AI Hardware Shortages: How Enterprises Can Do More With Existing Infrastructure

AI Hardware Shortages: How Enterprises Can Do More With Existing Infrastructure Key takeaways - AI hardware shortages are forcing enterprises to rethink how they plan, deploy and scale AI infrastructure. - GPU shortages, procurement delays and rising infrastructure costs are delaying enterprise IT projects, not just AI initiatives. - Organizations can reduce the impact of hardware constraints by improving utilization, increasing portability, optimizing existing infrastructure, and using MSP or cloud infrastructure before buying more AI hardware. - A flexible, software-defined infrastructure strategy can help enterprises build an AI strategy that is not limited by hardware availability. AI hardware shortages are creating new pressure for enterprise infrastructure teams. GPUs, specialized servers, memory, storage and networking capacity are all under strain as organizations race to support new AI initiatives. For many enterprises, the question is no longer whether they want to invest in AI. It is whether they can get the AI compute infrastructure they need, when they need it, at a cost the business can justify. A recent SUSE Customer Reference Panel survey of 110 practitioners and decision-makers found that AI hardware shortages are already affecting regular IT projects. Thirty percent of respondents reported significant or extreme delays, while 71% said they are frustrated by the impact of AI hardware shortages on their IT work. That frustration is understandable. AI infrastructure conversations often focus on the newest chips, the largest clusters and the most advanced accelerators. But the reality for most enterprise teams is more complicated. They must balance AI hardware demand against procurement delays, rising costs, budget constraints, security requirements, skills gaps and the need to keep existing systems running. The survey points to a clear conclusion: organizations cannot depend on hardware availability alone to keep innovation moving. They need an AI infrastructure strategy that helps them make better use of what they already have while maintaining the flexibility to run workloads across current and future infrastructure. What are AI hardware shortages? AI hardware shortages refer to limited availability, long procurement timelines or rising costs for the physical infrastructure needed to build, train, tune and run AI workloads. This can include GPUs, specialized accelerators, servers, memory, storage, networking equipment and the broader datacenter capacity required to support AI systems. For enterprises, AI hardware shortages are not just a procurement issue. They can affect infrastructure planning, application modernization, cloud strategy, security, compliance and the pace of AI adoption. As demand for AI compute infrastructure grows, many organizations are finding that the hardware they need is expensive, delayed or unavailable. This creates a practical challenge: how do enterprises keep AI and IT initiatives moving when the infrastructure required to support them is constrained? The answer is not simply to wait for more hardware or move to the cloud. Enterprises need to optimize existing infrastructure, improve workload portability and build an AI strategy that can adapt as hardware availability changes. How AI hardware shortages are delaying enterprise IT projects The hardware crunch is not theoretical. When asked how AI hardware shortages impacted their regular IT projects, 24.5% of respondents reported minor delays of one to three months. Another 23.6% reported significant delays of four to 12 months. Another group reported extreme delays of more than 12 months. The top constraints were exactly what many IT leaders would expect: - 39% cited not enough GPUs or specialized servers. - 36% pointed to hardware procurement and long lead times. - 18% cited memory or storage throughput limits. But hardware is not the only bottleneck. When respondents were asked about their biggest AI workflow challenges, the top answer was security, privacy and compliance hurdles at 43%. Skill gaps in AI infrastructure and workflows followed at 37%. The high cost of scaling and token usage sat close behind at 36%. That matters because it reframes the problem. Buying more hardware may help, but it does not solve the broader challenge of running AI in production. Organizations also need secure, compliant, efficient and manageable enterprise AI infrastructure that their teams can actually operate. How organizations are responding to AI hardware constraints When hardware is expensive or hard to get, organizations are not all responding in the same way. The survey found four main strategies. The most common response is to extend the lifecycle of existing servers and storage, selected by 30% of respondents. Another 15% are accelerating their shift to public cloud or infrastructure-as-a-service. Thirteen percent are waiting for the hardware they want to become available. Eleven percent are increasing virtualization and containerization to optimize existing