Fermi stock ($FRMI) deep dive: Rick Perry's 17 GW Project Matador power-and-data-center campus, the AP1000 nuclear and gas plan, why the anchor tenant walked and the stock fell 80%, the supply chain, and Fermi vs Vistra ($VST).
More Macroplane supply-chain guides · Baskets · Investment theses · Macro trends
Fermi ($FRMI) is a development-stage REIT trying to build Project Matador, a planned 11-to-17 GW behind-the-meter power-and-data-center campus in the Texas Panhandle, co-founded by former US Energy Secretary Rick Perry. It has zero revenue, no signed anchor tenant, and a market cap near $5.5B. The stock is down roughly 70% from its first day of trading after its primary customer walked away in December 2025. This is a look at what Fermi has actually committed to, what it has not, and why the whole equity is a bet on one signature it hasn't gotten yet.
👉 See it in context: $FRMI competes for the same AI-power dollars as the operators in the Power & Grid basket on Macroplane — $VST, $CEG, $GEV, $OKLO and the rest of the names actually selling electrons today.
Fermi stock has been one of 2026's sharpest round-trips. The setup:
So the price isn't tracking earnings — there are none. It tracks one question: will a hyperscaler sign a multi-gigawatt lease at a campus that is still mostly permits and dirt?
Project Matador, branded the "HyperGrid," is a roughly 5,800-acre campus near Amarillo, Texas. The plan is to generate power on-site and lease it — together with the buildings — to AI tenants, so the data centers never touch the public grid. Fermi calls it the largest energy-and-data complex in the US, and the numbers are deliberately enormous:
The structure matters: Fermi is a REIT — a landlord. It does not run the GPUs. It develops the land, builds the power and the shells, and collects long-term lease payments from whoever fills them. That is the same picks-and-shovels logic behind the rest of the data-center power trade, pushed to its extreme: own the megawatts, rent them to the AI build-out.
The headline that got Fermi attention is nuclear. The long-term base load for Project Matador is meant to come from four Westinghouse AP1000 reactors — about 4 GW — and the company has lined up the supply chain to build them:
The reactors are real engineering commitments, but they are a 2030s story. No AP1000 in the US has ever been built on its original schedule or budget — Vogtle ran years late and billions over. Fermi's nuclear base load is the part of the pitch that will take the longest and cost the most.
What actually powers Project Matador's first phase is gas, not atoms. Fermi has acquired more than 600 MW of natural-gas turbines and won preliminary approval from the State of Texas for the first 6 GW of gas generation on the site. Gas is what gets the first ~1 GW to a tenant by late 2026; nuclear, solar, and storage layer in over the following decade.
This is the same bet GE Vernova ($GEV) is monetizing across the industry: gas turbines are sold out for years because they are the only dispatchable power that can be stood up fast enough to meet AI demand. Fermi is buying into that queue rather than waiting for reactors. The risk is that buying turbines and permitting gas is the easy 10% — finding the tenant and financing the other $30B-plus is the hard part.
One genuinely differentiated asset: location. The campus sits next to the DOE's Pantex Plant, the only facility in the US that assembles and disassembles nuclear weapons. That adjacency gives Fermi built-in physical security, an existing federal-security footprint, and a "critical defense infrastructure" designation under executive order. For a campus pitching itself to government and frontier-AI tenants who care about secure, sovereign compute, sitting next to the nation's warhead plant is a moat competitors can't replicate.
Open $FRMI on Macroplane and the supply-chain graph is nearly empty — and that absence is the point. The deals Fermi has signed are all upstream, the inputs to a campus:
A REIT with a supplier list and no tenant is a construction project, not a cash flow. Everything downstream of the campus is still a slide.
Readers keep comparing Fermi to Vistra ($VST), and the pairing is useful precisely because the two are opposites. Both are Texas-based, both pitch nuclear-plus-gas power to AI data centers — but one sells electrons from plants that exist, and the other sells a render.
Vistra is an 1882-vintage integrated power producer with about 41 GW of generation running today across gas, nuclear, coal, solar, and storage, roughly 5 million retail customers, and $5B-plus of EBITDA. Crucially, it has already signed the contracts Fermi is chasing: a 20-year PPA with Amazon ($AMZN) for up to 1,200 MW from its Comanche Peak nuclear plant in Texas (whose NRC license now runs to 2053), and about 2,609 MW of 20-year nuclear supply to Meta ($META) across its PJM plants. That's roughly 3.8 GW of signed, decades-long AI-power demand against generation Vistra already owns. The market has rewarded it: ~$46B market cap, a stock that's compounded on the same AI-power wave as $CEG and $TLN.
Fermi is the venture-stage mirror image. It owns no operating generation, has no customers, and carries a ~$5.5B market cap entirely on the promise of a 17 GW campus. Vistra is what Fermi wants to become in a decade; Fermi is the call option on whether a brand-new developer can leapfrog into the incumbents' market.
| Fermi ($FRMI) | Vistra ($VST) | |
|---|---|---|
| What it is | Development-stage power-and-data-center REIT | Integrated power producer + retailer |
| Generation operating today | ~0 GW (permits + turbines) | ~41 GW (gas, nuclear, coal, solar, storage) |
| Revenue | $0 | Tens of billions; $5B+ EBITDA |
| Signed AI-power contracts | None (anchor tenant terminated) | ~3.8 GW (Amazon at Comanche Peak, Meta in PJM) |
| Nuclear | 4× AP1000 planned (2030s, licensing) | Comanche Peak operating to 2053 |
| Market cap | ~$5.5B | ~$46B |
| The bet | Will it sign a hyperscaler and get built? | Will AI demand keep re-rating existing nuclear? |
Different risk entirely. $VST is a cash-generating incumbent re-rating on AI demand; $FRMI is a binary on execution and a signature. Track both in the Power & Grid basket.
