What They Said, and What the Record Shows

This page puts the project’s public statements next to the documents. Quotes on one side, the record on the other, and a source number on everything.

I’m not claiming anyone lied. Most of these statements are technically accurate and still leave you with the wrong impression, which is a different problem and a more common one. Where the project is right, I say so. Read both columns and make up your own mind.

For the full chronology — every document from March 2023 forward, in order — see the timeline. This page is about the claims.

If you think anything here is wrong, tell me. I’ll fix it and say that I did.

1. Whose facility is this?

This is a specialized research hub, not a commercial data center … dedicated to advanced scientific research and high-performance computing in areas such as medicine, energy, climate science, engineering and national security.

Project FAQ2

What the record shows

The application to the State describes 100 megawatts of capacity for Los Alamos National Laboratory and a separate dedicated 10MW academic computing center for the University’s non-classified open research initiatives.1

By the University’s own numbers that’s roughly 91% federal, 9% University. The budget splits the same way: $685 million for the Los Alamos building against $100 million for the University’s. Los Alamos would pay $285 million in net construction costs plus operating costs estimated at $78 million a year, and the parties anticipated long-term lease or use agreements for its occupancy.1

The FAQ backs that up on floor area: a Federal Research Facility of 240,000 square feet supporting national research and security challenges, including classified research, next to an Academic Research Facility of about 50,000 square feet.2 That’s 83% federal by floorspace, and it settles a question people here had been asking: classified work will happen on this site.

At the September 9 town hall a resident asked directly about the split. After the moderator turned to project leader Steven Ceccio for confirmation, the panel acknowledged roughly 90/10, with the large majority to Los Alamos.3

Director Mason’s letter describes the project as expanding Los Alamos’s presence in Michigan — not as supporting the University’s research — and says some computational research there could relate to nuclear modernization, using modeling and simulation to assess the safety and reliability of the existing stockpile without testing.4 That’s computation rather than manufacturing, exactly as he put it. It’s also work supporting the nuclear weapons program, and people living nearby are entitled to know it.

None of that is new, incidentally. The 2023 non-disclosure agreement between the University and Los Alamos states its purpose as a computing facility and office building in support of the Advanced Simulation & Computing Program and Weapons Physics Directorate.10

And the signed agreement points somewhere else. Memorandum of Understanding #204, signed November 25–26, 2024 — two weeks before the state grant was approved — says the Center for HPC is “currently planned to be a central feature of the Innovation Campus,” and that the parties intend to work toward approvals, design, funding and “eventually establishing a Center for HPC near UM’s Ann Arbor campus.”6

To be precise about this: the Innovation Campus is a separate University initiative in Ann Arbor, and the MOU lists supporting it as its own area of cooperation. But the computing center isn’t placed somewhere else in that document. It’s described as a central feature of that campus, near Ann Arbor, with Los Alamos possibly taking office space within it — in the same paragraph that names the 200 LANL jobs and the intent to seek Michigan state assistance.6

Ypsilanti Township appears nowhere in it.

An MOU is a framework, not a site plan, and it says outright that it authorizes no work. Plans change. But the agreement that defines this partnership puts it beside the Ann Arbor campus, and the project is now fourteen miles away, on a river, in a township that has formally objected.

The question this raises

Article VIII, Section 5 of the Michigan Constitution gives the Regents supervision over the University’s own institution, and that autonomy is the only reason this project is exempt from local zoning. If 91% of the capacity is a federal laboratory operating under a long-term lease and running nuclear stockpile modeling, what’s the basis for extending that exemption to the whole facility? No Michigan court has ever been asked.

2. On the size of it

This facility will use less than 1/10th the energy of commercial data centers, starting at 50 megawatts with a maximum planned capacity of 100 megawatts. The physical footprint and energy usage of this facility will be comparable to a modest industrial facility rather than a commercial data center.

Project FAQ2

What the record shows

The two descriptions don’t obviously reconcile. The state filing gives 100 MW for Los Alamos plus a dedicated 10 MW academic center — 110 in total.1 The FAQ gives a maximum planned capacity of 100 MW without distinguishing them.2 Either the total came down, or the public figure covers only the Los Alamos side. That should be easy to clear up.

