There is one thing the AI buildout cannot scale without.
Not chips. Not capital. Not data.
**POWER.**
ERCOT just told Texas lawmakers that AI-driven peak demand could quadruple by 2032. Utility interconnection queues are five-year backlogs. Hyperscalers need firm redundant power years before any utility can deliver it.
There is exactly one publicly traded company sitting on 7,500 acres in Texas with a 6 GW Clean Air Permit (the second-largest of its kind in U.S. history), $1.4B of installed infrastructure, $785M of fresh financing, and a permitting path for 4.4 GW of advanced nuclear on top.
It should be trading at a premium.
Instead, seven months after its $21 IPO, the stock is down 73%.
Because here's what the market saw in the last six months:
1. Tenant: gone.
2. Earnings: missed.
3. CEO: fired for cause.
4. Founder: suing the company.
It makes sense why investors ditched this stock post-IPO. But there are some huge reasons to get back in.
1. The assets at this company actually GREW by $441M during the same quarter the stock collapsed.
2. The new institutional team that just took over is the most all-star roster I've seen step into a turnaround in years.
3. A public-market peer just ran this exact playbook and returned 736% in twelve months.
The ticker is FRMI. Here's the full thesis.
# What FRMI Actually Is
Fermi Inc. (FRMI) is building the largest private power grid in America.
The flagship asset is Project Matador. 7,500 acres in the Texas Panhandle on a 99-year ground lease with Texas Tech. At full build-out, the campus is designed to deliver up to 17 GW of integrated power: combined-cycle natural gas, advanced nuclear (Westinghouse AP1000), solar, and battery storage. Up to 15 million square feet of AI-ready powered shells colocated on top.
The business model is simple. Rent indexed to GW of reserved power. Per management, a SINGLE signed gigawatt = roughly $1B in annual NOI.
The setup we hit in the hook is the entire reason this company exists. Here's how Chairman Marius Haas framed it on this morning's earnings call:
>
Fermi builds a private grid behind-the-meter, sticks AI-ready powered shells on top, and leases the whole thing out by the gigawatt. Months to energize, not years. Haas's own words: Fermi was "purpose-built to relieve that constraint."
# So Why Is The Stock at $5?
Three credibility hits in five months.
**October 2025:** IPO at $21. Stock rips to $36.99 within weeks. Bullish.
**December:** The only disclosed anchor tenant terminates their $150M Advance in Aid of Construction Agreement. Stock drops 33% in a single session.
**March:** Q4 earnings confirm no signed cornerstone tenant. Strategy pivot from single-anchor to multi-tenant mid-call. $486M FY25 GAAP loss. Class-action lawsuit follows.
**April:** The Board fires the founder/CEO Toby Neugebauer FOR CAUSE. The CFO transitions out (he's now back on the Board). Neugebauer, who controls roughly 40% of the float with affiliates, immediately sues the company for wrongful termination and publicly demands a sale.
That's the headline reel. Stock cratered from $36.99 to $5.72. 73% off IPO. 85% off highs.
That's the bear case. It’s real, it’s ugly, and it’s well-documented. But everything has completely changed the last couple of months. I’ll tie this back in later on when we talk about the new management and their 90-day plan to secure a hyperscaler tenant.
READ that AGAIN. They have a 90-day plan to secure a hyperscaler tenant. The clock is ticking.
# Here's Something that Nobody Talks About: APLD Already Ran This Playbook
Applied Digital (APLD) is the proof.
If you haven't been paying attention to APLD, here's the short version: the stock is up 736% in the past twelve months. Market cap is north of $13B, and the market now treats it as a contracted-revenue infrastructure bet.
Here's the part that matters for FRMI: 18 months ago, APLD looked EXACTLY like FRMI looks today. Heavy capex story. No signed hyperscaler tenant. Balance sheet that the market viewed as inadequate to the ambition. Endless skepticism.
Then three things happened. Each one is a template that FRMI can replicate.
**ONE. The first investment-grade lease (2024).** CoreWeave signed a 400 MW lease at APLD's Polaris Forge 1 campus in North Dakota. Roughly $11B of contracted lease revenue over 15.5 years. That single contract took APLD from a speculative microcap to an institutional infrastructure REIT in the market's mind. The stock 5x'd from pre-announcement levels within months.
**TWO. The institutional capital partnership (2025).** Macquarie Asset Management committed up to $5B of preferred equity, structured at roughly $2.25M per MW of leased capacity. Macquarie called the relationship a potential unlock of $25B of total data center capital. That partnership made the entire AI campus business model financeable at scale without continuous equity dilution.
