The Nuclear Question

I keep hearing a version of this. That the facility will generate its own electricity with nuclear power. That small reactors or microreactors will be installed on the site. That Los Alamos is bringing a reactor to Ypsilanti Township.

It isn’t happening, and we need to stop saying it.

I understand why it has legs. A national laboratory founded on the Manhattan Project is partnering on the facility. Its director has confirmed that some computational work there could relate to nuclear modernization. And across the country, technology companies really are signing nuclear deals to power data centers. Put those three things in a room together and the conclusion assembles itself.

It’s still wrong, and repeating it costs us more than it buys. Everything we say about megawatts, water, ratepayers, taxes and local authority is checkable and documented. The moment somebody stands up at a public meeting and says there’ll be a reactor on site, the project team gets to spend its answer correcting that instead of answering the questions that matter — and everyone watching learns to discount the rest of what we say.

What’s actually planned for power

The project’s own documents describe grid electricity, delivered by DTE through a new substation.

The December 2024 application to the Michigan Strategic Fund itemizes a $1.25 billion budget: $685 million for the Los Alamos building, $100 million for the University building, $417 million for computers, $30 million as a substation capital contribution, $15 million for a network line, and $3 million for land.¹ There’s no generation line item. Nobody building their own power plant contributes $30 million toward someone else’s substation.

On-site generation is limited to backup: generators sized at roughly 10% of load, running on hydrotreated vegetable oil.² They’re combustion engines — the loudest thing on the site, with real emissions implications I’ve written about elsewhere. Diesel-type engines burning a bio-based fuel. Nothing nuclear about them.

Why an on-site reactor couldn’t happen here

No commercial small modular reactor is operating in the United States

Not one. Only a single design — NuScale’s 77 MWe module — has full NRC design certification, granted in January 2023. Every other design you’ll hear named is in pre-licensing or early application. As of 2026 there’s exactly one commercial small modular reactor under construction anywhere in North America, and it’s in Canada.³⁴

Microreactors are earlier still. The demonstration units that exist sit at national laboratory sites and federal test facilities, not on commercial campuses.

Licensing alone takes longer than this project has

Every commercial reactor in the United States needs a license from the Nuclear Regulatory Commission, either a combined license under 10 CFR Part 52 or the two-step process under Part 50. The review covers safety, security, environmental impact and design certification. It’s a multi-year process and it’s the single biggest constraint on any reactor project.⁴⁵

Typical deployment for a small modular reactor, licensing and construction together, runs five to ten years or more.⁶ The earliest realistic dates for a first-of-a-kind unit serving a data center are around 2028 to 2030, and most projects cluster in the early-to-mid 2030s.³

This facility is meant to be running by 2030. The arithmetic doesn’t work.

The money isn’t there

A twelve-module NuScale plant runs about $3 billion.⁵ Even one module is a capital project on the scale of this entire facility.

The published budget has no room for it. The funding sources — university and special purpose vehicle debt, a federal contribution, a state grant — are fully allocated to buildings, computers and utility connections.

It couldn’t be done quietly

This is the part worth holding onto, because it protects you from the next version of this rumor.

NRC licensing is a public federal process. Applications, safety evaluations, environmental reviews and correspondence all go into a public docket, published in the Commission’s ADAMS system.⁸ There are mandatory hearings. Siting brings emergency planning requirements that pull in state and local emergency management, and physical security requirements nobody could hide.

Anyone can check this. Search the NRC’s public records for an application naming this site. There’s nothing there. And if there ever is, it’ll be visible years before a reactor exists, with a formal chance to object.

What the tech company nuclear deals actually are

This is where most of the confusion starts, and it’s simple once you see it.

Almost every headline pairing a technology company with nuclear power is one of two things: a power purchase agreement for electricity from an existing plant delivered over the grid, or an investment in a reactor developer whose product doesn’t exist commercially yet.

Amazon’s deal with Talen Energy is a seventeen-year contract for power from the existing Susquehanna plant in Pennsylvania — and it got restructured from behind-the-meter to front-of-the-meter, meaning the electricity runs through the grid rather than straight from plant to data center.⁶ Google’s agreement with Kairos Power targets reactors that don’t exist yet; the Hermes unit at Oak Ridge is a non-power demonstration reactor that still needs an operating license.³

None of them is a reactor inside a data center fence. The reactor sits where reactors sit, licensed the way reactors get licensed, and the electrons travel.

One thing that could happen, and isn’t the same

Nothing stops this facility from eventually buying nuclear-generated electricity through the grid. Michigan has operating nuclear plants, and the Palisades restart brought another back. A power purchase agreement for nuclear electricity would be an ordinary commercial arrangement.

If that happens, it means nuclear power generated at an existing licensed plant somewhere else in Michigan, delivered over DTE’s wires. It doesn’t mean a reactor in Ypsilanti Township. Those two things will get reported in the same words by people who don’t care about the difference. We care about the difference.

What is true about nuclear and this project

There’s a real nuclear dimension here, and it isn’t about power generation. It’s about what the computers get used for.