Fermi ($FRMI) is a development-stage real estate investment trust building Project Matador, a planned behind-the-meter power-and-data-center campus near Amarillo, Texas. It was co-founded by former US Energy Secretary Rick Perry and CEO Toby Neugebauer, incorporated in January 2025, and went public on Nasdaq and the London Stock Exchange in October 2025. It generates no revenue yet and intends to make money by leasing on-site power and data-center space to AI tenants.
Fermi stock fell more than 80% from its highs after disclosing on 11 December 2025 that its primary prospective tenant had terminated a $150M construction-funding agreement. Because the company is pre-revenue and the entire thesis depends on signing tenants, losing its anchor customer gutted the stock and triggered securities class-action lawsuits alleging the IPO overstated tenant demand and funding.
As of mid-2026, no. Fermi's one disclosed anchor-tenant agreement was terminated in December 2025. Management says it is in active discussions with multiple prospective tenants but has not announced a binding lease. The June 2026 rally was driven by reports of OpenAI's separate multi-gigawatt data-center demand, not by any Fermi contract.
Yes — Fermi is structured as a specialty REIT. It plans to own the land, power generation, and data-center buildings of Project Matador and earn long-term lease income from tenants, rather than operating compute itself. That makes it a landlord-and-utility hybrid for AI infrastructure rather than a cloud or chip company.
Project Matador is Fermi's flagship campus: roughly 5,800 acres near Amarillo, next to the DOE's Pantex plant, designed for up to 17 GW of on-site generation (gas, nuclear, solar, wind, battery) and 18 million square feet of data centers. The "HyperGrid" branding refers to powering the data centers entirely from on-site generation, independent of the public grid. The first ~1 GW is targeted for late 2026, powered initially by natural gas.
They're different risk profiles. Vistra ($VST) is a ~$46B operating power producer with ~41 GW running, ~5 million customers, and signed 20-year nuclear PPAs with Amazon and Meta — a cash-generating incumbent re-rating on AI demand. Fermi ($FRMI) is a pre-revenue developer with no operating power and no signed tenant, so it's a high-variance bet on execution. Vistra is the lower-risk, already-contracted exposure; Fermi is the speculative option on a new entrant getting built. This is not financial advice — review the filings and size for the risk.
That depends entirely on your tolerance for binary, pre-revenue risk. The bull case: a uniquely sited 17 GW campus next to Pantex, a Rick Perry-led team with regulatory pull, signed reactor and gas supply deals, and a market starved for AI power. The bear case: zero revenue, a $486M loss, an anchor tenant that already walked, active securities litigation, a tens-of-billions funding gap, and a nuclear base load a decade out. The equity is an option on signing a hyperscaler. This is not financial advice — read the 10-K and the lawsuit complaints before sizing anything.
Fermi ($FRMI) is a development-stage real estate investment trust building Project Matador, a planned behind-the-meter power-and-data-center campus near Amarillo, Texas. It was co-founded by former US Energy Secretary Rick Perry and CEO Toby Neugebauer, incorporated in January 2025, and went public on Nasdaq and the London Stock Exchange in October 2025. It generates no revenue yet and intends to make money by leasing on-site power and data-center space to AI tenants.
Fermi stock fell more than 80% from its highs after disclosing on 11 December 2025 that its primary prospective tenant had terminated a $150M construction-funding agreement. Because the company is pre-revenue and the entire thesis depends on signing tenants, losing its anchor customer gutted the stock and triggered securities class-action lawsuits alleging the IPO overstated tenant demand and funding.
As of mid-2026, no. Fermi's one disclosed anchor-tenant agreement was terminated in December 2025. Management says it is in active discussions with multiple prospective tenants but has not announced a binding lease. The June 2026 rally was driven by reports of OpenAI's separate multi-gigawatt data-center demand, not by any Fermi contract.
Yes — Fermi is structured as a specialty REIT. It plans to own the land, power generation, and data-center buildings of Project Matador and earn long-term lease income from tenants, rather than operating compute itself. That makes it a landlord-and-utility hybrid for AI infrastructure rather than a cloud or chip company.
Project Matador is Fermi's flagship campus: roughly 5,800 acres near Amarillo, next to the DOE's Pantex plant, designed for up to 17 GW of on-site generation (gas, nuclear, solar, wind, battery) and 18 million square feet of data centers. The "HyperGrid" branding refers to powering the data centers entirely from on-site generation, independent of the public grid. The first ~1 GW is targeted for late 2026, powered initially by natural gas.
They're different risk profiles. Vistra ($VST) is a ~$46B operating power producer with ~41 GW running, ~5 million customers, and signed 20-year nuclear PPAs with Amazon and Meta — a cash-generating incumbent re-rating on AI demand. Fermi ($FRMI) is a pre-revenue developer with no operating power and no signed tenant, so it's a high-variance bet on execution. Vistra is the lower-risk, already-contracted exposure; Fermi is the speculative option on a new entrant getting built. This is not financial advice — review the filings and size for the risk.
That depends entirely on your tolerance for binary, pre-revenue risk. The bull case: a uniquely sited 17 GW campus next to Pantex, a Rick Perry-led team with regulatory pull, signed reactor and gas supply deals, and a market starved for AI power. The bear case: zero revenue, a $486M loss, an anchor tenant that already walked, active securities litigation, a tens-of-billions funding gap, and a nuclear base load a decade out. The equity is an option on signing a hyperscaler. This is not financial advice — read the 10-K and the lawsuit complaints before sizing anything.