**THREE. The multi-tenant validation (2026).** Two MORE investment-grade hyperscalers signed: 200 MW at Polaris Forge 2 ($5B contracted) and 300 MW at Delta Forge 1 ($7.5B contracted). Total contracted lease revenue is now over . More than 50% is backed by investment-grade counterparties. APLD then issued a $2.15B high-yield bond against those contracts. The public debt markets accepted the cash flows as bankable.
So, the playbook isn't speculative anymore. It is documented and proven. The per-MW pricing, the contract tenor, the financing architecture, the operating model. ALL of it has cleared at scale.
Per-MW contracted lease value is roughly $25M per MW of IT load. That number matters. We'll come back to it.
This is what FRMI is about to run. Same product. Same buyer pool. Same regulatory tailwind. Same financing template.
And the buyers know it. Per Co-President Anna Bofa on this morning's call:
>
The only real difference between FRMI and APLD-mid-2024 is that FRMI is starting from a much bigger asset.
# FRMI Is APLD But Bigger
Same trade. Bigger base.
Let me walk through it.
* **Same product:** 15-year power-capacity leases to investment-grade hyperscalers, structured as REIT-like rent streams.
* **Same buyer pool:** Microsoft, Meta, Amazon, Google, plus the neo-clouds (CoreWeave, Lambda, Crusoe) and the frontier-compute operators (OpenAI, Anthropic, xAI).
* **Same regulatory tailwind:** behind-the-meter permitting unlock. Trump admin EPA reforms in 2025 specifically targeted data center air permitting. Fermi's 6 GW TCEQ permit is the SECOND-LARGEST power-generation air permit ever granted in the United States.
* **Same emerging financing structure:** APLD has Macquarie at $5B preferred. Fermi has MUFG's $500M infrastructure facility as the first piece of the same architecture. Building out the broader programmatic capital partnership is one of the five explicit items on Fermi's near-term plan.
Now where they DIVERGE.
* **Pipeline:** APLD's ceiling is 4 GW of data center critical IT load. Fermi's permitted pipeline is 11 GW of gas plus a planned 4.4 GW of AP1000 nuclear. That is roughly 4x the pipeline.
* **Vertical integration:** APLD plays in ONE slice of the value chain. They lease the shell. They don't own the power generation. Fermi plays in all three: they generate the electricity, they deliver it behind-the-meter (no transmission losses, no utility markup), and they operate the shell. Every dollar of generation margin that would otherwise go to Constellation, Vistra, or NRG flows to Fermi's P&L instead.
* **Per-MW NOI:** roughly 40-60% higher than APLD because of that integration.
* **Speed:** this is the one management is leaning on hardest, and they have the receipts. Co-President Jacobo Ortiz Blanes on this morning's call:
>
* **Market cap:** APLD $13B & FRMI $3.65B
APLD is the certainty. FRMI is the optionality. The market is currently paying 3.5x more for the certainty than the optionality.
And we haven't even talked about the nuclear yet.
# The Nuclear Kicker
Of FRMI's 17 GW planned campus, 4.4 GW is Westinghouse AP1000 advanced nuclear.
APLD has zero nuclear exposure. Not in the asset base, not in the pipeline, not in the model. None.
This matters a lot. Let me show you why.
* **Nuclear capacity factor:** 90-95%. Natural gas capacity factor: 50-60%. Each nameplate MW of nuclear produces nearly 2x the actual electricity of a gas MW over the year.
* **Nuclear marginal cost once built:** $5-15 per MWh. Hyperscalers are paying $80-100+ per MWh for clean firm power. The Microsoft / Constellation Three Mile Island restart **deal is the public benchmark for those prices.**
* **Per-MW gross margin on nuclear:** roughly $0.6-0.8M per year. That is 2-3x the margin on gas.
* **Asset life:** 40-60 years on nuclear, against 7-15 year data center equipment refresh cycles. The nuclear plant outlives multiple generations of tenant hardware.
And the licensing path is materially clearer than people realize. From this morning's call, per Co-President Bofa:
>
Translation: front-end engineering by one of the world's largest EPC contractors, reactor vessel forging dies already in preparation at one of the only firms on Earth that builds them, and a regulatory fast-track that no other private company has.
Now the math. A 4.4 GW captive nuclear fleet, leased at Constellation-style economics, is by itself worth more than APLD's ENTIRE current $13B enterprise value.
