Crusoe Energy
The Stargate AI-Factory Deep Dive — a $10B round, a $23.6B secondary mark, and a 4× revenue guide carrying $7B of construction debt.
Crusoe is the most direct private pure-play on the AI-factory buildout — the developer behind OpenAI's first Stargate campus in Abilene, Texas, with a March 2026 deal to build a 900MW AI campus for Microsoft and a 45GW identified power pipeline. Revenue ran ~$276M in 2024 → ~$500M in 2025 (+81%), and management guides to ~$2B in 2026 — a 4× jump in a single year. The market has already voted: Crusoe raised at $10B in October 2025, and by June 2026 secondaries print near ~$23.6B (+136%) — the inverse of the usual secondary discount, with buyers outnumbering sellers 2.3-to-1. The tension is underneath: a ~$7.1B construction loan for Abilene Phase 2, heavy customer concentration, GPU-hour price deflation, and two June 2026 cracks — a lost Wyoming campus and a CEO publicly fielding demand questions. The December 2025 hire of MongoDB's IPO CFO says the listing is coming; the guide has to land first.
The Numbers at a Glance
1 · Business Overview
From flared-gas bitcoin miner to the AI-factory company
| Metric | Value |
|---|---|
| Founded | 2018 · Chase Lochmiller & Cully Cavness |
| HQ | Denver, Colorado |
| CEO / COO-CFO | Chase Lochmiller · Michael Gordon (ex-MongoDB, hired Dec 2025) |
| Round Valuation | $10B · Series E, October 2025 ($1.375B, Valor + Mubadala, ~$84/share) |
| Secondary-Market Price | ~$23.6B implied · ~$198/share, June 2026 · +136% vs round |
| 2024 Revenue | ~$276M · +82% YoY |
| 2025 Revenue (est.) | ~$500M · +81% YoY |
| 2026–2028 Guide | ~$2B → ~$3.6B → ~$5.5B |
| Equity Raised | ~$2.9–3.9B · $15B+ total capital incl. debt facilities |
| Anchor Projects | Stargate Abilene (OpenAI/Oracle) · 900MW Microsoft campus (Mar 2026) |
| Key Investors | Valor, Mubadala, Founders Fund, NVIDIA, Fidelity, T. Rowe Price |
| IPO Status | No S-1 filed · IPO-caliber CFO hired December 2025 |
Crusoe started in 2018 with a contrarian energy idea: put compute next to stranded and flared natural gas that would otherwise burn into the sky, and monetize wasted energy as bitcoin hashrate. The pivot is what makes the company interesting: Crusoe exited bitcoin mining entirely (selling the mining business to NYDIG) and redeployed the same energy-first playbook into AI data centers — where the binding constraint is no longer chips but power. Today Crusoe is a vertically integrated "AI factory" developer-operator: it finds or builds the energy, constructs the campus, and sells GPU compute on top — on-demand at roughly $2–3 per GPU-hour and under reserved contracts running six months to three years.
2 · Product Suite & Moat
Energy-first, vertically integrated — because power is the bottleneck
| Product | Description | Strategic Role |
|---|---|---|
| Crusoe Cloud | GPU compute — on-demand (~$2–3/GPU-hr) + 6mo–3yr reserved | Revenue engine · 150% ARR growth in 2025 |
| AI Factories / Data Centers | Build-and-operate campuses — Abilene 1.2GW flagship | Long-lease infrastructure backbone |
| Energy Portfolio | Stranded gas, wind, solar, on-site turbines · ~7GW identified, 45GW pipeline | The moat — speed-to-power |
| Digital Flare Mitigation | Original flared-gas capture heritage | Cost + ESG differentiation |
| Crusoe Spark | Modular / edge data-center units | Fast-deploy optionality |
The moat thesis rests on four pillars:
- Energy vertical integration — in a market where every hyperscaler is power-starved, Crusoe controls generation, land, and interconnects. Goldman projects data-center power demand compounding ~15% through 2030; owning the electron supply chain is the scarce capability.
- Speed-to-power — Abilene went from dirt to a 1.2GW AI campus on a timeline hyperscalers struggle to match internally; that speed is why OpenAI, Oracle, and now Microsoft anchor with Crusoe rather than waiting on utility queues.
- Marquee anchor stack — OpenAI/Oracle (Stargate), Microsoft (900MW, March 2026), NVIDIA on the cap table as investor and supplier. Bookings grew ~5× across the first three quarters of 2025; added total contract value grew ~17× YoY.
- Structural cost edge — stranded-energy economics mean power costs competitors can't replicate at grid prices, which matters more as GPU-hour pricing deflates.
