For years, data center power strategy had a familiar shape. The utility provided the primary source. Generators provided backup. UPS systems bridged the gap. Batteries covered short-duration events. The model was not simple, but the hierarchy was clear: grid first, backup second.
AI is changing that hierarchy.
The largest AI facilities are not asking for ordinary commercial service. They are asking for industrial-scale, always-on electricity at a pace many utilities were not built to serve. A facility running high-density AI training or inference cannot treat power as a background input. It has to treat power as the product behind the product. No electricity, no compute. No compute, no revenue.
That is why the industry is now talking seriously about nuclear power, battery energy storage, on-site generation, microgrids, and hybrid systems. Not because every data center wants to become a power plant, but because the old utility-dependent model is too slow and too fragile for the next wave of AI infrastructure.
iRecruit’s 2026 nuclear data center analysis frames the scale of the issue directly: AI-focused facilities now require about 80 MW of power, compared with roughly 32 MW for standard data centers, while U.S. data center energy demand is expected to rise from 17 GW in 2022 to 35 GW by 2030. The same analysis points to grid interconnection delays that can stretch up to a decade, which is why companies like Amazon, Google, and Microsoft are now pursuing nuclear power as part of their long-term data center strategy.
That is the real story. Nuclear is not back because it is fashionable. Nuclear is back because AI requires baseload power, and baseload power is getting harder to secure.
AI does not want intermittent power
A traditional enterprise workload may tolerate some flexibility. It has peaks and valleys. It can be shifted, deferred, or throttled in certain situations. AI training and large-scale inference are different. They need enormous amounts of power, and they need that power to be stable, clean, predictable, and available.
This is where renewables face a practical challenge. Solar and wind are valuable, and they will remain part of the power mix. But they are variable by nature. A massive AI data center cannot operate only when the sun is shining or the wind is cooperating. Batteries can bridge gaps, smooth peaks, and support resilience, but batteries are not infinite energy sources. They are storage systems. They have to be charged by something.
That is why operators are moving toward hybrid power stacks. A future AI campus may include grid service, nuclear PPAs, battery energy storage, natural gas generation, renewable procurement, microgrid controls, and traditional backup generators. The point is not to pick one source and declare victory. The point is to build a power stack that can deliver continuous capacity while managing cost, carbon, reliability, and regulatory risk.
CoreSite’s 2026 outlook says data centers are shifting from passive energy consumers to grid stakeholders, co-investing in infrastructure upgrades, enabling load flexibility, and deploying on-site power generation and storage to improve reliability and manage costs. The same outlook notes that data center operators are expected to increase behind-the-meter power arrangements, with technologies ranging from fuel cells to small nuclear reactors.
That phrase, “grid stakeholders,” is important. The data center is no longer just a customer at the end of the line. In many cases, it is becoming part of the power system itself.
Nuclear is being pulled back into the data center conversation
The strongest signal that nuclear is moving from theory to strategy is not a white paper. It is the actual deal activity.
Talen Energy announced in June 2025 that it had expanded its nuclear energy relationship with Amazon, entering a power purchase agreement to supply AWS operations with carbon-free electricity from the Susquehanna nuclear power plant. At full contract quantity, Talen says it will provide Amazon with 1,920 MW of carbon-free nuclear power through 2042, with power delivery ramping over time and full volume expected no later than 2032.
Google has moved in the same direction with Kairos Power. Kairos announced that it will develop, construct, and operate a series of advanced reactor plants and sell energy, ancillary services, and environmental attributes to Google under PPAs. The first deployment is targeted by 2030, supporting Google’s 24/7 carbon-free energy and net-zero goals.
Microsoft is tied to one of the most visible nuclear comeback stories in the country. Constellation announced a 20-year power purchase agreement with Microsoft to support the launch of the Crane Clean Energy Center and restart Three Mile Island Unit 1, which had been shut down for economic reasons.
These are not small pilots. They are the kind of moves companies make when they believe power availability will determine who wins the AI infrastructure race.
Small modular reactors are part of this conversation because they promise a different deployment model than large traditional nuclear plants. iRecruit describes SMRs as compact reactors that can produce up to 300 MW of power, scale in modules, require less land, and offer 24/7 baseload output. It also compares SMRs with solar and wind by noting SMRs can provide a 95%+ capacity factor, while solar and wind are weather-dependent.
That does not mean SMRs are an easy button. They are not.
