•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
The Financial Times reports that the AI boom is pushing major technology companies into an unprecedented effort to build new data-center infrastructure. However, power supply, labor availability, and regulatory constraints are increasingly turning large, multi-billion-dollar plans into delayed timelines.
While investors focus on chips such as Nvidia’s H100 and on large language models, a physical infrastructure bottleneck is emerging as a key risk. SynMax, a data and AI analysis group, told the Financial Times that nearly 40% of U.S. data-center projects scheduled to be completed this year are at risk of delays.
The affected pipeline includes facilities intended to serve the Microsoft-OpenAI alliance. SynMax also said that major construction sites are likely to miss deadlines by more than three months, a shift that could affect how quickly hyperscalers convert capital into revenue.
More than ten executives cited by SynMax said complexes requiring hundreds of megawatts are being held back by licensing hurdles, skilled-labor shortages, and shortages of electricity and equipment.
These projects are racing to deliver data centers designed for at least one gigawatt of output, widening the gap between investment scale and execution capacity. Wes Cummins, CEO of Applied Digital, said raising capital at this scale is difficult, but logistics, construction, and operations are even harder.
SynMax data, cross-checked with IIR Energy’s industry standards, suggests a different picture from some public schedules.
For example, SynMax said a 1.4 GW complex in Shackelford County, Texas, is being developed by Oracle to serve OpenAI. Vantage Data Centers said the first building is due in the second half of 2026, but satellite imagery from early April showed that among six planned sites, only one shows clear development.
SynMax estimated that the earliest handover for the first building could be December 2027 if the pace remains average.
In Milam County, Texas—where OpenAI cofounder Greg Brockman previously said a 1.2 GW site was “taking shape”—construction has started on only one facility. Among the group’s large Texas projects, only the Abilene site is expected to meet this year’s schedule.
OpenAI and Oracle said operations are proceeding as planned. OpenAI said its historic data-center build is accelerating in close collaboration with Oracle and SB Energy.
Executives cited by the report said the delays are not driven by funding alone. Two project executives said the market is severely short of skilled trades, including electricians and pipefitters, to meet the needs of increasingly complex campuses.
They also pointed to grid overload and shortages of hardware such as gas turbines and transformers. In remote locations, labor costs have risen by as much as 30%.
Satellite images reviewed by SynMax showed slow progress on six sites, with SynMax projecting that more than 60% of projects scheduled for next year have not yet started.
Josh Price, Capstone’s energy director, argued that the mismatch between rapid AI development and slower regulatory processes will likely lead to more scrutiny and potentially additional delays.
With AI expected to become a new growth engine for the global economy, the on-site delays raise a central question: whether tech giants can sustain the pace of AI expansion when the physical foundations of their operations show signs of strain.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