The industry threshold for calling something hyperscale — the largest category there is — is at least 5,000 servers, 10,000 square feet, and 40 megawatts.19 This is roughly two and a half times that.

The “1/10th” comparison never says what it’s comparing to. For 100 MW to be a tenth of its comparison class, that class would have to average about a gigawatt — which fits only the largest AI training campuses being built anywhere, not commercial data centers generally.

On “modest industrial facility”: at 110 MW running flat out, that’s roughly 2,640 megawatt-hours a day. Set against the average American household’s 10,791 kilowatt-hours a year,20 it’s comparable to the daily electricity use of about 89,000 homes. Ypsilanti Township has fewer than a quarter that many. The filing’s own estimate is $78 million a year just in electricity, water and maintenance.1

The substation description changed. The filing says DTE anticipated building a new substation with two-thirds of its capacity serving the project and a third for community benefit, against a $30 million capital contribution.1 The FAQ now describes a dedicated substation that will exclusively power the facility.2 Exclusive and two-thirds aren’t the same deal, and the third earmarked for community benefit — which helped support a $100 million grant application — has quietly dropped out of the public description.

One more detail worth reading closely. The FAQ says the University secured large-scale solar agreements to account for the center’s academic energy needs.2 On the December 2024 figures the academic share is 10 of 110 megawatts. So the clean energy commitment, as written, covers the small part of the facility and not the federal load that makes up most of it.

The question this raises

Is the public 100 MW the whole facility or just the Los Alamos portion? What’s the substation’s design rating, and what total interconnection capacity has been requested from DTE? If two-thirds of a substation serves this one site, the remaining third caps every other future connection around it. And with 144 acres now in hand, what’s the ultimate build-out capacity of the parcel?

3. On water

What the record shows

Start with why Michigan got picked. The University’s own application lists the state’s geography and long-term climate resilience, room for expansion, a reliable grid — and ample access to fresh water for equipment cooling operations.5

Fresh water for cooling was one of the reasons Michigan won this project. That’s hard to square with a public position that the facility won’t draw from the river or use groundwater and might use closed-loop cooling, which uses no water at all. Either the water mattered to the siting decision or it didn’t.

The filing says water-cooling technology is being considered, lists non-potable or reclaimed water and heat re-use among the concepts, and records that YCUA had confirmed the location already has water and wastewater capacity to immediately support the proposed initial demand.1

That’s changed. On April 22, 2026, following a Township resolution, YCUA voted a twelve-month moratorium on delivering, committing, reserving or approving water and sewer service to data centers and AI computing facilities.15 YCUA is a public authority run by the Township and the City — and zoning exemption doesn’t compel a municipal utility to extend service.

The current FAQ says the facility won’t draw from or discharge into the Huron River and won’t use local groundwater, that any municipal water comes through YCUA, and that the University is reviewing designs to consider closed-loop cooling and reclaimed wastewater. It also confirms direct-to-chip cooling, which is a liquid technology.2

So closed-loop is under consideration, not committed. And the municipal supply the FAQ leans on runs through the authority that just stopped serving facilities like this one.

At the town hall the panel confirmed the cooling design is still open, described closed-loop as one option that uses more electricity but no water, and called it a significant decision still to be made.3 In three hours, nobody gave a figure in gallons.

Worth holding onto: a facility can truthfully say it doesn’t withdraw from the Huron River while consuming very large volumes of municipal water. Those are different claims. There’s more on how all this works on our water page.

The question this raises

What’s the projected consumptive water use in gallons per day, at initial build and at full capacity? Will closed-loop be committed to, or only considered? What exactly did YCUA confirm in 2024, and on what terms?

4. On emissions

As a fully electric facility, the operation will not produce smoke, odors or harmful emissions. For essential backup power during emergencies, we have limited the system to just 10% of the facility’s load. These backup generators will utilize Hydro-treated Vegetable Oil (HVO), a renewable, bio-based fuel, which is significantly cleaner than traditional diesel and reduces greenhouse gas emissions by up to 95%.

