As AI-driven demand for memory and flash storage puts pressure on the broader technology supply chain, state, local, and education organizations are likely to feel the impact through rising costs, longer lead times, and greater uncertainty around infrastructure planning. These market conditions arrive in an already constrained operating environment, marked by tight budgets, lean teams, and aging systems.
As a result, IT leaders have a purchasing problem – and a challenge to modernize, strengthen cyber resilience, and keep essential services running while market conditions keep changing, Stephen Savas conveyed in a recent discussion with MeriTalk. The conversation with Savas, vice president of U.S. state, local, and education at the enterprise data storage and data management company Everpure, explored how AI demand is reshaping the data infrastructure market and how public sector organizations can respond effectively.
MeriTalk: You’ve worked closely with public sector technology leaders throughout your career. As you continue stepping into your role leading SLED at Everpure, what are you hearing from IT leaders about the pressures they’re facing right now?
Savas: It’s really the same almost everywhere: Do more with less. Budgets are not growing, but leaders still need to modernize aging systems, defend against cyber threats, and get ready for AI – all at once, with historically lean teams.
What has changed is the foundation underneath all of that. Storage and memory used to be easier to plan around. Now costs are up, lead times are longer, and a refresh that was budgeted 18 months ago no longer stretches as far as leaders planned. That kind of uncertainty makes it difficult for organizations to achieve their goals.
MeriTalk: The current memory and flash storage shortage is driving costs and lead times up. What makes this a structural infrastructure challenge, and what risks do organizations take if they try to wait it out?
Savas: It is structural because the demand is coming from the AI build-out. Hyperscalers are consuming memory, flash, and GPUs so quickly that they are reshaping the entire supply chain. When they lock up supply, public sector customers are competing for what is left, and they usually have less leverage because their procurement processes move more slowly.
We have all heard forecasts that suggest this will clear up in a year, but I do not think leaders should treat this like a short-term spike. This is a multi-year problem. Waiting it out is really a bet against the market continuing to move in the same direction.
Organizations that defer usually end up paying more. They face longer wait times, and they still have aging hardware in production, which creates delivery risk and cyber risk. If something fails, they can end up in an emergency buying situation, and that is the most expensive way to procure anything. The better move is to change the model, not try to time the market.
MeriTalk: For state, local, and education organizations specifically, how do fixed budget cycles, procurement rules, and staffing constraints affect how IT leaders need to respond to volatility in the memory and flash storage market?
Savas: The public sector is different. A business can absorb a midyear price increase or shift capital around. State and local governments, agencies, and school districts are working with appropriated budgets, a procurement process built for fairness rather than speed, and often a multi-year capital cycle.
When the market moves faster than the budget, government leaders get squeezed. Their dollars were set under last year’s prices, but in some cases, prices could increased by as much as 300 percent, according to Gartner.
The Arkansas Department of Transportation is a good example of changing the model instead of fighting the system. The department runs critical 24/7 traffic management and emergency dispatch systems, and unpredictable IT costs make it almost impossible to plan years in advance. By moving to storage as a service, their infrastructure manager can now tell leadership with real accuracy what costs will look like five years out.
That consumption model also helped the department avoid building a second data center, saving more than $9 million. That money went back into critical services and infrastructure. And because upgrades happen during normal business hours, the team saved more than 40 hours a day in maintenance. The lesson is: Do not fight procurement. Adopt a model that makes a fixed budget behave predictably.
MeriTalk: Many state and local governments and education organizations are trying to modernize applications, strengthen cyber resilience, and prepare for AI-driven use cases. How can leaders balance those innovation goals with the need to control costs and keep core systems running?
Savas: It is a constant struggle. Under budget pressure, the instinct is to treat innovation and stability as if they are competing for the same dollars. That usually shortchanges both.
The better move is to stop overprovisioning the core, because that frees up money and capacity for new work. A lot of state, local, and education environments carry extra storage they bought ahead of demand because that used to be the safe play. In today’s market, idle capacity is incredibly expensive.
Leaders need to right-size the footprint and redirect savings to modernization and resiliency. AI readiness does not mean making a huge purchase on day one. It means building a foundation that is efficient and flexible enough to scale as AI workloads arrive.
Keeping the lights on and innovating are not opposites. An efficient, predictable foundation makes both affordable.
MeriTalk: What should state, local, and education IT leaders look for in a data infrastructure model that can help them move from volatility to predictability?
Savas: Three things. The first is predictable cost. You know what you will pay, and you can align spending to what you actually use.
Second, do not overprovision. Scale capacity as needed instead of buying years ahead and hoping your forecast is correct.
The third is flexibility, and I think this is the biggest one. Leaders need to grow into AI and modernization without re-architecting or repurchasing every time something changes.
The test is simple: Does the model let me plan with confidence inside a budget cycle I cannot easily change? If the answer is yes, you are operating on your own terms instead of constantly responding to the market.
MeriTalk: How does Everpure help state and local and education organizations build a more predictable, flexible, and efficient operating model, especially when budgets, workloads, and technology demands keep changing?
Savas: We start with what organizations actually use, because most savings come from cutting the overprovisioning that quietly drains budgets. From there, we structure infrastructure to scale with real demand, which helps protect organizations from the price and lead-time swings that the AI crunch has created.
The Mississippi Department of Revenue is a good example. The department collects more than $9 billion a year in funds for roads, public safety, hospitals, and schools, and its tax and vehicle registration systems have to stay available across 82 counties. As their data grew, backup and recovery became a pressure point.
Their architecture now scales in a linear way, so capacity grows with data instead of forcing expensive upgrades. That scale-as-you-grow means the organization can add capacity in step with demand rather than overbuying.
An earlier platform move more than doubled performance and saved the department hours in maintenance. That kind of headroom lets a lean team focus on the mission instead of maintenance. That is what a flexible, predictable operating model looks like. In a volatile market, predictability itself is the advantage.