AI is running out of power. Can nuclear startups like Aalo keep it alive?
Artificial intelligence is expanding rapidly, but its growth is increasingly constrained by a fundamental resource: electricity. Across the United States and other major markets, data centres powering AI systems are consuming rising amounts of energy, placing pressure on electricity grids and reshaping infrastructure planning.
This shift is no longer theoretical. Governments, utilities and technology companies are already responding to increased demand linked to AI workloads. The U.S. Department of Energy has warned that data centre electricity consumption is expected to grow significantly this decade, with AI a key driver.
In response, a new category of energy providers is emerging. Startups such as Aalo Atomics are developing modular nuclear reactors designed to deliver consistent, low-carbon electricity directly to data centres—bypassing some of the limitations of traditional grid infrastructure.
Why AI data centres need more power
The energy requirements of AI differ from earlier generations of digital infrastructure. Traditional data centres typically operated in the range of 5 to 10 megawatts (MW). Today’s AI-focused facilities can exceed 100 MW, with some projects planning significantly higher capacity.
This increase is driven by the computational intensity of training and running large AI models. Unlike conventional cloud workloads, these processes require sustained, high-density power over extended periods.
Two constraints are emerging:
-
Grid limitations: Electricity networks are not designed for sudden, concentrated demand at this scale
-
Sustainability targets: Technology companies are under pressure to reduce carbon emissions
Renewable energy sources such as wind and solar remain central to decarbonisation strategies. However, their variability makes them less suited to applications requiring uninterrupted power. Data centres, by contrast, operate continuously.
This has renewed interest in nuclear energy as a source of stable, low-carbon electricity.
Why nuclear energy is being reconsidered
Nuclear power is re-entering strategic discussions as a potential solution to AI-driven energy demand. Governments and industry players are exploring advanced nuclear technologies, particularly small modular reactors (SMRs).
According to the International Atomic Energy Agency, SMRs are designed to be smaller, factory-built and deployable closer to industrial users. This reduces reliance on grid infrastructure and can improve energy reliability.
Large technology companies are already moving in this direction. Microsoft, Google and Amazon have all taken steps to secure access to nuclear-generated electricity as part of their long-term energy strategies.
However, most SMR projects remain in development, and commercial deployment at scale is not expected before the 2030s.
Aalo Atomics: building nuclear for the AI era
Aalo Atomics represents a new approach to nuclear energy, designed specifically for data centre applications.
Founded in 2023, the Texas-based startup is developing modular reactors intended to be co-located with data centres. Rather than feeding electricity into the grid, Aalo’s systems are designed to generate power at or near the point of use.
Its proposed “Aalo Pod” system is designed to deliver approximately 50 MW of electricity through multiple smaller reactor units. This modular design allows for incremental scaling, aligning energy supply with data centre demand.
The company’s reactors are sodium-cooled, reducing reliance on large water sources, and are intended to be manufactured in factories before on-site assembly. This model aims to improve cost predictability and reduce construction timelines compared to traditional nuclear projects.
Aalo is currently developing its demonstration reactor, Aalo-X, at Idaho National Laboratory as part of a U.S. government-supported programme for advanced nuclear technologies.
Progress towards deployment depends not only on reactor design, but also on securing critical components such as nuclear fuel. In March 2026, Aalo announced a fuel fabrication agreement with Global Nuclear Fuel (GNF), a joint venture led by GE Vernova with Hitachi, Ltd. Under the agreement, GNF plans to supply fabricated fuel for the Aalo-X reactor in early 2026. The company positions Aalo-X as a step towards deploying advanced reactors for data centre applications, though it remains in the demonstration phase.
Rather than a standalone milestone, the agreement reflects a broader transition: from concept to early execution, and from startup ambition to integration with established nuclear supply chains.
A broader shift: modular nuclear startups
Aalo is part of a broader wave of companies rethinking nuclear energy deployment. Firms such as Oklo and Last Energy are also developing modular reactors aimed at industrial users, including data centres.
These companies are addressing longstanding challenges in nuclear energy:
-
High capital costs
-
Long construction timelines
-
Project complexity
Modular designs aim to mitigate these issues through standardisation and factory-based production. For data centre operators, this could enable more flexible and scalable energy sourcing.
However, most of these technologies remain in early development stages.
The barriers to scaling nuclear energy
Despite renewed interest, nuclear energy faces several structural challenges.
Regulation
Nuclear projects require extensive safety approvals, often resulting in long development timelines. While governments are exploring ways to accelerate innovation, regulatory processes remain stringent.
Financing
Even smaller reactors require significant upfront investment. Long payback periods can limit investor appetite.
Supply chains
Scaling modular nuclear depends on developing specialised manufacturing capabilities and ensuring access to nuclear fuel.
These constraints mean nuclear is unlikely to meet short-term energy demand from AI, but could play a larger role over time.
The nuclear waste challenge
Nuclear energy also faces ongoing scrutiny over radioactive waste.
Spent nuclear fuel remains hazardous for long periods and requires secure, long-term storage. While the volume of waste is relatively small compared to other industrial waste streams, its management presents both technical and political challenges.
According to the International Atomic Energy Agency, most countries currently store spent fuel on-site or in interim facilities. Long-term solutions, such as deep geological repositories, are widely considered the safest approach.
Progress varies by country. Finland is among the most advanced, with its Onkalo repository expected to become the first operational deep geological storage facility. In contrast, the United States has faced prolonged delays in establishing permanent disposal infrastructure.
For critics, this raises questions about expanding nuclear capacity without fully operational long-term waste solutions. Increasing the number of reactors inevitably increases the volume of waste requiring management.
Developers of advanced reactors argue that newer designs may reduce waste volumes and improve fuel efficiency. However, these approaches are not yet widely deployed.
The issue extends beyond engineering. It involves public trust, regulatory frameworks and long-term accountability.
Can nuclear keep pace with AI demand?
The key challenge is timing.
AI-driven electricity demand is increasing rapidly, while most advanced nuclear technologies are still under development. In the near term, demand is likely to be met through a mix of grid expansion, renewable energy and, in some cases, fossil fuels.
Over the longer term, nuclear energy offers a stable, low-carbon option aligned with the needs of data centres.
For companies like Aalo, the challenge lies in execution—moving from demonstration to deployment at scale.
A new infrastructure reality
The convergence of AI and energy marks a shift in how technological progress is understood.
Digital innovation is no longer decoupled from physical infrastructure. Energy availability is becoming a defining factor in where and how data centres are built.
Nuclear startups are positioning themselves within this shift, offering a model that combines reliability with low emissions. Their success will depend on technology, regulation, financing and public acceptance.
Powering the future of AI
AI is not running out of potential, but it is encountering physical limits. Electricity is becoming central to the future of digital infrastructure.
Startups such as Aalo Atomics are attempting to address this challenge by rethinking nuclear energy for a new generation of applications. Their progress reflects a broader recognition that scaling AI will require advances in infrastructure as well as software.
Whether nuclear energy can meet this demand remains uncertain. What is clear is that the future of AI will depend not only on algorithms, but on the systems that power them.
Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.