Project FAQ2

What the record shows

Where they’re right: limiting backup to roughly 10% of load is a legitimate engineering choice for high-performance computing. An HPC site can checkpoint running calculations and shut the compute floor down in order, keeping only storage, networking and critical cooling alive. That’s a real technical distinction, not marketing.

On “no harmful emissions”: that first sentence is about the building, which has no gas-fired boilers or chillers. It sits immediately before a sentence describing combustion engines. Ten percent of 110 MW is about 11 MW of on-site generation.

And emergency generators don’t only run in emergencies. EPA rules let them run without time limit during a genuine emergency, plus up to 100 hours a year for maintenance and readiness testing, of which up to 50 can cover certain non-emergency purposes.2122 Monthly load-bank testing is standard practice.

On the 95% figure: that’s a lifecycle greenhouse gas number, not a local air quality one. Caterpillar puts HVO carbon intensity at 15–78% of fossil diesel depending on feedstock, using the Department of Energy’s GREET model.23 Cummins puts well-to-wheels emissions roughly 70% below diesel.24 The reduction comes from the carbon being biogenic, not from cleaner exhaust.

On what neighbors actually breathe, the picture narrows fast. Particulate matter does drop — Cummins measured 30–60% lower PM on HVO than ultra-low-sulfur diesel in generator testing.25 Nitrogen oxides mostly don’t. Cummins says criteria pollutants like NOx are comparable to diesel,24 and Caterpillar’s own testing found NOx equivalent to diesel at loads above 35% and below 15% of rated power.23 NOx is an ozone precursor. More on this on our noise page, which covers the same equipment.

The question this raises

How many generator sets, at what rating, for what total emergency generating capacity? How much fuel stored on site, in what containment, with what spill response plan? What’s the testing schedule and the EGLE air permit basis? Is the HVO commitment contractual, for what term, enforceable by whom — given the same engines run on ordinary diesel with no modification?

There’s also a commitment implied here that hasn’t been offered. A facility that can shed most of its load on command is an interruptible load, which burdens the grid very differently from a firm one that has to run through peak demand. If the 10% figure describes a real capability, committing to curtailment during grid stress would cost them nothing.

5. On jobs

The project is expected to create approximately 300 union construction jobs, more than 200 high-paying permanent research positions, and 30–50 technical support roles.

Project FAQ2

What the record shows

Where they’re right: the construction jobs are real. A build this size draws heavily on regional trades, they’re good jobs, and the people who take them will have earned them. They’re also temporary, against a facility meant to run for decades.

The 200 is described differently everywhere you look. The state filing calls it a minimum of 200 direct permanent jobs over 10 years, then Los Alamos estimating up to 200 jobs in Michigan over the next 10 years, then plans to accommodate up to 200 National Lab employees.1 “Minimum of 200” and “up to 200” are not the same commitment, and they’re in the same document.

The signed MOU settles what they are: subject to the availability of NNSA funds, approximately 200 LANL jobs created at UM over ten years.6 Los Alamos jobs, at the University, across a decade, contingent on federal money that hasn’t been appropriated.

And it used to be 800. MEDC’s briefing memo of October 16, 2024 told the board to expect over 800 direct permanent jobs over ten years, with the University then asking for $135 million.8 By the fiscal modelling of November 8, and the final memo of December 10, it was 200.9 Three weeks. Nothing released explains it.

The bigger numbers you’ve heard come from that same modelling, and it states its assumption outright: the industry modelled is NAICS 541511, Custom Computer Programming Services, giving a job multiplier of 3.25 and about 650 total jobs in the final five years.9 That’s the classification for a software development firm — a business whose product is people writing code. A computing facility is infrastructure with a few dozen people in it. Model one as the other and the answer comes out wrong in a predictable direction.

Researchers don’t work inside data centers anyway. They submit jobs to a scheduler from wherever they happen to be. A position enabled by a computing facility isn’t a position located at one.

At the town hall, asked how many permanent jobs would go to Township residents, the answer opened with the panel not being sure they could legally say any would.3 I think that’s an honest answer — hiring commitments to one municipality do raise real legal questions. It’s also, plainly, the answer.

