Fermi’s first 10-K: a bold blueprint for an 11 GW AI power campus in Texas
Fermi Inc. has filed its maiden 10-K, and it’s a meaty read. The dual-listed developer (Nasdaq and LSE) is building “Project Matador” – a private, on-site power and AI infrastructure campus in the Texas Panhandle – with ambitions up to 11 GW of generation and roughly 15 million square feet of AI-ready powered shells. Think of it as an energy-first data campus designed to sidestep grid bottlenecks by putting the turbines next to the compute.
It’s early stage, capital intensive and audacious. Here’s what matters for investors.
Project Matador: permits, phasing and sheer scale
Fermi controls 5,236 acres under a 99-year ground lease with the Texas Tech University System and expects to expand the campus to about 7,570 acres through adjacent parcels. In February 2026, the Texas Commission on Environmental Quality (TCEQ) granted final approval for an approximately 6 GW natural gas-fired air permit. On 27 March 2026, Fermi applied for an additional 5 GW permit. If approved, the site could be authorised for up to ~11 GW of gas generation alone, with potential to push to ~17 GW if further land and permits are secured.
The roadmap runs in phases:
- Phase 0 (site enablement) – substantially complete: roads, water (including a 2 million gallon tank), early substations, ~11.3 miles of fencing, initial gas lines and engineering.
- Phase 1 – initial energisation: up to 200 MW of grid power from Southwestern Public Service (86 MW expected H2 2026; a further 114 MW targeted by year-end 2026 but may slip into 2027), plus early on-site gas units and battery systems to support first tenant commissioning.
- Phase 2 – scale-up and combined cycle conversions: aim for ~2.5 GW of gas-fired capacity as soon as end-2028, subject to financing, construction and permits.
- Phase 3/4 – nuclear buildout: long-term plan for Westinghouse AP1000 reactors (COL application filed June 2025; NRC accepted Sept 2025; EIS scoping notice Mar 2026) targeting up to 6.0 GW of nuclear baseload over time, with batteries for power quality and solar for gas displacement.
Power ramp: equipment in hand, timelines reset
Fermi has changed its timeline. The earlier target of ~1.1 GW online by end-2026 is off the table. The company now expects “nearly 2.0 GW” of gas + grid by the end of 2027, subject to project financing, installation/commissioning and tenant readiness.
Crucially, they’ve banked hardware to de-risk the ramp:
- Seven leased GE TM2500 mobile units – ~132 MW.
- Three GE 6B turbines – ~114 MW (refurb in progress).
- Six Siemens SGT-800s – ~300 MW (arrived Port of Houston in February 2026).
- Three Siemens SGT6-5000F (F-class) – ~720 MW contracted; turbine cores expected to begin shipping Q2 2026.
Those blocks total around 1.5 GW of simple-cycle capacity under Fermi’s control before efficiency gains from combined-cycle conversions. Early batteries are planned to smooth intra-second AI load swings and provide ride‑through.
Commercials: tenant LOI still non-binding, broader pipeline engaged
On customers, Fermi’s biggest near-term swing factor is converting interest into contracts. The 10-K confirms a non-binding LOI with an investment-grade “First Tenant”, but the associated advance agreement for up to $150.0 million was terminated by the tenant on 11 December 2025 after exclusivity expired. Discussions continue, and Fermi says it has opened talks with several other potential hyperscaler and AI infrastructure tenants, with some parties exchanging drafts or carrying out confirmatory diligence. As at the filing date, however, there are no definitive leases.
Why it matters: binding leases unlock non-recourse project finance and dictate sequencing. Without them, schedules and capital costs can drift.
Financing: sizeable runway, bigger cheques still to write
Fermi raised $745.6 million net in its October 2025 IPO. Since then, it has stitched together a stack of equipment and warehouse facilities to secure long-lead kit and grid hardware:
- MUFG: $500.0 million non-recourse turbine warehouse (Feb 2026).
- Keystone: $120.0 million high-voltage equipment facility, expandable by up to $100.0 million (Feb 2026).
- Beal/CSG: $165.0 million term loan for six SGT-800s (Mar 2026).
- Yorkville: $156.3 million senior unsecured promissory note (Mar 2026) – undrawn at filing.