The Director of Los Alamos wrote to this community in September 2026 that some computational research at the facility could relate to nuclear modernization — modeling and simulation to assess the safety, security and reliability of the existing stockpile without testing.⁷ That’s stockpile stewardship. Computation rather than manufacturing, exactly as he described it.

The same letter says no plutonium or other nuclear or hazardous materials would be produced, used or stored there, that it wouldn’t be a military installation or a weapons production site, and that plutonium pit production could never happen there because that work is confined to facilities in New Mexico and South Carolina.⁷

I accept that, and I’ve said so before. The University’s FAQ does confirm classified research will happen in the federal building, and that’s a fair thing for people here to have views about. But classified computing isn’t a reactor, and weapons modeling isn’t weapons material.

The test ban treaty claim

Here’s a newer one, and it’s been circulating locally: that the University will be responsible for evading two nuclear test ban treaties, including the one Kennedy negotiated at the height of the Cold War.

That isn’t right either — and it’s close to backwards.

The treaties are real enough. The Limited Test Ban Treaty of 1963, negotiated under Kennedy, bans nuclear test explosions in the atmosphere, in outer space and underwater. The Comprehensive Nuclear-Test-Ban Treaty of 1996 would ban all nuclear explosions; the United States signed it, never ratified it, and it has never entered into force.

Both prohibit nuclear explosions. Running physics codes on a supercomputer isn’t an explosion. There’s nothing in either text to evade.

And the deeper problem with the claim is that it gets the history exactly the wrong way round. Stockpile stewardship exists because of the test bans. When the United States stopped nuclear explosive testing in 1992, it needed some way to certify that aging warheads still worked without setting any off. The answer was to build enough computational physics to substitute for the tests — which is what the Advanced Simulation and Computing program is. Mason’s letter describes it accurately: modeling and simulation to assess safety, security and reliability without nuclear testing.

A facility supporting that work is doing the thing the treaties were meant to produce. Calling it evasion inverts the point.

There is a real argument nearby, and it’s worth knowing the difference. Some arms control advocates argue that simulation capability this good lets the United States maintain and modernize nuclear weapons indefinitely, and that this sits badly with the disarmament obligation in Article VI of the Non-Proliferation Treaty. That’s a genuine, contested policy argument. It’s about the NPT and the spirit of disarmament, not about test ban compliance, and it’s an opinion rather than a fact. Someone could make that case honestly. It just isn’t the same claim.

Same pattern as the reactor rumor, really. Three true things sitting near each other — Los Alamos is a weapons lab, the facility supports weapons work, test ban treaties exist — and the conclusion assembles itself. Say it at a public meeting and the panel gets to spend its answer on a treaty-law correction instead of on why there’s still no water figure.

Why this page exists

We’re asking the University of Michigan to be precise with this community. We can’t ask for that while being sloppy ourselves.

The real objections don’t need embellishing. A 100 megawatt facility is going into a township with no vote over it, on the strength of a constitutional exemption written for something else. Ninety percent of its capacity belongs to a federal laboratory. The water figure has never been published. The cooling design is undecided. The state was offered $400 million for a laboratory at a different site. Nobody has said what this will cost the township in services, or what it’ll pay toward them.

Every one of those is documented and sourced on this site. None of them needs a reactor.

If you hear somebody repeat the reactor claim, send them here.

Sources

Michigan Economic Development Corporation memorandum to the Michigan Strategic Fund Board, December 10, 2024. Board packet (PDF)
University of Michigan Office of the Vice President for Research, LANL FAQ. research.umich.edu
World Wide Technology, “Big Tech’s Nuclear Bet: Key Small Modular Reactors for Cloud Power.” Design certification status, deployment timelines, and the status of the Kairos Hermes demonstration reactor. wwt.com
American Bar Association, Trends, “Pressure to succeed: Small modular (nuclear) reactor approvals on the horizon?” March–April 2026. NRC licensing pathways under 10 CFR Parts 50 and 52, and the status of North American SMR construction. americanbar.org
Davis Graham, “Small Modular Reactors and Data Centers: The Emerging Regulatory Landscape.” NRC licensing as the principal timeline constraint. davisgraham.com · Cost anchor for a twelve-module NuScale plant from industry analysis.
Industry tracking of SMR data center deals, including the Amazon–Talen Susquehanna power purchase agreement and its move from behind-the-meter to front-of-the-meter, and typical five-to-ten-year deployment timelines. Trade source. irecruit.co
Thom Mason, Director, Los Alamos National Laboratory, letter to the Ypsilanti Township community, September 2026. The University Record · my response
U.S. Nuclear Regulatory Commission. Licensing applications, safety evaluations and environmental reviews are public and searchable through the Commission’s ADAMS document system. nrc.gov

Last updated September 2026. If a reactor application naming this site is ever filed with the Nuclear Regulatory Commission, I’ll say so here. If you think anything on this page is wrong, tell me and I’ll fix it and say that I did.