For what it’s worth, the University’s own Ford School reached the same conclusion about data centers generally: infrastructure projects rather than job-creating businesses, requiring relatively few employees once built.26 Independent analysis puts operations at 0.15 to 0.35 full-time equivalents per megawatt, with highly automated campuses above 100 MW running on 20 to 30 permanent staff.27 Brookings calls data centers among the least labor-intensive structures in the economy.28

The project’s own 30 to 50 technical support roles sits right in that range.

The question this raises

Of the 200: how many have a duty station in Ypsilanti Township? How many on University payroll versus Los Alamos payroll? How many are new rather than reallocated? And is it a minimum or a ceiling — the state filing says both.

6. On talking to us

Once a site is selected, U-M will work with local leaders and residents to identify opportunities for the project to provide meaningful long-term value to the Ypsilanti Township community.

Project FAQ2

What the record shows

The site has been selected. Textile Road was announced as the preferred location in August 2026.13 By its own terms, that commitment is now due.

The commitment itself has no figure, no timeline, no obligation and nothing measurable in it. It’s a promise to have a conversation.

Here’s the conversation so far.

It starts in November 2024, with a Zoom meeting. Supervisor Stumbo described it at the September town hall: her, State Representative Jimmie Wilson, and Chris Kolb from the University. What they were told, in her account, was twenty acres, two office buildings, and two hundred jobs on site at $200,000 a year.3

The University had closed on those twenty acres on November 5.

That meeting doesn’t appear in the University’s own published timeline, which begins its engagement story with a campus open house in September 2025 and the first Ypsilanti Township open house in January 2026.

Then the University bought the 124-acre parcel and told the Township Supervisor by text message after the sale closed; the Township attorney recorded the notification and objected in writing.1615

Every household got a postcard inviting them to the September 9 town hall. Registration had been full, with a waitlist, for weeks before those postcards landed.3 At the meeting the panel confirmed it had no authority to change, alter or stop the project — that sits with the Regents. One regent attended. Los Alamos, which by the University’s own figures accounts for roughly 90% of the facility, sent nobody.3

The Strategic Site Readiness Program also requires applicants to document local support. Asked how much of a priority the project was for local government or the economic development organization, the filing cited Ann Arbor SPARK — a regional body, not Ypsilanti Township. Asked about the extent of regional and local support, it offered that support is evidenced by the preferred site being in Ypsilanti Township.1

Choosing a place isn’t evidence that the place agrees. There’s no letter from Ypsilanti Township anywhere in that packet.

The same filing cites the Township’s own Future Land Use Plan to show the site sits in an Innovation and Employment District intended for “technology, office, craft manufacturing or light industrial uses.”1 So the Township’s land use plan was useful enough to quote when applying for a state grant.

A second town hall was held on October 1, with the same panel as September 9. The Regents have also voted to appoint C. Cybele Raver as the University’s next president.14 New leadership is a chance to settle an inherited fight on better terms, and I’d welcome that.

The question this raises

The site is selected. Will the University enter a binding community benefits agreement, negotiated in public and enforceable in court, including an annual payment in lieu of taxes? Is the promise to keep most of the 144 acres as natural space legally binding, and by what instrument — a conservation easement, a deed restriction, or a press release? And if the Township’s Future Land Use Plan was worth citing to the State, is it worth complying with?

7. On what it costs to run

What the record shows

Two versions of the same application give very different operating costs. The December 2024 board packet says $78 million a year for electricity, water, building maintenance and operations.1 An earlier version in the MEDC release puts it at $115 million for the 100MW facility and $35 million for the 10MW — $150 million all in.5

Scopes change between drafts and there may be a perfectly ordinary explanation. But electricity is a big share of any number like that, and a community trying to understand this facility’s appetite now has two figures from the same institution that differ by nearly half.

The same release also records that a road will need widening up to the main entrance, with the specifics redacted.5 Nobody has said what that costs or who pays.

The question this raises

Which operating cost figure is current, and what’s the projected annual electricity consumption in megawatt-hours? Which road gets widened, at what cost, paid by whom?

