Financing is catching up with physics in the race to build the next generation of artificial intelligence (AI), and for the first time, some of the most ambitious projects are being quietly scaled back rather than expanded.
From compute-heavy research teams to mega-data centers, the story across the industry is the same. The money, the power grid, and the hardware can no longer keep up with every big idea.
The Stargate Expansion That Didn’t Happen
Oracle and OpenAI have cancelled plans to expand their flagship Stargate data center project in Abilene, Texas, a project that was supposed to develop roughly 2 gigawatts of capacity. It was a planned 600-megawatt addition that would have turned the site into one of the most powerful AI data centers in the United States.
The existing Abilene campus, which is a 1,000-acre, eight-building facility, remains under construction and is not affected. But the reasons the expansion fell apart reveal the structural problems now facing large-scale AI infrastructure across the board.
Negotiations between both companies broke down over financing terms and OpenAI’s frequently shifting demands on data center capacity.
There is also the power problem. Expensive power usage as well as its scarcity has shifted from a temporary bottleneck to the primary constraint limiting AI and hyperscale growth. As such, utilities and regulatory bodies across the U.S. and Europe are now revising tech companies’ usage of electricity and grid interconnection.
Community Opposition Is Blocking Billions
Beyond financing and infrastructure, a third factor being public resistance is reshaping the AI buildout. Data center projects worth billions of dollars have been delayed or altogether blocked since 2023, according to Data Center Watch. In the second quarter of 2025 alone, an estimated $98 billion in projects were blocked or delayed, which was more than in all previous quarters since 2023 combined.
This is so due to concerns that include the environmental and health impacts of these AI projects, rising local electricity bills, and low expected job creation.
The Financing Structure Is Under Strain
Even the money flowing into these projects isn’t as straightforward as these tech companies explain it to be. A funding method known as circular financing has defined much of the sector’s investment activity, where investors pour capital into AI startups, which then spend that money on the investors’ own chips, cloud services, or infrastructure. OpenAI, Microsoft, Nvidia, Oracle, AMD, and SoftBank all took part in this model throughout 2025, and are still taking part in it.
Additionally, Oracle is reportedly carrying more than $100 billion in debt to fund its Stargate commitments. And fellow Stargate investor SoftBank is seeking $40 billion in loans to fund OpenAI’s expansion.
What This Means Going Forward
The scale of investment planned for AI infrastructure remains enormous. The largest hyperscalers including Google, Amazon, Meta, Microsoft, Oracle, OpenAI, are expected to spend a collective of over $700 billion in 2026 alone, much of it on data centers and the hardware that fills them.
The amount of money going behind these projects implies that the largest AI efforts are no longer governed only by ambition or technical possibility. They are also tightly bound by capital costs, energy availability, and the politics of resource allocation inside these tech companies.
And this also means the downsizing of some of the world’s most ambitious AI projects is a sign that the industry has hit the point where economics and engineering constraints have to be managed as carefully as the tech itself.
