Washington is fast-tracking new reactors to feed the AI power boom

The federal government has moved to accelerate a new generation of nuclear reactors expressly to power the data centers driving artificial intelligence, folding some of the industry’s best-known startups into a funded program with tight deadlines. The latest step, taken in the second half of July 2026, brought two prominent reactor developers into a $200 million effort backed by the Department of Energy and joined by major technology companies. It is the clearest sign yet that Washington now treats nuclear power as core infrastructure for the AI era.

The push reflects a projection that has alarmed grid planners: the electricity demand of AI computing is climbing faster than the power system can comfortably absorb. Data centers draw enormous, steady loads, and their spread has begun to raise electricity prices and stretch interconnection queues in several regions. Nuclear has emerged as a favored answer because it supplies large amounts of around-the-clock power without direct carbon emissions, and the government’s response has been to compress the timelines that normally keep new reactors years from operation.

The $200 million selection

On July 21, the administration selected reactor developers Oklo and X-Energy for a $200 million program meant to speed the construction of reactors aimed at AI data centers, according to reporting from Bloomberg. Technology giants including Microsoft and Nvidia are collaborating on the effort, aligning the companies that consume the most computing power with the firms trying to build dedicated generation for it.

The federal money is spread across more than the two headline names. The initiative includes awards touching roughly two dozen companies and allocates a portion, reported at $60 million over three years, to several national laboratories and academic institutions, among them the University of Texas at Austin. Splitting the funding between commercial developers and research centers is meant to advance both the reactors themselves and the supporting science that regulators and operators will lean on.

The scale the government is planning for

The urgency is anchored in a striking official estimate. The Department of Energy projects that roughly 300 gigawatts of new nuclear power will be needed by 2050, a figure that would require building far faster than the country has managed in decades. That number frames the current programs not as isolated demonstrations but as the opening moves in a much larger construction wave, and it explains why the government has been willing to attach deadlines that would have looked reckless a few years ago.

Meeting a target of that magnitude means proving that new designs can be built quickly and repeatedly rather than one at a time. The department has leaned on its national laboratories to shepherd early reactors through testing, work it has documented through the specialists helping developers reach key milestones, as described by the agency. The strategy treats the first units as templates whose lessons can be copied across a fleet.

How the fast-track has already delivered

The July funding builds on a separate federal effort that had already shown results earlier in the summer. Under a pilot program that set the July 4 holiday as a target, several advanced reactors reached criticality, the point at which a reactor sustains its own chain reaction. The first of them crossed that line under the program in late June, a milestone documented by World Nuclear News, and it was quickly followed by others.

By the deadline, four reactors had reached criticality, a pace that would be unremarkable for software but is extraordinary for nuclear hardware, per a separate account from the same outlet. The demonstrated ability to move designs from groundbreaking to a working chain reaction in under a year is precisely the momentum the newer $200 million program is trying to carry into commercial construction.

What still stands between funding and power

Federal money and expedited testing do not erase the hardest parts of building a reactor. Selling power commercially still requires clearing the Nuclear Regulatory Commission’s design and licensing reviews, a process that has historically stretched for years and that a funding award cannot short-circuit. Oklo, for instance, has assembled a large customer pipeline reported at 14 gigawatts, but the company must still secure full design approval and hit a commercial-startup target set for late 2027, leaving a gap between commitments on paper and electricity delivered to a data center.

There are broader risks in compressing timelines as well. Advanced-reactor projects have a long record of slipping past their original dates and budgets, and pairing first-of-a-kind reactors with the novel, fast-swinging loads of AI clusters adds engineering uncertainty on top of the usual construction challenges. Critics of the accelerated approach warn that shortening review periods could trade away scrutiny, while supporters counter that the reforms streamline testing rather than weaken safety standards. The tension between speed and caution will shape how quickly the program’s paper commitments turn into operating plants.

What the summer’s moves establish is a clear federal posture. A funded program, named developers, participating technology giants, a documented 2050 power target, and a string of reactors that have already reached criticality together mark a deliberate bet that small nuclear can be scaled fast enough to feed the AI build-out. Whether that bet pays off will depend on the unglamorous work still ahead, including regulatory approvals, financing, and construction that no executive order can complete on its own. For now, Washington has signaled that it intends to treat reactors as essential AI infrastructure, and the coming reviews and build-outs will test how far that intent can carry.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview