Crusoe's $3.9B Round Pays 124x Sales for a Turbine-to-Rack Builder — and That's the Public-Market Warning Shot

Crusoe announced the initial closing of an oversubscribed $3.9B Series F at a $30.9B post-money valuation on Sept. 17, 2026, co-led by Atreides Management, Mubadala Capital, and Valor Equity Partners, with NVIDIA, Founders Fund, GIC, Qatar Investment Authority, TPG, and Radical Ventures participating. The round tripled Crusoe's primary mark from $10B in October 2025.

Crusoe is private, so the multiple has to be triangulated. Sacra estimates 2026 revenue around $250M; even on a conservative $300M run-rate, the $30.9B post-money implies an EV/Sales of roughly 103-124x. The scale-colo peers trade at 12x, the power-electronics specialist Vertiv at 8x, the turbine maker GE Vernova at 5.7x, and the merchant nuclear IPPs Constellation Energy and Vistra under 4x.

McKinsey projects 156 GW of AI-related data-center capacity will be needed by 2030. The U.S. grid is not adding anywhere near that fast. Crusoe's premise — and the bet this $3.9B round is funding — is that AI compute demand has shifted from being GPU-constrained to being power-constrained, and that the firm which owns the gas turbine, the switchgear, the module factory, and the GPU rack replaces the utility queue with an internal procurement schedule.

That is a different business from leasing a colo cage. Crusoe's flagship 1.2 GW Abilene, TX campus for Oracle and OpenAI was built gas-turbine-on-site, with GE Vernova supplying 29 LM2500XPRESS aeroderivative turbines capable of 5-minute cold starts — a flexibility profile the grid cannot match. The new vertical-integration bet is to compress the next 5 GW of buildout into a 352,000 sq-ft factory in Brighton, Colorado, producing Crusoe Spark modules deployable in \"as little as three months.\"

Crusoe is private, but the stack that gets built around it is overwhelmingly public. Three vendors have already signed on, with the combined gas-turbine and genset deals alone covering more than 1.75 GW of Crusoe's contracted pipeline.

Downstream, Crusoe's contracted campus roster is the most visible signal of demand quality: a 1.2 GW Abilene campus for Oracle/OpenAI (Stargate), a 900 MW second Abilene campus for Microsoft, and two further Texas sites plus a Missouri campus in advanced development. The OpenAI Stargate alone is a $100B+ program, with the Abilene build-out accounting for roughly $12B of cost.

The implication for the public market is mechanical. If Crusoe's 124x sales is the right private mark, then the listed picks-and-shovels feeding that stack are mispriced on the cheap, and the listed scale-colo peers — who rent cages inside someone else's power story — are mispriced on the rich. Three groups matter.

Three catalysts sit inside the next six months that will tell investors whether the Crusoe premium is a fair value or a private-market bubble.

Crusoe's $30.9B is a bet on a specific 2028 — that power remains the binding constraint and that integrated turbine-to-rack builders continue to win share against both utility-fed colos and merchant IPPs. Three structural risks could collapse that bet.