The AI Race Is Exposing a Costly Blind Spot in Corporate IT

The race to build the infrastructure powering artificial intelligence is creating an uncomfortable contradiction for businesses. While technology companies pour billions into expanding data-centre capacity, organizations that depend on traditional commercial infrastructure are finding it even more difficult to secure affordable space. The pressure is exposing a problem that has been accumulating for years, with many businesses making expensive technology decisions without fully understanding their long-term financial consequences.

According to CBRE's North American Data Center Trends report, vacancy across primary North American markets fell to a record low of 1.4% in the first half of 2026, despite supply increasing by 33.7% year-on-year. More than 80% of the capacity under construction had already been preleased, leaving businesses competing for an increasingly restricted pool of available infrastructure.

John McCabe, president of The Cavan Group, believes the industry is confronting a structural challenge that has received insufficient attention. Public discussion around data centres has become increasingly dominated by artificial intelligence, creating a perception that infrastructure expansion is primarily an AI issue. However, commercial organizations continue to depend on these facilities to support their everyday operations.

"There is a conflation of the current drive for AI data-centre build-out and the need to support commercial data centres for small and medium businesses," McCabe explains.

Growing public resistance to new developments, particularly over electricity consumption and local infrastructure demands, is making expansion difficult. Communities across the United States have challenged proposed developments, with some municipalities introducing restrictions on new projects. For businesses dependent on colocation facilities, the consequences are becoming increasingly tangible.

McCabe points to a growing supply crunch as AI companies absorb available commercial data-centre capacity. Colocation providers that previously offered relatively stable pricing are now facing pressures that increasingly affect their customers. "Now we have a big supply crunch on the colocation opportunities for regular businesses," he observes. Floor space and power, two fundamental components of any hosting arrangement, are becoming more expensive.

The challenge is compounded by the economics of cloud computing. McCabe explains that cloud adoption was often positioned as a more efficient approach to infrastructure management, allowing businesses to reduce their dependence on physical data centres. Yet the financial benefits are not guaranteed. Poorly optimized workloads can generate substantial expenditure, particularly as consumption increases.

McCabe has already observed organizations reconsidering their reliance on public cloud services. Infrastructure repatriation, where selected workloads move back into privately managed or on-premises environments, is gaining attention as businesses reassess their hosting costs.

"The cost of cloud computing, when not well thought through, can be enormous if you're not optimizing your workloads to run in the cloud," he explains.

This does not make cloud infrastructure an inherently poor investment. Some workloads benefit from its scalability, while others may be better suited to dedicated infrastructure or colocation arrangements. Hybrid architectures allow businesses to distribute applications across different environments according to their financial and operational requirements.

The more pressing question for CIOs and CTOs is whether they understand the economics of those decisions. McCabe suggests that infrastructure choices are often made around immediate requirements, leaving long-term operating expenditure insufficiently examined. As market conditions change, arrangements that once appeared commercially attractive can become progressively expensive.

McCabe believes infrastructure assessments should receive the same scrutiny applied to major corporate transactions. His observations draw on a growing overlap between traditional data-centre strategy and IT due diligence in mergers and acquisitions. Both require an examination of existing systems, contractual commitments, and the infrastructure supporting business operations.

He finds that the connection is relevant in acquisitions, where technology costs and operational dependencies can materially influence a transaction's value. An organization may inherit infrastructure arrangements that carry substantial financial exposure or limit future growth.

Artificial intelligence introduces another dimension to this assessment. Businesses considering AI adoption must understand where their workloads should operate and how resource consumption will affect expenditure. McCabe points out that disaster recovery also deserves renewed attention, as hybrid environments introduce additional dependencies across cloud platforms and physical infrastructure.

According to McCabe, infrastructure governance must become an ongoing management priority. He contends that organizations need to understand how their hosting arrangements affect operating expenditure and whether their existing workloads remain appropriately positioned. "Companies really need to make closer note of their application workloads and make more frequent changes to deal with the uncertainty of pricing in the market," he says.

His warning points to a responsibility that technology leadership can no longer afford to delegate entirely to infrastructure teams. McCabe believes that CIOs and CTOs must regularly reassess whether their technology investments remain commercially viable as cloud pricing, colocation availability, and business requirements evolve.

The AI boom has made infrastructure a more visible corporate concern, but its financial implications extend well beyond companies developing artificial intelligence. Ultimately, McCabe argues that the most expensive infrastructure decision for CIOs could be the one they never contemplated questioning.