AI Data Centers Fuel Nuclear Power Boom, but the Nuclear Waste Problem Demands Urgent Solutions

Tech Giants Spark a Nuclear Renaissance

American nuclear energy is experiencing a rare historical moment: across the political spectrum, from left to right, nuclear power enjoys unprecedented broad support. The key force driving this shift is none other than AI tech companies with seemingly bottomless appetites for electricity.

To meet the ever-growing energy demands of massive data centers, tech giants including Microsoft, Google, Amazon, and Meta have all turned their attention to nuclear power. Microsoft has signed power purchase agreements with nuclear fusion startups and even plans to restart the Three Mile Island nuclear plant. Amazon has acquired a nuclear-powered data center campus. Google has partnered with small modular reactor (SMR) companies. Capital and attention have come flooding into a nuclear industry that had been dormant for years.

Yet it is precisely this newfound enthusiasm that forces us to confront a problem that has been shelved for decades: What do we do with nuclear waste?

A Problem Deferred for Half a Century

The nuclear waste disposal challenge is nothing new. Since the birth of commercial nuclear power in the 1950s, the question of how to safely and permanently handle highly radioactive spent fuel has remained nuclear energy's Achilles' heel.

In the United States, more than 88,000 metric tons of spent fuel are currently stored at nuclear plant sites across the country, most of it kept in "temporary" cooling pools or dry storage casks. Some of these so-called "temporary" arrangements have persisted for over forty years. The federal government once planned to build a permanent geological repository at Yucca Mountain in Nevada, but the project was shelved in 2010 amid political controversy, and no alternative plan has materialized since.

In other words, after seventy years of civilian nuclear power operations, the United States still does not have a single permanent nuclear waste disposal facility.

The AI Era's Energy Hunger Heightens the Urgency

Why has this issue become especially pressing today? The answer lies in the AI data centers springing up around the globe.

According to projections from the International Energy Agency (IEA), global data center electricity consumption could double by 2026, reaching approximately 1,000 terawatt-hours — equivalent to Japan's entire national electricity consumption. The power consumed to train a single large language model could supply thousands of average households for a year. As training scales for next-generation models like GPT-5 and Gemini Ultra continue to climb, the power gap will only widen.

Caught between carbon neutrality commitments and surging power demands, nuclear energy has become the "perfect answer" in the eyes of tech companies — zero carbon emissions, stable output, and immunity to weather conditions. But if nuclear capacity expands dramatically to meet AI-driven demand, nuclear waste generation will surge accordingly. When there is nowhere to put even the existing waste, reckless expansion is tantamount to building the road while driving on it.

Technical Solutions Exist — What's Missing Is the Will

From a technical standpoint, nuclear waste disposal is far from unsolvable. The international community has explored multiple viable pathways:

Deep geological disposal is currently the most widely recognized and reliable approach internationally. Finland's Onkalo project is nearing completion and will become the world's first operational permanent spent fuel repository, sealing nuclear waste in granite bedrock approximately 450 meters underground. Sweden has also approved a similar project. These cases prove that the technology is entirely feasible.

Advanced reactor technology also offers new approaches to the waste problem. Fourth-generation nuclear reactors, including some of the small modular reactors (SMRs) favored by tech companies, can theoretically "burn" a portion of the waste produced by conventional reactors, reducing its radioactive lifespan from tens of thousands of years to just a few hundred. While these technologies have not yet been commercialized at scale, they represent a direction well worth investing in.

Spent fuel reprocessing technology has been operating maturely in France for years, using chemical methods to extract still-usable uranium and plutonium from spent fuel, reducing the volume of final waste. However, due to concerns about nuclear proliferation risks, the United States has long avoided this approach.

The core of the problem has never been technology — it has always been political will and social consensus.

Tech Companies Can't Just Buy Power and Ignore the Waste

A question worth asking is: Are tech companies playing a responsible enough role in this nuclear renaissance?

Currently, most tech giants' nuclear strategies focus on the "power purchasing" level — signing long-term power purchase agreements (PPAs) to secure stable, clean electricity for their data centers. But when it comes to the waste inevitably generated by nuclear power operations, these companies have rarely spoken publicly, let alone assumed any corresponding responsibility.

This approach of enjoying the dividends of clean energy while ignoring the costs of waste disposal may work in the short term, but it is unsustainable in the long run. If the tech industry truly wants nuclear power to become the backbone energy source for AI infrastructure, these companies have a responsibility to promote — and even fund — the implementation of waste disposal solutions.

Some argue that tech companies could leverage their strengths in technological innovation and project management to participate in nuclear waste disposal R&D. For example, AI itself could be used to optimize site selection simulations for geological repositories, materials science research for waste containers, and even safety planning for nuclear waste transportation routes. Having AI help solve the problems it creates may be an ironic but pragmatic approach.

The Policy Window Won't Stay Open Forever

The current window of political consensus is precious. In the United States, nuclear energy has earned rare bipartisan support, with multiple bills promoting nuclear power development advancing through Congress. But history shows that such consensus windows are often fleeting — a single nuclear accident or a shift in political winds could send everything back to square one.

Therefore, now is the best time to put nuclear waste disposal on the nuclear renaissance agenda. Specifically, breakthroughs are needed in at least the following areas:

• Restart or replace the Yucca Mountain project, with a clear site selection and timeline for a permanent repository
• Establish transparent mechanisms for using the Nuclear Waste Fund, ensuring that fees paid by electricity consumers (including tech companies) are actually used for waste disposal
• Accelerate R&D and regulatory approval of advanced nuclear fuel cycle technologies, providing technical support for waste reduction
• Promote public communication and community engagement, addressing the "Not In My Backyard" (NIMBY) effect — the greatest social barrier to nuclear waste disposal

Don't Let Enthusiasm Outrun Responsibility

The AI-driven nuclear renaissance is a positive signal, offering a viable path to combat climate change and meet computing power demands. But if we only embrace the benefits of nuclear energy while continuing to ignore the elephant in the room — waste disposal — this renaissance will ultimately sow the seeds of future crises.

Finland spent forty years solving its nuclear waste problem, while the United States has been stalling for seventy. Before pressing the "accelerate" button on nuclear power in the AI era, it's time to get serious about a nuclear waste disposal plan.

The tech industry loves to say "Move fast and break things." But when it comes to nuclear waste, what we need is "Move fast and fix things."