How Much More Power Can Nuclear Plants Squeeze Out? The Largest US Operator's New Plan

Rising electricity demand is making an old question urgent again: how much more power can existing nuclear plants actually deliver? Constellation, the largest nuclear operator in the United States, recently offered an answer—not new reactors, but squeezing more out of existing cores.

Why the Focus Is Now on "Existing" Nuclear Capacity

Tech companies' appetite for power is reshaping the entire industry's calculus. Data centers and AI training clusters have steep, urgent demand curves, while building a new nuclear plant takes a decade or more—far too slow to address immediate needs.

So uprating—increasing the output of existing units without building new reactors—is back in the spotlight. The core appeal: a shorter approval chain, better return on investment, and direct connection to existing grid nodes.

According to industry reports, Constellation recently announced plans to add substantial generating capacity to its nuclear fleet. As the largest nuclear operator in the US, its moves carry signal value—if even the top player is taking uprates seriously, it suggests the economics and regulatory pathways are now viable.

Baseload power refers to electricity sources that can deliver stable output 24/7 on demand, as opposed to intermittent sources like wind and solar that fluctuate with weather. Once connected to the grid, a nuclear plant can typically run continuously for 12 to 24 months before shutting down for refueling, with nearly constant output throughout—making it inherently baseload. For the grid, the value of baseload power isn't just generation volume; it provides an "anchor" for frequency and voltage, allowing dispatchers to schedule other sources around it. AI training and data centers are rigid loads—once a GPU cluster starts, it needs continuous power, and outages or fluctuations can cause training interruptions and data corruption. This load characteristic aligns perfectly with baseload power, which is precisely why tech companies are willing to pay a premium for long-term nuclear power purchase agreements.

How Uprating Actually Works

Nuclear plant power uprates aren't as simple as "overclocking." There are roughly two main pathways:

Measurement Uncertainty Recapture (MUR): Using more precise instruments and computational models to demonstrate that the reactor actually has safety margin remaining, and releasing that conservatively reserved margin. These uprates are typically smaller but easier to approve.

Extended Power Uprate (EPU): More thorough modifications involving replacement of turbines, main pumps, transformers, and other key equipment. Single units can gain 10% or more, but require more comprehensive safety reviews and larger capital expenditure.

The common prerequisite for both paths: safety analyses must be redone, and regulators must be convinced that margin genuinely exists. This is why uprating looks "cost-saving" on paper but isn't easy to execute in practice.

Safety margin is a core concept in nuclear engineering. When designing reactors, engineers reserve margin on key parameters—thermal-hydraulic limits, fuel cladding temperature, critical heat flux—to account for instrument precision limitations, conservative computational models, and material performance variations, ensuring that even off-design conditions won't breach safety boundaries. As measurement technology advances (more precise in-core neutron flux detectors, digital instrumentation) and computational power increases (3D neutron kinetics simulations replacing traditional 1D approximations), engineers can more accurately demonstrate how much headroom actually exists between operating points and safety boundaries. The essence of Measurement Uncertainty Recapture (MUR) is legally releasing this "margin reserved due to uncertainty" to increase power, without changing any physical hardware.

What It Means for AI and Data Centers

If uprates can scale, the most direct beneficiaries are tech companies that have signed long-term power purchase agreements (PPAs). Nuclear provides 24/7 stable, low-carbon baseload power—particularly critical for training clusters, since wind and solar intermittency struggles to match the rigid loads of GPU clusters.

Tech giants have already locked in nuclear supply through restarting old units and signing long-term contracts. If uprates roll out broadly, it's essentially adding another layer of supply on top of existing assets, which would meaningfully help ease localized grid strain.

But it's important to note: the incremental supply from uprates remains a drop in the bucket relative to total demand. It's more of a "stopgap" than a cure—new large-scale reactors and small modular reactors (SMRs) are the long-cycle solution.

Power Purchase Agreements (PPAs) are contract forms in electricity markets where buyers commit to purchasing a certain volume of power at agreed prices over an extended period (typically 10 to 20 years), while sellers gain stable revenue expectations to support investment decisions. For tech companies, signing nuclear PPAs not only locks in electricity prices and hedges against fossil fuel price volatility, but also provides traceable low-carbon electricity certificates to meet their own carbon neutrality commitments. For nuclear operators, PPAs provide long-term cash flow security, making capital-intensive projects like uprate modifications easier to pass internal rate of return reviews. In recent years, multiple tech giants have bound themselves to nuclear operators through PPAs, with some agreements even including financial support for unit restarts or uprates—forming a new cooperation model where "the demand side pays, the supply side delivers."

Real Constraints and Controversies

The first hurdle for uprates is economic calculation. Modification investments, outage windows, and elevated ongoing O&M costs must all be weighed against long-term revenue from added electricity. Electricity price volatility and policy subsidy uncertainty both affect investment decisions.

The second hurdle is regulatory pace. Approval pathways vary significantly across units—MUR projects are typically faster than EPU—but specific timelines remain subject to regulatory staffing and review priorities.

The third hurdle is public communication. Any action involving reactor power adjustments can easily be interpreted as "safety margin being consumed," and operators need to communicate the technical rationale clearly or risk losing ground in public discourse.

There's actually disagreement within the industry about whether uprates are worthwhile: supporters see them as the lowest-cost low-carbon increment, while opponents worry about material aging under extended operation and excessive depletion of safety margins. This debate won't be resolved anytime soon.

Small Modular Reactors (SMRs) are nuclear reactors typically under 300 MW per unit, using factory prefabrication and modular assembly. Compared to traditional large reactors, SMR design goals include: shorter construction timelines (modules can be mass-produced in factories and assembled on-site), lower upfront investment thresholds (modules can be added in phases), and simplified approval through passive safety systems. However, SMRs are still in early commercialization, with most designs not yet completing regulatory licensing and first-of-a-kind construction costs uncertain. Even with smooth progress, SMRs would still need years from licensing to grid connection, unable to fill the urgent power gap of AI data centers in the near term. This explains why uprates are seen as a "stopgap"—they leverage already-licensed, already-grid-connected existing assets, operating on a far faster timescale than any new-build option.

Signals Worth Watching

Constellation's move is just one data point. What's truly worth tracking: whether uprate projects can smoothly obtain regulatory approval, whether tech companies' PPAs will further bind to existing nuclear capacity, and whether SMR commercialization can keep pace. These three variables together determine nuclear power's actual share in the electricity mix over the next decade.

For general readers, the value of understanding uprates lies here: the energy transition isn't only about "building new"—revitalizing existing assets can also shift the supply curve. It's just that the ceiling is lower than one might imagine.

One-line Summary

Constellation plans to squeeze more power from existing nuclear units through uprates to meet AI data center demand.

Paragraph Summary

Facing rising power demand from AI and data centers, Constellation, the largest US nuclear operator, plans to boost output from existing nuclear units through uprates rather than building new reactors. There are two main technical pathways: Measurement Uncertainty Recapture (MUR), which releases conservative margins and is easier to approve, and Extended Power Uprate (EPU), which involves equipment replacement, delivers larger gains, but requires stricter review. As stable baseload power, nuclear aligns well with AI training clusters' rigid loads, and tech companies have already locked in nuclear supply through long-term PPAs. However, uprates face three constraints—economic calculation, regulatory pace, and public communication—and the incremental supply relative to total demand remains limited. It's essentially a short-term stopgap; new builds and SMRs are the long-cycle solution.