The Department of Energy is clear that next-generation reactors will take time. DOE says licensing, demonstrating, and deploying additional reactors will take years, and that widespread commercial reactors are likely to arrive in the 2030s. DOE also notes that first-of-a-kind deployments are expensive, with most nuclear costs concentrated in upfront capital construction, even though modular construction and factory fabrication may reduce cost and schedule risk over time.
That is the honest picture. Nuclear is strategically important, but it is not a short-term fix for every project. Existing nuclear can support nearer-term deals. New advanced reactors may matter more in the next decade. SMRs will have to clear licensing, cost, workforce, fuel, public acceptance, metering, and interconnection hurdles.
Still, the direction is unmistakable. AI has made nuclear relevant again because AI needs reliable, large-scale, low-carbon power that does not depend on the weather.
Batteries are not just backup anymore
Batteries used to be discussed primarily as UPS infrastructure. They bridged the gap between utility failure and generator start. They were part of the backup chain, not the operating strategy.
That is changing.
Schneider Electric describes the future model as “grid interdependence,” where data centers and utilities coordinate more actively. For that model to work, Schneider argues that data centers need to generate or store some energy on-site, allowing them to continue operating even when disconnected from the grid. Schneider also notes that microgrids with battery energy storage systems are increasingly being deployed at some facilities and can allow operators to disconnect from the grid during peak hours to optimize energy consumption.
That is a different role for batteries.
In a hybrid power stack, batteries can do several jobs. They can provide short-duration backup. They can smooth transitions between grid, generator, and on-site sources. They can support peak shaving. They can help the facility ride through grid disturbances. They can make renewable generation more usable. They can support demand response or grid-interactive strategies. They can reduce the need to start generators for every short event.
Batteries do not replace baseload power. They make the overall system more flexible.
That flexibility matters because AI facilities are operating in a world where power supply is no longer a single-source assumption. A site may have grid service, on-site generation, batteries, solar, a nuclear PPA, traditional backup generators, and controls that decide which source should serve the load under different operating conditions.
At that point, power architecture becomes less like a building utility system and more like an engineered energy platform.
Hybrid systems are where the real design complexity lives
The future power stack will not be simple.
A nuclear-backed data center still needs switchgear, transformers, protection, distribution, monitoring, backup, and contingency planning. A battery-backed microgrid still needs fault studies, controls, transfer logic, metering, and maintenance procedures. A natural gas bridge solution still has permitting, emissions, fuel, noise, run-hour, and resiliency considerations. A grid-parallel facility still has to coordinate with the utility and understand how fault current, power quality, and protection change across operating modes.
The hardest part is not naming the sources. The hardest part is making them work together safely.
A hybrid AI data center may need to operate in several modes. Normal grid-connected mode. Islanded mode. Generator-supported mode. Battery-supported mode. Nuclear-supplied PPA mode. Curtailment mode. Emergency restoration mode. Maintenance bypass mode. Expansion phase mode.
Each mode can change the electrical behavior of the system.
Available fault current may be different when the facility is on the utility than when it is on local generation. Protective devices that coordinate in one operating state may not coordinate the same way in another. Ground fault behavior can change. Arc flash exposure may change. Controls become critical. Metering becomes political and commercial, not just technical. Documentation must be precise because the system is no longer a one-source, one-direction power path.
DOE highlights one version of this issue in the nuclear context. It notes that metering can become a hurdle, pointing to Amazon and Talen’s earlier behind-the-meter arrangement at Susquehanna and concerns from federal regulators that some arrangements could allow data centers to benefit from transmission systems without paying for them, potentially raising costs for utility customers.
That is the kind of complexity owners have to expect. Power architecture is becoming technical, commercial, regulatory, and political at the same time.
This is also why the phrase “always-on power” needs to be understood carefully. Always-on does not simply mean having a big enough generator or signing a large enough PPA. It means the full electrical system can support the load through normal operation, source changes, grid events, equipment failures, maintenance windows, and future expansion without creating hidden failure points.
Always-on power is not a source.
It is a system.
Where Moonshot fits in the future power stack
This is where Moonshot’s positioning can move up a level.
Moonshot is not a nuclear developer. It is not a battery manufacturer. It is not a utility. It does not need to be.
Moonshot operates in the layer that allows all of those sources to become usable inside a facility: switchboards, transformers, PDUs, modular data centers, controls, EPC support, and electrical infrastructure.
That layer becomes more important as power systems become more hybrid.
