Not only "power shortages," but also "delays" will become the "key theme" for U.S. data centers in 2026.

Not only "power shortages," but also "delays" will become the "key theme" for U.S. data centers in 2026.

```

As the construction of AI infrastructure enters deeper waters, "delay" is gradually replacing the simple issue of "power shortage" and becoming the focus of the industry. Several executives and investors have pointed out in recent discussions that project delays will become the key theme in 2026, and this trend of divergence will distinguish world-class operators from ordinary participants.

The current "delay" is an extremely broad concept, with complex and diverse causes. From utilities failing to fulfill power commitments, key equipment arriving late, to construction delays caused by labor shortages on site, all kinds of situations abound.

For builders aiming to construct large-scale AI infrastructure, the complexity of delivery speed and scale is beginning to show: the waiting time for generators, transformers, and liquid cooling components is growing longer, and because Nvidia’s most advanced hardware requires liquid cooling, this still-experimental technology faces many unresolved challenges.

This widespread industry lag is triggering a chain reaction in the capital markets. Recent cases show that not all delay impacts are the same. What truly worries the market are delays involving political resistance, financing failures, and binding contracts falling through. These factors not only directly impact the stock prices of the relevant companies but also force some capital providers to reevaluate the economic viability of the projects.

Meanwhile, industry giants are attempting to tackle this bottleneck at its source. Last week, Nvidia CEO Jensen Huang secretly convened a summit at the company’s California headquarters, aiming to promote the "behind-the-meter" power generation model, in which data centers build their own power facilities to bypass the long wait for grid interconnection.

For the industry, whether power supply and engineering delays can be resolved quickly will determine who survives in the AI race.

Political Resistance and Financing Challenges

Some project delays do not stem from technical issues, but instead encounter political resistance from local governments—especially as data center scale jumps from hundreds of megawatts to gigawatt (GW) level.

Take for example a project in Saline Township, Michigan, where last September a vote opposed rezoning farmland for a large AI data center. Developer Related Digital soon sued the township, and eventually local officials changed their stance a few weeks ago. The project, planned by Related Digital and Oracle for OpenAI, will reach a total capacity of 1.4 gigawatts. The Michigan energy regulatory body also approved an expedited power request for the site this week.

Nevertheless, such local opposition has had real consequences. Before construction begins, Oracle and Related Digital need to arrange over $10 billion in project financing. Well-informed sources told The Information that investor Blue Owl decided last week not to participate, concerned that the project's lease and debt terms were less attractive than other Oracle-backed deals, and that persistent local resistance might cause further delays.

Though Oracle claims Blue Owl's decision hasn't affected the schedule and that other financing channels may be available, the news still caused Oracle’s stock to drop by 5%.

Non-exclusive Agreements Falling Through

Another common type of delay occurs when developers and customers sign only letters of intent or are in exclusive negotiations, but the customer ultimately decides not to sign a formal lease.

Data center developer Fermi, co-founded by former Texas Governor Rick Perry, recently suffered such a setback. The company disclosed in regulatory filings that its West Texas AI server campus failed to sign a lease with tenants before the deadline.

According to Business Insider, the tenant that ultimately pulled out was Amazon Web Services (AWS). This news sent Fermi’s stock plummeting by over 40%.

Such events are not rare in the industry. Earlier this year, Microsoft decided not to sign further data center deals with CoreWeave in Texas; tech giants like AWS and Microsoft also abandoned leasing facilities developed by Applied Digital in North Dakota. Because lenders typically only begin funding once a developer holds binding contracts with top customers, letters of intent are insufficient to unlock billions in funding.

Contract Defaults and Damaged Economics

Compared to the cases above, more disastrous delays involve projects that have secured funding and are already significantly under construction.

Sources told The Information that a few AI data center projects have suffered so many problems that customers exited binding agreements before completion.

Such situations pose a direct threat to the project’s survival. In some extreme cases, due to severe delays, developers have had to offer to halve the rent for Nvidia servers as a concession. Industry veterans say data center contracts typically favor tenants, so if a facility isn’t completed on time, clients don’t have to pay full price and can walk away altogether. When delays result in deep price discounts for customers, the project's economic viability is severely impacted.

For investors, the core concern is no longer whether a project will be delayed, but whether the team behind it has the ability to solve problems and deliver.

Nvidia's "Secret" Power Summit

Facing infrastructure bottlenecks, Nvidia is taking more proactive steps. Last week, Jensen Huang convened about 100 executives for a closed-door meeting at the Santa Clara headquarters, focusing on the core limiting factor in the AI boom: power.

According to sources cited by The Information, attendees included global heavyweights in energy and electrical infrastructure such as Schneider, GE, Hitachi, and Siemens Energy, as well as some cloud service providers. Huang reiterated that AI has turned data centers into "factories"—power goes in, intelligence comes out. Nvidia’s message was clear: whoever can quickly ramp up power supply will seize the next wave of AI demand.

The summit highlighted the need for "behind-the-meter" electricity. Companies no longer wait months or even years for public grid interconnection—instead, they opt for onsite power generation. For example, OpenAI’s facility in Abilene, Texas, uses this model. Huang also called for the power industry to improve the energy efficiency of data centers, increasing the number of AI tokens generated per watt of electricity.

In the data center and power infrastructure sector, mergers and new initiatives abound:

Google acquires Intersect: Google parent company Alphabet announced a $4.75 billion acquisition of Intersect, which has helped Google locate land adjacent to high-quality wind, solar, and battery power—so-called "electrified land."Regulatory calls: Vermont Senator Bernie Sanders has called for a halt in the construction of AI data centers.New project partnerships: Texas Pacific Land Corporation announced a strategic agreement with Bolt Data & Energy to develop large-scale data center campuses. Bolt was co-founded by former Google CEO Eric Schmidt.

Risk Warning and DisclaimerThe market involves risks, and investment should be made cautiously. This article does not constitute individual investment advice, nor does it take into account specific investment goals, financial situations, or needs of any particular user. Users should consider whether any opinions, views, or conclusions in this article fit their individual circumstances. Investments made according to this article are at your own risk. ```