei, AWS, and others — so that a shortage at one vendor doesn't stall your AI strategies. Few competitors can match this breadth, and it turns silicon choice into a way around scarcity and a competitive advantage in the race to AI business value.

Run everything with less effort and less disruption 

Needless complexity and manual operations vacuum up capacity and create at least the illusion that you need more hardware. You shouldn’t have to buy more machines to support the ones you already have.

• Manage tens of thousands of mixed Linux endpoints from one console instead of standing up new servers to manage other servers.

• Bring multi-cluster Kubernetes under control across hybrid and multi-cloud setups. Automated rebalancing and live migration keep utilization high without babysitting your infrastructure.

• Move off VMware one cluster at a time, on existing hardware, using the Coriolis-by-Cloudbase migration tooling that removes most of the pain enterprises experience in that shift.

• Go from community to supported Kubernetes with no change to the underlying setup.  

• Leverage certified software to prolong the productive life of legacy hardware and avoid premature equipment retirement.

The hours all of these moves free up are hours enterprise teams get back, and you can get started without a new budget cycle or an outage window.

Add capacity on-demand without owning it

• The same SUSE stack that makes the above strategies possible also runs in public cloud environments, MSPs, and AI neoclouds. That means you have a viable, attainable option for expanding capacity even when owning hardware isn't possible or just doesn't make financial or technical sense.

• When we talk about solving capacity challenges with software instead of buying more hardware, here’s an excellent example, especially as AI workloads put new and intensive demands on your infrastructure: SUSE customers can burst AI workloads to outside capacity without re-platforming, because it's the same software running everywhere.

• That gives you access to modern GPUs and compute without the delivery delays, CAPEX hit, or lock-in to one provider.

• This ties it all together:  own and optimize what makes sense, rent the rest, keep everything consistent, resilient and open as your capacity and other needs continue to evolve.

The financial impacts of this approach, especially when you rely on SUSE solutions to help implement it, are considerable. In fact, IDC research found a 258% average three-year ROI for organizations using SUSE Rancher Prime with Virtualization.*

Stack that and the costs of recovering and optimizing your existing capacity against the costs of buying new hardware in 2026, which are becoming exorbitant. Increasingly, it’s simply impossible to acquire new hardware at all. That means your strategic business initiatives will sit idle, either because the hardware costs will be untenable or because you just can’t get the hardware in the first place.

Capacity is not just a line item in the budget, either. It shows up in your sustainability reporting, since maximizing your existing hardware investments will reduce energy consumption and emissions from your data center and other infrastructure. That’s a particularly hot topic in the public eye right now, too.

It’s just as important to protect your IT estate from being held hostage by a single chip vendor or untenable lead times for new hardware. You retain control to make the right decisions for your business.

While this strategic shift might seem overwhelming, it doesn’t need to be. Start with actionable, attainable first steps that prove value and grow from there:

• Measure the current state of utilization and complexity in your estate today.

• Consolidate and run a pilot, add automation, then focus on continuous optimization.

• Make informed decisions about what to own and optimize versus what to offload to a cloud or neocloud.

• Define and track several clear metrics and then identify one step worth taking this quarter to improve.

Enterprises that overcome the hardware shortage won’t be the ones with the biggest purchasing power or the newest chips and GPUs. They are the ones who stop leaving half their infrastructure idle and keep their infrastructure – and their options – open.

SUSE makes that happen by turning underutilized compute into usable capacity, taking the manual work out of running a mixed IT estate and letting you modernize and add capacity on your own terms. We can help you do this across the hardware you own, the silicon you can get and the clouds and neoclouds you choose.

Learn more at suse.com.

* IDC Business Value White Paper, sponsored by SUSE, The Business Value of SUSE Rancher Prime with Virtualization, February 2025 | IDC #US52704924