Moonshot’s UL891 switchboard page says its engineering team designs solutions around supply chain constraints and maximum flexibility, can use major breaker manufacturers, and evaluates site design during the sales process to provide code-compliant, cost-effective, quick solutions.
Moonshot’s transformer offering is also relevant because the company is an authorized distributor with direct partnerships with leading transformer manufacturers worldwide, offering pole mount, pad mount, substation transformers, and mobile substations for industries including utilities, data centers, and solar power.
Moonshot’s broader EPC experience matters as well. The company states that it has data center EPC experience, including an 8 MW facility in West Texas and a 5 MW facility in South Carolina, with more than 150 MW of collective data center construction experience across the team.
Those capabilities fit the future power stack because the customer problem is no longer just “build me a switchboard.” The customer problem is, “Help me build power infrastructure that can accept multiple sources, survive multiple operating modes, remain code-compliant, and scale with AI demand.”
That is a more valuable conversation.
Moonshot can help customers think through the practical implications of hybrid power before the design hardens. How does the switchboard need to be rated if the source configuration changes? How should transformer strategy be planned if future phases add more generation or storage? Where do PDUs, metering, and controls need to sit if the facility participates in grid-interactive operation? How should modular data center infrastructure be designed if the site later adds a battery system, on-site generation, or a different utility service path?
These are not distant questions. They affect projects being designed today.
A facility built in 2026 may still be operating when advanced nuclear, larger-scale batteries, 800 VDC distribution, or new hybrid microgrid architectures are far more common. If the electrical backbone is designed only around today’s power source, it may become the constraint tomorrow.
The strategic point: power optionality is becoming a competitive advantage
The future data center will not rely on one answer.
Some sites will use existing nuclear through PPAs. Some will pursue SMRs or advanced reactors over the long term. Some will use natural gas as a bridge. Some will pair solar with batteries. Some will develop microgrids. Some will use grid power while investing in demand response and on-site storage. Most serious AI campuses will use a combination of these strategies over time.
The winning operators will not be the ones that make the boldest announcement. They will be the ones that preserve optionality.
Optionality means the site can adapt as technologies mature, regulations shift, tenants change, and utility conditions evolve. It means the facility is not trapped by early electrical decisions. It means switchboards, transformers, PDUs, controls, metering, and modular infrastructure are designed with future sources and future operating modes in mind.
That is how Moonshot should frame its role.
Not just as a manufacturer of electrical gear, but as an infrastructure partner that helps customers build power systems capable of adapting to the next decade of AI growth.
The real takeaway
AI infrastructure is forcing the data center industry to build a new power stack.
Grid power alone may not be fast enough. Renewables alone may not be firm enough. Batteries alone are not energy sources. Generators alone are not a sustainability strategy. Nuclear is powerful, but new reactors take time, capital, permitting, and regulatory alignment.
The answer is hybrid.
The future power stack will combine grid service, nuclear power, batteries, on-site generation, renewables, microgrid controls, and traditional redundancy into systems designed for continuous operation. That shift creates a new challenge: the electrical infrastructure has to be flexible enough to support sources and operating modes that may change over the life of the facility.
For Moonshot, this is the opportunity to sit at the top of the infrastructure conversation. The company’s switchboards, transformers, PDUs, modular data centers, and EPC experience are not just project components. They are the connective tissue between the power source and the compute load.
If you are planning AI infrastructure, the question is no longer only where the power comes from.
The better question is whether your electrical architecture can support the power stack you will need five years from now.
https://moonshotus.com/request-form/
Because always-on power is not created by one source.
It is created by a system built to survive change.
Sources:
https://www.irecruit.co/insights/smr-nuclear-powered-data-center-developments
https://www.energy.gov/ne/articles/advantages-and-challenges-nuclear-powered-data-centers
https://ir.talenenergy.com/news-releases/news-release-details/talen-energy-expands-nuclear-energy-relationship-amazon/
https://www.kairospower.com/updates/google-and-kairos-power-partner-to-deploy-500-mw-of-clean-electricity-generation
https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html
https://blog.se.com/datacenter/2026/01/13/rethinking-power-grid-interdependence-future-of-data-centers/
https://www.coresite.com/blog/data-center-outlook-2026-power-and-cooling-challenges-and-solutions-are-top-of-mind
https://moonshotus.com/ul-891-switchboards/
https://moonshotus.com/transformers/
https://moonshotus.com/construction/

