A crack is forming in what has seemed like an unstoppable wave of AI infrastructure investment. OpenAI and Oracle have abandoned plans to expand their flagship data centre facility in Abilene, Texas, according to reporting from Bloomberg, citing sources familiar with the matter. The proposed expansion would have increased capacity at the site from 1.2 gigawatts to as much as 2 gigawatts. Instead, negotiations stalled over financing constraints and OpenAI's inability to accurately forecast its own computing demand.
This represents the first visible breach in the tech industry's grand AI infrastructure narrative. For months, executives at OpenAI, Meta, and other hyperscalers have committed hundreds of billions to data centre buildouts with barely a hint of compromise. Yet here we are, barely weeks after OpenAI secured a $110 billion funding round, announcing a retreat from a major expansion.
The timing is instructive.The Abilene facility was planned to grow to as much as 2 gigawatts of capacity, up from the 1.2 gigawatt campus currently under development.The cancellation was attributed to a combination of financing and OpenAI's inability to forecast demand effectively. This is not a trivial issue. Demand forecasting sits at the heart of rational capital allocation. If OpenAI cannot predict its own compute requirements, confidence in the broader investment thesis should reasonably diminish.
The market appears to be moving quickly to fill the gap.Meta is reportedly in the running to lease the untapped and unbuilt capacity from the site's developer Crusoe after Nvidia stepped in to broker a deal. According to reporting, Nvidia put down a $150 million deposit on the future capacity before approaching Meta about possibly moving in together. This arrangement has merit. Meta, whichannounced its intention to plow up to $135 billion into capital expenditures with an eye for GPU compute capacity during its Q4 earnings call in January, has the firepower to commit. The swap also illustrates how quickly capital can pivot when opportunities shift.
Yet the broader Stargate initiative remains intact.OpenAI and Oracle entered an agreement in July to develop up to 4.5 gigawatts of additional capacity through a partnership that exceeds $300 billion between the two companies over the next five years.Construction of the flagship Abilene facility is progressing, with parts of the facility now up and running. Oracle began delivering the first Nvidia GB200 racks and early training and inference workloads have commenced. The pullback is not a collapse; it is a recalibration.
The broader spending environment remains intense.The eight largest hyperscalers, including Google, Amazon, Meta, Microsoft, Oracle, Tencent, Alibaba, and Baidu, are expected to spend a collective $710 billion on data centre infrastructure and related equipment in 2026 alone. This is rational behaviour given the stakes. Whoever controls the most efficient computing capacity will dominate AI deployment. Meta, Google, and Amazon are not building for vanity; they are building for competitive survival.
Yet OpenAI's retreat raises legitimate questions about project viability and financial discipline. The company has demonstrated extraordinary confidence in its own growth trajectory.OpenAI has been telling investors that it is now targeting roughly $600 billion in total compute spend by 2030, after CEO Sam Altman previously touted $1.4 trillion in infrastructure commitments. The company is providing a lower number and more defined timeline for its planned spending as broader concerns mounted that expansion ambitions were too great for the potential revenue that would follow. A more honest revision of forecasts is overdue across the industry.
The question now is whether this represents an isolated stumble or the beginning of a wider adjustment. Investors in the hyperscaler ecosystem should be watching demand forecasts and utilisation metrics closely. Confidence in AI infrastructure spending has been built on faith that capacity will fill. If major users cannot justify their own buildouts, others will struggle to justify theirs. For now, the core Stargate initiative continues. But the industry's confidence in its own projections should have just declined.