3 · Financials & Unit Economics
Four straight years of ~80% growth — then a guided 4× leap
Crusoe's disclosed-and-estimated revenue path is remarkably consistent: ~$80M (2022) → ~$152M (2023, +79%) → ~$276M (2024, +82%) → ~$500M (2025E, +81%). What the $23.6B secondary mark is actually paying for is the next leg: management's guide of ~$2B in 2026, driven by Abilene energization and the Microsoft campus, then ~$3.6B in 2027 and ~$5.5B in 2028. The company has not disclosed clean profitability figures — the economics rest on long-hold, contracted infrastructure returns, funded by layered debt.
Signals under the topline: cloud bookings grew ~5× across the first three quarters of 2025; the company's February 2026 update cited ~17× YoY growth in added total contract value, 150% cloud-ARR growth, and ~70% new-logo growth. On the other side of the ledger sit the facilities: a $225M Upper90 line, $750M Brookfield facility, $175M Victory Park facility, and a ~$7.1B J.P. Morgan-led construction loan for Abilene Phase 2 — plus the $11.6B in financing commitments assembled for the Abilene buildout in 2025. Equity of ~$2.9–3.9B against $15B+ of total capital means the common stock sits beneath a very tall stack.
4 · IPO Status & Timeline
No S-1 yet — but the CFO hire is the tell
Crusoe has not filed an S-1, confidentially or otherwise, as of early July 2026. The strongest path-to-public signal is personnel: in December 2025 Crusoe hired Michael Gordon — the COO/CFO who led MongoDB through its IPO — into the same role. Companies do not hire IPO-proven public-company CFOs to stay private. Add an October 2025 tender offer (interim liquidity at $108/share, ~$12.8B implied) and a CoreWeave public benchmark trading near ~7× forward revenue, and the setup reads as a 2027-window listing, contingent on the 2026 ramp printing.
Secondaries at a 136% premium to the round
Our Kraken deep dive documented the rare case of secondaries printing below a defended round. Crusoe is the mirror image: the October 2025 Series E priced at ~$84/share ($10B), an October tender cleared at $108 (~$12.8B), early 2026 prints ran through the $120s, and by June 2026 marks sit near $198 (~$23.6B) — with reported buyer demand outstripping supply ~2.3:1.
- A premium this large means secondary buyers are pre-paying for the 2026 guide — at ~$23.6B the stock trades ~47× trailing (2025E) but ~12× the $2B forward guide.
- The reference public comp, CoreWeave, trades near ~7× forward revenue — so even the forward math needs the guide to land and a premium multiple to hold.
- Momentum premiums built on 2.3:1 demand imbalances can compress as fast as they form — the June Wyoming headline is exactly the kind of catalyst that flips the ratio.
Base case: an S-1 once 2026 revenue visibility is provable — H1 2027 listing window. The catalyst chain to watch: Abilene Phase-2 energization → 2026 bookings-to-revenue conversion → confidential S-1 → EDGAR flip.
5 · Competitive Landscape
The neocloud pack — public comps set the multiple
| Company | Status | Positioning vs Crusoe |
|---|---|---|
| CoreWeave (CRWV) | Public · Mar 2025 | Largest GPU neocloud; the valuation benchmark at ~7× forward revenue; chips-first, buys power |
| Nebius (NBIS) | Public | International GPU cloud; strong balance sheet, less US power depth |
| Applied Digital (APLD) | Public | Data-center developer converting to AI hosting; closest infrastructure analog |
| Lambda Labs | Private | GPU cloud for AI labs; developer brand, thinner energy story |
| Together AI | Private | Inference/serving layer; a customer-competitor hybrid for capacity |
| Hyperscalers (MSFT · GOOGL · AMZN · ORCL) | Public | Both customers and rivals — Oracle contracts Crusoe capacity; Microsoft anchors a campus and builds its own |
Crusoe's differentiation in this pack is singular: it is the only name whose primary asset is energy, not GPUs. CoreWeave buys power to feed chips; Crusoe builds generation and sells the compute on top. In a power-constrained decade that ordering matters — but it also means Crusoe carries construction-and-energy execution risk that pure GPU clouds don't, financed with utility-scale debt on a startup's equity base.
6 · The Stargate / Abilene Bet
One campus carrying the guide — and the first cracks
Abilene is the flagship: OpenAI's first Stargate campus, ~1.2GW, developed and operated by Crusoe with Oracle as the compute counterparty — financed by $11.6B in commitments assembled across 2025 and a ~$7.1B JPMorgan-led construction loan for Phase 2. In March 2026 Crusoe added a 900MW AI-factory campus for Microsoft — proof the model extends beyond one anchor. Sacra-level analyses attribute roughly $250M of 2026 revenue to Abilene alone, with the balance of the ~$2B guide coming from cloud growth and new campuses ramping.
