The global “chip shortage,” a familiar explanation for rising hardware costs and delayed projects, is an increasingly broad and unhelpful label for enterprise technology leaders. The real pressure is actually from memory, particularly high-bandwidth memory (HBM) driven by surging AI demand.
This reality is keenly felt at enGen, the technology and operations arm supporting a large, regulated healthcare enterprise. Operating at the intersection of ambitious AI adoption and its inherent physical, financial, and operational constraints, enGen understands this shift first-hand.
"There is just a huge spike of requests for these new GPUs," said John "Rez" Rzeszotarski, vice president of infrastructure strategy at enGen. This surging demand for HBM-reliant GPUs and accelerators is profoundly reshaping the entire hardware supply chain.
Rez explained that this shift has caused manufacturers to reprioritize. "The same providers basically stopped producing the normal memory and only started investing in this high-bandwidth memory because they were making more profit."
The ripple effects are already widespread: as HBM is prioritized, availability tightens for traditional components, impacting the servers, networking equipment, and storage systems organizations rely on daily.
From Supply Chain Shift to Leadership Problem
At enterprise scale, this kind of disruption shows up in budgets. Routine infrastructure refreshes are now colliding with pricing that no longer reflects historical norms. In some cases, core infrastructure components are seeing increases of 150% to 200%, with outlier quotes reaching several times expected cost.
“If we bought everything right now, we’d be significantly over budget,” said Rez. “That’s a tremendous amount of money.”
At the same time, waiting isn’t always viable. Systems age out. Capacity needs increase. New initiatives depend on infrastructure being in place.
A More Deliberate Way to Decide
In response, organizations are moving away from static planning cycles and toward more deliberate, structured decision-making.
Rather than reacting to volatility, enGen applies a consistent decision framework to infrastructure planning that is designed to keep the organization moving forward without absorbing unnecessary risk.
Every major investment is evaluated through a small set of consistent lenses:
- Need now: investments that cannot be deferred without operational or business risk
- Maximize: increasing utilization of existing infrastructure to reduce new demand
- Hold: delaying purchases where systems can be extended safely
- Pivot: reconfiguring components or architectures to achieve similar outcomes more efficiently
“It really comes down to asking the right questions,” Rez said. “Do we need this now? Can we get more out of what we already have? Or is there another way to solve the problem?”
This discipline allows enGen to continue modernizing infrastructure even as AI demand accelerates, without locking the organization into unsustainable cost structures or reactive purchasing decisions.
“It really comes down to asking the right questions,” Rez said. “Do we need this now? Can we get more out of what we already have? Or is there another way to solve the problem?”
-John "Rez" Rzeszotarski, VP of Infrastructure Strategy at enGen
Efficiency Becomes a Strategic Lever
One of the clearest lessons from enGen’s experience is that constraint often forces better architecture. As hardware becomes more expensive and less predictable, the incentive to increase efficiency grows. enGen has focused on operating models that emphasize shared capacity, dynamic workload placement, and higher infrastructure density - approaches that resemble cloud principles applied within enterprise environments.
“Density becomes a huge lever,” Rez explained. “If you can architect environments so you’re using what you already have more effectively, you don’t need to buy nearly as much…even in a tough market.”
Those gains reduce spend and, for enGen, they also reduce exposure to supply‑chain volatility and create a more resilient foundation for both AI and non‑AI workloads.
Looking Ahead: Memory Today, Power Tomorrow
While memory remains today’s most visible constraint, enGen is already planning for what comes next.
As AI systems scale, infrastructure readiness increasingly depends on power availability and delivery. Compute‑dense environments require reliable electricity, cooling, and grid capacity - factors that extend beyond traditional IT control.
“You can build the systems,” noted Rez, “but eventually you have to answer how you’re going to get enough power to them, reliably and at scale.”
From enGen’s perspective, this “last mile” of infrastructure may soon become as critical to AI progress as compute itself.
“You can build the systems, but eventually you have to answer how you're gonig to get enough power to them, reliably and at scale.”
-John "Rez" Rzeszotarski
The Bottom Line
At enterprise scale, AI success is shaped by how organizations manage the unglamorous work underneath: infrastructure planning, cost discipline, and decision‑making under uncertainty.
“Everybody wants to talk about the AI you see on the front end,” said Rez. “Very few people talk about everything that has to exist behind that to make it real.”
The organizations that are willing to have those difficult conversations will be better positioned to scale what comes next.