That optionality is embedded in FRMI's $3.65B market cap at zero. The market is paying you nothing for the single largest piece of long-term value in the company.
This is exactly the kind of thing that doesn't show up in a quick screen. It is exactly the kind of thing that re-rates an entire equity story the day a contract gets signed.
# The Team (This Is The Whole Thesis)
Now the **most important section** in this entire write-up.
The APLD chart proves the trade works. The Fermi assets are real. The macro is in place. The near-term catalyst window is publicly committed. None of that matters if the team can't execute.
If they sign tenants, you'll be reading about FRMI the same way people are talking about APLD right now. If they don't, the activist sale case takes over and at best you get a sale process at some premium. Either path works from this entry point. But the upside case lives or dies on this group.
That is why the April reset matters. Not because they swapped one CEO for another. Because they took the company from a founder-led startup to an institutional public-company operating team in a single Board meeting. And the new team is PRECISION CAST for the work in front of them.
Let me walk through it.
Marius Haas, Chairman.
Former President and Chief Commercial Officer of Dell EMC. A $90B+ revenue commercial sales engine. He spent a decade selling into the EXACT customer profile that Project Matador needs as tenants: Fortune 500 and federal. He also led ops through the Dell take-private LBO and the $67B EMC integration, so he knows what it looks like to run a large industrial business under capital and time pressure. This guy doesn't sit on the Board of a $3B company for fun. He owns the permanent CEO search.
And the CEO search is already running. Per Haas on this morning's call:
>
**Anna Bofa, Co-President.**
Google to early Dropbox to early Pinterest to Global Head of Product Programs at Meta to Co-founder of Crate AI. That resume tells you ONE thing very clearly: she has working relationships inside the hyperscaler product orgs that decide WHERE compute capacity gets sited. The single most important deal Fermi needs to close in the near term is a binding hyperscaler tenant. She owns that interface.
**Jacobo Ortiz Blanes, Co-President.**
Former COO of Fermi since inception. He's the institutional memory carrier and the day-to-day operational owner. The construction milestone list (11+ miles of perimeter fencing, 5 miles of high-pressure gas pipeline, 7 miles of water distribution, 86 MW of Xcel power brought on site, Siemens SGT-800 gensets in customs, GE 6B Frame turbines refurbishing in Houston, F-class 1.1 GW combined-cycle scheduled for Q3) is what he is delivering quarter over quarter.
**Larry Kellerman, Head of Power.**
40+ years in power. Former Partner at Goldman Sachs, where he built the firm's on-balance-sheet power generation business and led the Cogentrix Energy acquisition. Former CEO of Atlantic Power & Utilities AND Quantum Utility Generation. Founding President of Citizens Power. Architect of Fermi's 17 GW power plan from day one. He owns first power, the pending 5 GW air permit, and the infrastructure partnerships.
**Rob Masson, Interim CFO.**
Career public-company CFO at Hypertherm, Latham, Noble Supply. 13 years at Raytheon in operational finance before that. Naval aviator, Naval Academy, Harvard MBA. He owns working capital, project finance, and audit-ready reporting through the buildout.
**Mesut Uzman, Head of Nuclear.**
6+ years at the Emirates Nuclear Energy Corporation. One of the few teams on Earth with actual hands-on AP1000-era new-build experience. He owns the 4.4 GW nuclear leg, which as we just walked through is the largest piece of long-term value in the entire equity story.
**Rick Perry, Co-Founder and Director.**
Former U.S. Energy Secretary. Former Governor of Texas. He kept his Board seat through the reset. He owns the political and regulatory channel into the Nuclear Regulatory Commission, TCEQ, ERCOT, DOE, and the Trump-administration energy team. The Nuclear Regulatory Commission NEPA pilot designation and the DOE Office of Energy Dominance financing track that management referenced on the call are early public proof points of how much that channel is worth.
That's seven unambiguous owners. Seven individually credible track records that map directly onto the work that has to be done.
It is the difference between a founder-led venture project and an institutional public-company operating team. It is the bridge between APLD's $13B certainty and FRMI's $3.65B optionality.
Every single one of these people saw the same balance sheet, the same litigation, the same 40% activist block, the same broken share-price chart. They took the job anyway. That tells you something.
If they execute, this name does not stay at current prices for long.
