DEEP RESEARCH · FRMI / Fermi America

Fermi America (FRMI) — A New AI-Infra Paradigm, or a Speculative Mega-Project?

An immediately-post-IPO REIT planning an 11 GW × 18M-sqft self-grid campus on 5,769 acres in Texas — a bipolar bet between vision and execution risk.

Date: 2025-10-02 · Personal research note · Sources: S-11 registration, company IR, NRC/DOE materials, peer comparisons

Reference video (mentor's analysis): YouTube — FERMI inside U.S. energy projects

0. Bottom line first

I. The Fermi proposition — a new AI-infra paradigm

A. Business model: a vertically integrated power+data campus

Official fact: Fermi is a hybrid of independent power producer and hyperscale data-center developer — generating electricity "behind the meter" and selling it directly to its own on-site data centers. Structured as a REIT (≥90% of taxable income distributed to shareholders).

Interpretation: The REIT wrapper buys tax efficiency but blocks the "retained earnings → self-funded expansion" path. The decades-long, multi-tens-of-billions plan will require permanent reliance on external capital (equity + debt).

B. The flagship — the Matador Project

C. Matador phased roadmap

D. Path to revenue & cash conversion

Official fact: Founded January 2025. As of 2025-06-30 there is zero revenue and a cumulative net loss of $6.4M. First tenant revenue is expected starting 2027, after Phase 1 infrastructure delivers.

II. Two megatrends — AI compute × the energy crunch

A. The AI-driven data-center boom

Official fact: Global DC market est. $347.6B (2024) → $652.0B (2030), CAGR 11.2%. A more aggressive scenario: $505.8B (2024) → ~$1.2T (2032), CAGR 11.51%. North America (chiefly the U.S.) holds 40%+ share with ~5,530 data centers.

Official fact: McKinsey — $5.2T in AI-related DC capex through end-2030. AI workloads expected to drive 160% incremental DC demand; AI workloads ~70% of all capacity demand by 2030.

B. The power bottleneck

Official fact: U.S. AI-DC power demand forecast: 4 GW (2024) → 123 GW by 2035 (~30×). Global DC power demand forecast: +165% 2023→2030. Some new utility interconnect queues stretch up to 7 years.

Interpretation: Fermi's business logic rests on the assumption that public grids cannot keep up with AI demand → private integrated solutions become necessary. Buying FRMI is implicitly a "bearish bet on U.S. public-grid modernization speed".

C. The nuclear renaissance

Official fact: The 2024 ADVANCE Act and a series of 2025 executive orders explicitly modernize NRC licensing, cut fees, and accelerate advanced-reactor deployment. Nuclear has best-in-class >90% capacity factor.

III. Competitive landscape & potential moats

Potential moats (conditional on execution)

• Unprecedented scale and integration — 11 GW × 18M sqft is staggeringly costly to replicate.
• High switching costs — once a hyperscaler's IT stack lands, leaving is operationally and financially prohibitive.
• First-mover at gigascale integration — locks in real estate, supply chain, and regulatory pathway.

Strategic partnerships — a global consortium for risk mitigation

Interpretation: The most recent U.S. AP1000, Vogtle, ran $20B+ over budget and 7 years late, effectively making U.S.-led nuclear construction "uninvestable." The Hyundai/Doosan tie-up is direct insurance against that trauma — though it introduces new geopolitical/logistical risk.

IV. Financial analysis & capital strategy

A. Initial capital structure

Official fact: The IPO was reported oversubscribed. Implied market cap $12.5B–13.8B. Dual-listed on NASDAQ and LSE (ticker FRMI).

B. Future capital needs

Official fact: Total estimated cost $70B–$90B; Phase 1 alone at least $2B.

Interpretation: The decisive financing card is a DOE Loan Programs Office loan; the company says it is in "pre-approval." A multi-billion-dollar DOE loan would simultaneously deliver (a) long-term, low-cost capital, (b) federal validation, (c) follow-on private financing momentum. Failure to secure it would be a near-existential blow.

C. Shareholders & capital return

• Co-founders include former U.S. Energy Secretary Rick Perry, his son Griffin Perry, and CEO Toby Neugebauer. Perry family combined stake reportedly valued $2.2B+ on day one.
• Macquarie remains a major institutional shareholder.
• 180-day IPO lockup expiry could enable insider selling and weigh on price.
• Despite REIT structure, management has explicitly told investors not to expect dividends in the near future.

V. Outlook & risks

A. Three-year milestones

1. Anchor tenants — a binding long-term lease with MSFT/GOOGL/AMZN/META would be the ultimate model validation.
2. DOE loan finalization — central to the financing strategy.
3. Phase 1 execution — 1.1 GW + 1M sqft by end-2026.
4. Nuclear licensing progress — NRC review milestones for the AP1000 COLA.

B. Key risks

C. Scenarios — Bull / Bear

Bull

Matador delivers each phase on time/budget; a major hyperscaler signs a binding long-term lease; DOE loan unlocked. FRMI re-rates as a foundational 21st-century U.S. infrastructure asset and the current market cap looks trivial in hindsight.

Bear

Nuclear-build delays + cost overruns + no anchor tenants + tightening capital markets compound. Follow-on financing fails. Significant probability of total capital loss.

VI. Risk-adjusted alternatives

• vs DLR/EQIX — proven operators with dividend track records. FRMI offers far more upside but exponentially higher downside.
• vs CEG — a more direct, lower-risk bet on AI-driven nuclear demand. Already cash-flowing nuclear assets and signed hyperscaler PPAs.
• vs NVDA/MSFT — direct exposure to AI value creation without single-mega-project dependency.

VII. Conclusion