Flexential Report: Enterprises Plan Years Ahead for AI Infrastructure

Flexential, a provider of secure and flexible data center solutions, has released new research showing that enterprises are planning their AI-related infrastructure requirements far earlier than they do for traditional IT needs. The company’s 2025 State of AI Infrastructure Report reveals that nearly four out of five organizations are mapping out future data center capacity more than a year in advance, with many projecting requirements up to five years ahead.

According to the study, 62 percent of IT leaders now plan one to three years in advance, while 17 percent are extending forecasts out three to five years. The lengthened planning cycle reflects the scale of investments and long procurement times tied to AI deployments, which demand significant amounts of computing, storage, and power. However, while 94 percent of respondents expressed confidence in their planning, the report suggests that this optimism may not be entirely justified. A notable 16 percent of organizations are planning less than a year ahead, yet 70 percent of them believe they are well prepared, despite evidence that the AI infrastructure market is becoming increasingly constrained.

Flexential’s survey included more than 350 IT leaders at companies generating at least $100 million in annual revenue, with around 100 participants representing firms exceeding $2 billion. Respondents were asked about readiness, workforce challenges, and the strategies they are adopting to scale AI. The findings show that data center vacancy rates are tightening, lead times for infrastructure are extending, and enterprises risk missing opportunities if they delay decisions. The report underscores that new capacity cannot be deployed quickly and that the traditional two-year planning window is now the minimum needed to remain competitive.

Even for organizations that have secured infrastructure, performance problems persist. Almost six in ten respondents reported experiencing bandwidth shortages in the past year, while more than half noted excessive latency. These challenges highlight that AI progress depends as much on robust network performance as on raw computing power. In response, companies are increasingly blending private and public solutions to balance performance and cost, with many adopting colocation to supplement cloud workloads.

Workforce constraints are another barrier. Only 5 percent of respondents said they experienced no AI-related skills gaps in the past year. The most acute shortfall lies in expertise with specialized computing infrastructure, cited by 61 percent of participants. Skills shortages in data science and engineering (53 percent) and advanced networking technologies (47 percent) were also common. Beyond technical know-how, enterprises reported broader hiring and retention difficulties, with more than half pointing to a lack of upskilling and retraining programs or difficulty sourcing candidates with advanced AI experience. Yet despite these obstacles, only 10 percent of organizations identified talent scarcity as their primary barrier to AI adoption, with most ranking infrastructure as the more urgent concern.

Public Cloud, Colocation, GPU-as-a-Service

The report also provides insight into how enterprises are structuring deployments. The public cloud remains the most widely used environment for AI training data, chosen by 68 percent of respondents. Colocation facilities are also gaining traction, with 54 percent using them to supplement storage and performance requirements. On-premises data centers are playing a reduced role, with just 20 percent storing training data locally and 36 percent planning to use those environments for training or inference workloads. GPU-as-a-service is becoming more common, now used by 40 percent of organizations, up from 34 percent a year earlier. Public cloud GPU usage also rose, from 30 percent to 34 percent.

Enterprises are simultaneously addressing workforce needs with training initiatives. Every organization surveyed reported offering some form of AI training, whether through embedded tools, in-house programs, or external certification opportunities. This reflects recognition that while infrastructure is critical, workforce preparation must evolve in parallel to fully realize AI’s potential.

The research emphasizes that building AI-ready infrastructure is a long-term endeavor requiring significant foresight. The ability to bring new capacity online takes years rather than months, and enterprises that fail to plan risk finding themselves without the necessary resources when demand peaks. Companies that move early are positioning themselves to align infrastructure with strategic ambitions, while also ensuring that energy efficiency and sustainability are integrated into future capacity. Flexential notes that demand is particularly strong in metropolitan hubs such as Atlanta, Dallas, and Denver, where it is continuing to expand.

The report concludes that as AI adoption accelerates across industries, planning horizons will continue to stretch further into the future. Enterprises that view infrastructure readiness as a competitive advantage, rather than a technical requirement, are likely to be better positioned to capture opportunities in an increasingly capacity-constrained market.