As at 31 December 2025, Fermi reported $408.5 million in cash and cash equivalents and $935.3 million invested in property, plant and equipment. Management estimates Phase 0 and Phase 1 capex could exceed $3 billion in aggregate, with about $2 billion expected over the next 12 months – explicitly subject to executing a definitive tenant lease and aligning with deployment timelines. The company intends to operate as a REIT from 2025, with some activities pushed into taxable REIT subsidiaries.
Key numbers from the 10-K
| Cash and cash equivalents (31 Dec 2025) | $408.5 million |
| Property, plant and equipment | $935.3 million |
| Total assets | $1,413.3 million |
| Total liabilities | $317.4 million |
| Net loss (period to 31 Dec 2025) | $(486.4) million |
| Shares outstanding (23 Mar 2026) | 629,839,790 |
| Air permits | ~6 GW approved; +5 GW application filed |
| Grid power via SPS | Up to 200 MW (86 MW H2 2026; 114 MW targeted YE 2026/2027) |
| On-site generation secured | ~1.5 GW simple-cycle equipment under control |
| Target by end-2027 | Nearly 2.0 GW (gas + grid), subject to financing and readiness |
Enablers: fuel, water, fibre and permitting momentum
Fermi’s site economics hinge on logistics and inputs. On gas, there’s a firm supply arrangement with Energy Transfer, contemplating scalable deliveries up to ~300,000 MMBtu/day over time, and the site is near major pipelines. For cooling, the City of Amarillo approved up to 2.5 MGD with a path to 10 MGD, and Fermi has a 30‑year groundwater lease over ~2,542 acres; a hybrid cooling approach aims to trim water intensity. Two major fibre corridors run along the campus boundary, and the company has progressed grid interconnections at 230 kV and 115 kV. On nuclear, the NRC has kicked off EIS scoping and Fermi is in a pilot programme for applicant-prepared EISs – potentially a helpful (if not guaranteed) accelerator.
The good, the bad, and what to watch
What looks positive
- Permitting step-change – the ~6 GW air permit is a major de-risking moment, and a +5 GW application signals intent.
- Hardware secured – locking in turbines and HV equipment addresses global lead-time pain points.
- Gas, water, and fibre foundations – key site inputs are progressing, and early battery plans fit AI load dynamics.
- Nuclear first-mover posture – COL application accepted; EIS scoping started; FEED with Hyundai; forging readiness with Doosan.
What needs scrutiny
- No signed tenant leases yet – the LOI is non-binding and a previous construction advance agreement was terminated in December. Contracts drive financing.
- Timeline reset – the 2026 power target is gone; “nearly 2.0 GW” is now pushed to end‑2027 and is conditional.
- Capital intensity – Phase 0/1 could top $3 billion; total multi‑phase needs are very large and will flex with scope and technology mix.
- Controls and litigation – a material weakness in internal controls persists; a putative securities class action was filed in January 2026.
- Share price context – from first trading day to year‑end 2025, the stock’s indexed line fell from $100 to $25 in the performance graph.
My take: a power-first bet on AI scale, with execution front and centre
Fermi’s pitch is simple: AI needs firm, abundant power faster than the grid can deliver, so bring gas (and later nuclear) to the campus and rent power capacity as an “incident of tenancy”. The 10-K shows solid progress on permits, equipment and site enablement. However, the investment case hinges on two near-term catalysts: binding anchor leases and project-level financing. Until those arrive, timelines remain elastic and equity could stay volatile.
If Fermi signs one or more multi-gigawatt tenants and closes non-recourse project finance against those cash flows, the 6 GW permit and hardware stockpile put them in a strong delivery lane. If not, the carrying costs of long-lead assets and development overheads will bite.
What to track next
- Execution of a definitive lease (or leases) with hyperscale/AI counterparties.
- Financial close on project-level, non-recourse debt and any tenant prepayments.
- Energisation milestones: 86 MW SPS in H2 2026; timing for the additional 114 MW; mobilisation of TM2500s and installation of GE/Siemens units.
- Outcome of the +5 GW TCEQ air permit application.
- NRC EIS scoping progress and any schedule colour on the first AP1000 unit.
Bottom line
This is one of the market’s most ambitious private-power plays for AI. The upside is obvious: capacity scarcity meets a campus permitted at gigawatt scale. The trade-off is equally clear: no revenue yet, rising capex, and an execution path that depends on big contracts and bigger financing. For high-risk, long-duration infrastructure investors who believe in sustained AI power demand, Fermi’s 10-K lays out a credible but challenging path. For everyone else, keep it on the watchlist and let the next few quarters answer the two big questions: tenants and funding.