# The 90-Day Clock
Chairman Haas closed today's call by laying out the management priorities in plain language. Five items. Each one is individually a stock-moving catalyst:
1. Attracting premier tenants who recognize the unique value of the platform.
2. Building the best private power grid on the planet, in close collaboration with suppliers and partners.
3. Ensuring sufficient capital to support liquidity needs.
4. Accelerating strategic partnerships in both power and data centers.
5. Investing in the people and talent pipeline, including key leadership additions.
I read this as a 90-day catalyst window. Why? Because the commercial momentum is already there, and the team is publicly committing to delivering inside it. Per Co-President Bofa:
>
And this one, which I think is the single most important quote from the entire call for the catalyst case:
>
Remember what happened to APLD after the CoreWeave lease in 2024. Within months, the stock 5x'd from pre-announcement levels. That single print took APLD from speculative microcap to institutional REIT in the market's eyes.
That same kind of print is what Fermi management is publicly working toward. The team that Marius Haas and Anna Bofa now lead was specifically reconstituted to close exactly that gap.
If they sign one tenant, FRMI is APLD-mid-2024 with a bigger asset and a deeper pipeline. If they sign two, it is a different conversation entirely.
The window matters. Hyperscaler agreements at this scale, once they get close, are pre-announced as the share price moves. You don't get to wait for the actual print. The asymmetry compresses fast.
# The Math
One more number that should be permanently lodged in your head.
APLD's realized contracted lease value is approximately $25M per MW of IT load. Use that as your peg.
A single 1 GW lease at FRMI on those economics = roughly $25 BILLION of contracted backlog.
For perspective: that is comparable to APLD's ENTIRE current $23B+ contracted portfolio. Built over three landmark hyperscaler deals across 2024-2026.
FRMI prints that in one signing event. From a $3.65B starting market cap.
The market reaction would not be subtle. It would not be slow. It would not be polite.
And that is just from a single gigawatt. FRMI is built to lease 11 GW of gas plus 4.4 GW of nuclear on top.
This is what asymmetric setup actually means.
# The Risks (Real Ones)
I'm not pretending this is risk-free. The bear case is real.
* No tenant signed yet. Until that happens, this is still pre-revenue and dependent on continued capital access. The entire thesis ultimately rests on a binding lease being signed.
* Capex per MW is roughly 2x APLD's. Vertical integration captures more margin per MW but costs more capital per MW. Total project cost per integrated MW runs $15-25M against $8-12M for data-center-only. A full 17 GW buildout will require $50-100B of cumulative capex over a decade.
* Time-to-first-revenue is 2-3x longer. APLD energized its first 100 MW within 12-14 months of groundbreaking. Fermi's first 2.2 GW of gas is arriving on site now, but the nuclear leg is a multi-year endeavor. Hyperscaler patience matters.
* Governance overhang. Neugebauer's 40% block and his wrongful-termination lawsuit are a live dispute. Sale-process forcing risk is real. That said, the Board has been firm. Haas on the call: 'A forced sale at this moment is not in the best interest of the long-term shareholders, especially with anchor tenant negotiations advancing and our financing structure intact...we're not going to be stampeded into a short-sighted decision.'
* Single-site concentration. FRMI's asset is one campus. APLD has three sites in two states. A single missed milestone at Project Matador affects the entire asset. On the flip side, Texas Tech publicly reaffirmed its support on today's call, so the ground-lease overhang from Q4 has been materially de-risked.
* Industry inexperience. No one on the current operating team has previously run a data center business. Hyperscaler customer relationships are being built from scratch through Bofa and Haas's networks. APLD has three years of operating-execution track record that FRMI does not.
None of those break the asset. None of them break the playbook. APLD has already proved the playbook works. The real question is whether this specific team can execute it.
That is the trade. And I think the team is good enough to clear it.
# Why I'm Positioning Now
Let me bring it all together.
APLD ran this playbook and returned 736% over twelve months. The trade is publicly proven. The buyer pool is hungry. The contract template is standardized. The financing architecture is documented.
FRMI is selling the same product, into the same buyers, with a bigger and more vertically integrated asset, on more aggressive permits, at a fraction of APLD's market cap.
The April reset was the unlock. The post-IPO implosion was a credibility crisis, not an asset crisis. The new team maps directly onto the catalysts. The catalyst window has been publicly committed.
And per Co-President Jacobo Ortiz Blanes today, the operational side is ready:
>
The only missing piece is one signed lease.
When that print comes, the asymmetry compresses immediately. You don't get to wait. You don't get to think about it. You don't get to read the press release and then back up the truck at $9.
You position before the catalysts, not after.
I'd rather be a few weeks too early than one day too late.
I'm long FRMI.
Sizing accordingly.
Have a great day.