Crusoe's $30.9bn price is a bet on unbuilt gigawatts

Crusoe has raised $3.9bn in a Series F that values the Denver company at $30.9bn, with Nvidia among the new investors and Atreides Management, Mubadala Capital and Valor Equity Partners co-leading the round. Equity of that size is rare enough in private infrastructure that the headline tends to absorb the attention, and it probably will; the more revealing figure sits underneath, in the ratio between what the market just paid and what the company actually runs — 6GW of contracted capacity across Crusoe's cloud and data center operations against a little more than 1GW in service.

Do the arithmetic and the spread is stark. The whole valuation against the entire contracted pipeline works out to about $5m a contracted megawatt, while the same valuation against the operating base — “more than 1GW,” as the company puts it — comes to no more than about $31m a megawatt in service. The two denominators are not strictly comparable and the coverage does not break the 6GW down by asset, so read the gap as directional: equity and debt alike have been priced off the pipeline rather than the plant all year.

What “contracted” means is undefined — it could be signed leases, cloud commitments, or land under control, and the difference matters enormously at a $30.9bn mark; a megawatt leased to a named counterparty is a cash flow, while a megawatt under option is an intention. The coverage does not settle which reading applies, and the round is priced as though the stronger one holds.

Crusoe's route to this position runs through crypto: it launched in 2018 as a cryptominer, moved into AI cloud services and data center development, and has since divested the mining operations. It is the developer behind OpenAI's first Stargate campus in Abilene, Texas, and it is working on a 1.4GW site in Texas, two sites in Missouri and natural gas-powered data centers in Alberta, while leasing space from third-party providers to host its cloud platform where that is faster than pouring concrete. The company describes the ambition as controlling the infrastructure “from electrons to tokens,” the phrase co-founder and chief executive Chase Lochmiller uses for the vertical integration the Series F is meant to fund; a business that sold its miners now carries a valuation built on gas turbines, land and cloud contracts.

On the cloud side there is at least one named customer, and it is large: earlier this month Crusoe signed a cloud services agreement with the trading firm Jane Street that Data Center Dynamics reports is worth $13bn over five years, a single contract reportedly worth more than three times the equity raised this week. The engagement has been carried as a rumored $13bn deal since September 4, with terms beyond that reported total absent from the coverage. Concentration of that shape is something the round's investors are underwriting.

Abilene is what puts Crusoe on the anchored side of this market's capital hierarchy. Anchored digital infrastructure earns infrastructure pricing while everything else fights for capital, and a named OpenAI project is the kind of counterparty that changes how lenders and landlords price a site. The rest of the portfolio will be underwritten on its own merits, and in Alberta the coverage discloses no counterparty at all.

Leasing is the bridge and the tell. Owning campuses takes years and a power strategy; leasing third-party space lets Crusoe bill for cloud capacity against hardware that is already racked, revenue today against construction tomorrow. The leased racks are carrying the company while the owned gigawatts are financed, and the owned gigawatts are where the valuation lives. If the campuses slip, the cloud business still bills; if the campuses land, the margin moves inside Crusoe's own walls.

The investor list carries the rest of the information. Alongside the three leads sit Founders Fund, GIC, Nvidia, the Qatar Investment Authority, Radical Ventures and TPG — growth capital, three sovereign-linked names and one chip company. The mix matters because growth funds price on revenue trajectories and sovereigns price on decades, and a 6GW pipeline asks for both. Nvidia's presence is the one that says something; the source does not disclose the size of its check, and equity is the smaller part of what a chip vendor brings to a compute developer. When Nvidia's $3bn SB Energy investment behind OpenAI's Ohio campus was announced, the argument was that the guarantee, more than the equity, was what made the project financeable. Here the logic runs the other way: a chip maker holding equity in the companies that rent out its silicon owns a position in the demand curve as well as the supply one, and the check is small next to the order book it defends.

Nvidia's name appears in 70 stories in PWD's tracking, against 12 for Crusoe and four for Jane Street — a fixture that turns up in nearly every capital event this buildout produces, whether or not it leads the round.

The lead economics deserve a second look. Atreides co-leads a $3.9bn round with $8.9bn in regulatory assets across 47 accounts and a 26-person team, which makes the round equivalent to about 44% of its book before Mubadala Capital and Valor write anything. Lead status in infrastructure is a coordinating job as much as a capital one, and a round of this size is nearly all coordinating: assembling growth funds and sovereigns around an asset that needs a decade of patience and a fuel contract.

Read the site list as a power portfolio and it leans firm: a 1.4GW project in Texas, two sites in Missouri and data centers in Alberta powered by natural gas put gas at the dispatchable end of the generation stack, and this publication has argued that firm capacity now carries the premium while merchant renewables get repriced toward a discount. The coverage does not say whether Crusoe is building the generation in Alberta or contracting for the fuel, which leaves the ownership question open. Whatever the answer, as we argued in covering Schneider's factory-built power modules, the interconnection queue sets the buildout's clock, not the equipment.

The pipeline also carries a consent question the coverage only gestures at. Crusoe recently backed off a project in Cheyenne, Wyoming — the coverage does not say why — and was recently named as a potential partner on a Nebraska project, a mention rather than a commitment. Six gigawatts under contract is, in the first instance, a permitting claim: land, fuel and local agreement have to hold in each jurisdiction before any of it becomes engineering. Pipeline megawatts are only as firm as the places willing to host them, and consent now functions as capital in this buildout.

Where this week departs from recent form is that it came with numbers attached. The infrastructure announcements of the past month have tended to disclose position and withheld price — 2.7GW with no capital stack, in the case covered on September 10. Crusoe's round has a valuation, and its cloud contract has a reported value, which is more than most of the class has offered lately. What remains unpriced is everything that decides the return: construction cost per megawatt, the debt, and the offtake terms behind the 6GW, none of which is in the coverage.

Equity of $3.9bn against a 6GW program is a down payment on a construction bill the source never sizes. The debt that funds the rest will be underwritten on the offtake, and the Abilene anchor and the Jane Street contract are the two names a lender would hold; the Alberta sites have none disclosed. That is the shape of the merchant data center market, where lenders are underwriting a leasing forecast rather than a customer. Crusoe has more named demand than most of the shelf, but it also has more gigawatts to fill than it has customers on the record.

The valuation's real test is arithmetic. If Crusoe converts contracted gigawatts into running ones — Texas first, then Missouri, with Alberta as the firm-power experiment — $30.9bn will read as a price paid early in a build that has barely started, and this week's equity will look like the cheapest layer of the capital stack. If the distance between 6GW and just over 1GW holds through next year, the round becomes a re-rating of a power book whose value is still mostly on paper, and the next capital event is the one that has to carry the difference. The number to watch is megawatts in operation, which the company has been quoting since before the round; it stood at just over a gigawatt when the $30.9bn was set.