The pattern

Read these together and they share a shape. Each one is defensible on its own terms, and each one leaves out the number that would let you judge it. The comparison class for “1/10th” goes unnamed. The capacity figure seems to describe part of the facility. The emissions claim describes the building rather than the site. The job figure went to the State as national laboratory positions spread across Michigan over a decade. The community commitment contains no commitment — and the site it was contingent on has now been picked.

None of it is hidden. Most of it sits in public state and township records the project’s own materials never mention. Every question on this page has a specific factual answer the University already has in hand.

Publishing them would settle most of this.

Sources

  1. MEDC memorandum to the Michigan Strategic Fund Board, December 10, 2024, with MSF Resolution 2024-176 and the Summary of Terms. Board packet (PDF)
  2. University of Michigan Office of the Vice President for Research, LANL FAQ. Retrieved August 28, 2026. research.umich.edu
  3. Public recording of the September 9, 2026 town hall. youtube.com · our account, with timestamps
  4. Thom Mason, Director, Los Alamos National Laboratory, letter to the Ypsilanti Township community, September 2026. The University Record · my response
  5. MEDC FOIA release, request 1068 (73 pages): the LANL/Regents Memorandum of Understanding, the University’s SSRP application materials, an economic impact analysis, and related correspondence. MEDC FOIA archive
  6. Memorandum of Understanding #204, Los Alamos and the Regents, signed November 25–26, 2024 (PDF)
  7. Letter from MEDC CEO Quentin L. Messer, Jr. to Dr. Santa Ono, October 27, 2023 (PDF) · our analysis
  8. Project L briefing memo to the MSF Board, October 2024 (PDF) · draft term sheet (PDF)
  9. National Research Lab fiscal impact estimates, November 8, 2024 (PDF)
  10. Non-disclosure agreement RSNDA-23-0076, March 2023 (PDF) · first amendment, July 2023 (PDF)
  11. Nexus Gas Transmission confidentiality agreement, December 2025 (PDF)
  12. MEDC confidentiality agreement, May 24, 2024 (PDF) — see Exhibit A
  13. The University Record, “University of Michigan selects Textile Road as preferred site,” August 2026. record.umich.edu
  14. The University Record, September 2026 — the additional October 1 town hall, and the Regents’ appointment of C. Cybele Raver. record.umich.edu
  15. Charter Township of Ypsilanti, Data Center Information archive, including Resolutions 2026-05 and 2026-08 and correspondence from McLain & Winters. ypsitownship.org
  16. The Michigan Daily, on the closing of the Textile Road purchase. michigandaily.com
  17. Planet Detroit, on the land purchase and the notification by text message. planetdetroit.org
  18. WDET, “Ypsilanti Township fight over U-M computing center intensifies with moratorium,” September 11, 2026. wdet.org
  19. Data Center Dynamics, “What is a hyperscale data center?” datacenterdynamics.com
  20. U.S. Energy Information Administration, “How much electricity does an American home use?” eia.gov
  21. U.S. Environmental Protection Agency, “Understanding the Stationary Engines Rules.” epa.gov
  22. 40 CFR Part 63, Subpart ZZZZ. ecfr.gov
  23. Caterpillar, renewable liquid fuels FAQ for 3500 and C175 series generator sets. cat.com
  24. Cummins, “Comparing emission reductions across alternative fuels.” cummins.com
  25. Cummins, “Generator Set Performance on HVO Fuel: QSK95 Test Report Summary.” cummins.com
  26. Terry Nguyen and Ben Green, What Happens When Data Centers Come to Town?, Ford School of Public Policy, University of Michigan, with the Michigan Environmental Justice Coalition, July 2025. PDF
  27. Hamm Institute for American Energy, Data Center Employment Forecast Analysis, November 2025. PDF
  28. Bahar and Wright, “New evidence on data center employment effects,” Brookings Institution. brookings.edu

Last updated September 2026. Every document referenced here is also on the sources page. Found an error? Tell me and I’ll fix it and say that I did.

Copyright 2026 — YpsiDatacenter.org. All rights reserved.