hardware. Secondary strategies include paying a premium to get preferred hardware, reducing project scope or purchasing refurbished, preowned or third-party hardware. This mix of responses shows how difficult the current environment has become. Some teams are buying time. Some are shifting platforms. Some are re-scoping projects. Others are trying to extract more value from infrastructure they already own. The good news is that there is substantial room for optimization. One-quarter of respondents are using only 50% or less of their current infrastructure capacity, including 10% operating at 25% capacity or less. Another 54% report utilization between 51% and 75%. Altogether, 79% are operating below 75% capacity, highlighting a significant opportunity to optimize existing AI infrastructure before investing in additional hardware. The best AI hardware strategy may start with software One of the strongest findings in the survey was support for portable, efficient software. Eighty percent of respondents agreed that software that can run workloads efficiently across existing and new hardware is a better business investment than the latest cutting-edge hardware. That is a significant signal. Enterprises are not rejecting new hardware. They are recognizing that hardware alone is not a strategy. Portable software gives organizations options. It helps teams avoid locking innovation roadmaps to a specific vendor, chip type or infrastructure environment. It makes it easier to shift between on-premises systems, public cloud, hybrid cloud, sovereign cloud and future hardware platforms as business needs change. As one respondent put it: “The current hardware supply and cost climate has encouraged our organization to focus more on software portability, virtualization, cloud flexibility and infrastructure optimization rather than depending on specific hardware vendors.” That is the core lesson for enterprise AI: the winning strategy is not simply getting access to scarce compute. It is building an architecture that can keep moving even when compute availability changes. Flexibility is a business requirement The survey also shows that vendor neutrality, portability and flexibility are essential parts of AI infrastructure planning. When asked how the current climate has forced them to decouple innovation roadmaps from specific vendor hardware, respondents pointed to several common themes: budgetary strain, supply chain delays, cloud and modernization acceleration and a stronger focus on hardware-agnostic, software-defined and virtualized architectures. One respondent said: “The current climate has reinforced our focus on flexibility and interoperability, but it has not fundamentally changed our roadmap.” For organizations that already have a vendor-neutral strategy, the disruption appears easier to manage. As one respondent explained: “We’ve always gone after a vendor-neutral approach so the current climate hasn’t been too bad.” This is especially important for organizations pursuing AI in regulated, distributed or cost-sensitive environments. A team may want to run some workloads in the cloud, keep others on-premises, use sovereign cloud providers for data control or modernize applications gradually over time. A rigid infrastructure approach makes those choices harder. A portable, open approach makes them easier. Flexibility also reduces risk. If a preferred hardware platform is delayed, too expensive or unavailable, organizations must keep projects moving. If a cloud strategy changes, workloads should not be trapped. If new accelerators enter the market, teams should be able to adopt them without rewriting everything from scratch. How to optimize existing infrastructure before buying more AI hardware The pressure to acquire more AI hardware is real, but the survey suggests that many organizations have room to optimize first. Before adding more infrastructure, IT teams should ask: - Are we fully using the infrastructure we already have? - Can virtualization and containerization improve utilization? - Can workloads run efficiently across different environments? - Are we locked into specific hardware or vendors? - Can our architecture support cloud, hybrid, on-premises and sovereign deployment models? - Do we have the skills, security and compliance foundation to scale AI responsibly? These questions do not replace the need for new hardware. But they can help organizations make smarter decisions about when to buy, where to run workloads and how to protect AI investments from supply chain volatility. For many enterprises, the most practical AI infrastructure strategy starts with making better use of current systems. That means improving utilization, extending the value of existing servers and storage, supporting virtualization and containers, and ensuring workloads can move across environments as needs change. Build an AI strategy that is not limited by hardware availability AI hardware shortages are exposing a larger infrastructure truth. Enterprises need optionality. They need the ability to extend the life of existing systems when budgets are tight. They