- Why it can work — contracted, multi-year infrastructure revenue from the most capital-rich counterparties on earth (OpenAI/Oracle/Microsoft), with NVIDIA allocation access as a cap-table investor.
- The June 2026 cracks — Bloomberg reported Crusoe lost a Wyoming AI campus after failing to secure anchor customers including Google, forcing project restructuring; days earlier the CEO was publicly addressing demand-softening questions. Neither is fatal; both are exactly what a 136% secondary premium cannot absorb many more of.
- The pricing overhang — GPU rental rates have already swung from ~$8 to ~$2 per GPU-hour. Long-term contracts insulate the ramp, but every renewal reprices into a deflating curve.
7 · Valuation Framework
36× trailing at the round — 12× forward at the secondary, if the guide lands
| Mark | Implied Value | vs 2025E Rev (~$500M) | vs 2026 Guide (~$2B) |
|---|---|---|---|
| Series E round · Oct 2025 | $10B | 20× | 5× |
| Tender · Oct 2025 | ~$12.8B | 26× | 6.4× |
| Secondary · June 2026 | ~$23.6B | ~47× | ~12× |
| CoreWeave (public benchmark) | — | — | ~7× forward |
The frame is simple: at the secondary mark, Crusoe must both hit a 4× revenue year and sustain a ~70% multiple premium to CoreWeave to justify today's ~$23.6B. The bull rebuttal is mix — Crusoe's revenue skews toward contracted, energy-integrated infrastructure rather than spot GPU rental, which arguably deserves a premium; and if the 2027 guide (~$3.6B) is credible, the secondary is ~6.5× two-year-forward. The bear frame: every turn of that multiple sits on top of ~$7B of project debt and 1x preference stacks that get paid before common — leverage cuts both ways at 47× trailing.
8 · Key Risks
What public-market diligence will price in
| Risk | Severity | Mitigant |
|---|---|---|
| 2026 guide execution (4× in one year) | High | Bookings 5×, TCV 17× YoY — but construction, power, and customer timelines must all land |
| Leverage / debt service | High | ~$7.1B construction loan is project-level and asset-backed; contracted revenue services it — until it doesn't |
| Customer concentration (OpenAI · Oracle · Microsoft) | High | Multi-campus, multi-anchor diversification underway; Microsoft deal is the proof point |
| Secondary-premium fragility (+136%) | Medium | 2.3:1 buyer demand today; premium compresses fast on execution slips — Wyoming was the warning shot |
| GPU-hour price deflation ($8 → $2) | Medium | Reserved multi-year contracts + energy cost edge cushion the curve |
| Anchor-walk repeats (Wyoming / Google) | Medium | 45GW pipeline gives redeploy options; but each walk hits the growth math and the narrative |
| No disclosed profitability | Low–Medium | Standard for the model pre-IPO; the S-1 will force the margin conversation |
9 · Pre-IPO Exposure Routes Today
Indirect vectors before the eventual listing
| Route | How | Caveats |
|---|---|---|
| Private secondary marketplaces | Crusoe common via secondary platforms (~$198/share, June 2026) | Accredited only · +136% above round · ROFR/company consent · nested-SPV fee stacks · debt + 1x prefs sit above you |
| CRWV | Liquid public neocloud proxy | Chips-first model, no energy vertical; the multiple anchor |
| NBIS / APLD | Sector beta on AI-infrastructure buildout | Different geographies and balance-sheet profiles |
| NVDA · ORCL · MSFT | Investor and counterparty exposure | Heavily diluted — Crusoe is a rounding error inside each |
| Patience | Wait for the S-1 with audited financials | Cleanest entry · the 2026 revenue print will be in it |
Bottom Line
Crusoe is the purest private expression of the decade's most crowded trade — the AI-factory buildout — with a real edge (energy) and real anchors (OpenAI, Oracle, Microsoft, NVIDIA). Four consecutive ~80% growth years and a 5×-bookings 2025 earn the excitement; a $23.6B secondary mark against ~$500M of trailing revenue means the market has already spent the 2026 guide before it prints. The MongoDB-CFO hire says an IPO is the plan; the ~$7.1B construction loan, three-customer concentration, and June's lost Wyoming campus say the path is levered and non-linear. For aggressive-growth investors, this is a high-conviction watchlist name with a valuation-discipline problem: the round was 20× trailing, the secondary is 47× — and the cleanest risk-adjusted entry remains the S-1 with audited 2026 numbers, where the guide either shows up or reprices everything above it.