need to improve utilization across servers, storage and compute resources. They need to modernize without being forced into a single vendor, platform or cloud. And they need a practical path to AI that does not depend on waiting for the newest hardware to become available. That puts the full SUSE portfolio at the center of the strategy. SUSE AI Factory gives organizations an open, flexible foundation for building and running AI workloads with more control over where and how they run. It helps teams move AI forward without tying their strategy to a single hardware vendor, cloud provider or deployment model. SUSE Rancher Prime helps teams manage Kubernetes consistently across environments, from the data center to the cloud to the edge. That consistency helps organizations improve efficiency, reduce operational complexity and move workloads where they make the most sense. As AI and application demands grow, Rancher Prime gives teams a way to scale cloud native operations without adding unnecessary complexity. SUSE Virtualization solutions help organizations bring cloud native virtualization to their infrastructure strategy. By running virtual machines alongside containers, teams can improve efficiency, simplify operations and create a more flexible foundation for both traditional and modern workloads. For organizations trying to get more value from the hardware they already own, virtualization is not just a stopgap. It is an optimization strategy. SUSE Linux Enterprise Server helps organizations continue running critical workloads on existing and legacy infrastructure while staying current, supported and secure. When hardware is difficult or expensive to replace, that matters. Extending the useful life of infrastructure can give teams more time, more control and more room to prioritize modernization on their own terms. Together, these offerings help organizations respond to hardware scarcity with a smarter infrastructure strategy: extend what still works, optimize what is underused, modernize what needs to evolve and build AI on a foundation designed for portability and control. The organizations that move fastest will not necessarily be the ones with first access to the newest hardware. They will be the ones that can run efficiently across the infrastructure they have today, adapt to the infrastructure they need tomorrow and avoid being boxed in by any single vendor, platform or procurement cycle. AI may be driving the demand for more compute, but the path forward is not just more hardware. It is a smarter, more portable, more flexible infrastructure. FAQs about AI hardware shortages What are AI hardware shortages, and why are they affecting enterprises? AI hardware shortages happen when organizations cannot easily access the GPUs, specialized servers, accelerators, memory, storage, networking or datacenter capacity needed to support AI workloads. These shortages are affecting enterprises because AI hardware demand has increased rapidly, creating procurement delays, higher costs and greater competition for limited infrastructure resources. How are AI hardware shortages impacting enterprise IT projects? AI hardware shortages are delaying enterprise IT projects by making it harder for teams to get the infrastructure they need on time. In the SUSE Customer Reference Panel survey, respondents reported delays ranging from one to three months to more than 12 months. These delays can affect AI initiatives, application modernization, infrastructure upgrades and regular IT operations. How can organizations reduce the impact of AI hardware shortages? Organizations can reduce the impact of AI hardware shortages by improving infrastructure utilization, extending the life of existing systems, using virtualization and containerization, adopting portable software and designing workloads to run across multiple environments. AI infrastructure optimization can help enterprises move projects forward even when new hardware is delayed or expensive. Why is software portability important for AI infrastructure? Software portability is important because it gives organizations more flexibility in where and how they run AI workloads. Portable software helps enterprises avoid dependence on a single hardware vendor, cloud provider or infrastructure environment. This makes it easier to adapt when GPU shortages, procurement delays or changing business requirements affect AI infrastructure plans. Is buying more GPUs the only way to scale AI initiatives? No. Buying more GPUs can be part of an AI infrastructure strategy, but it is not the only way to scale AI initiatives. Many organizations can make progress by optimizing existing infrastructure, improving utilization, using virtualization, adopting cloud or hybrid deployment models, and building a more portable software foundation. The strongest strategy combines smart hardware investment with flexible, efficient infrastructure design. Build an AI strategy that isn’t limited by hardware availability. Find out more here. Related Articles Jun 03rd, 2025 The brains behind the books: Sushant Gaurav Mar 11th, 2025