Construction AI BriefSubscribe →
Issue
№050
Pillar
Trend
Audience
GC ops
Dated
2026.07.05

Crusoe's valuation just tripled to $30 billion. The way it builds data centers skips the traditional GC hierarchy.

Crusoe is in talks to raise $3 billion at a $30 billion valuation to keep building AI data center campuses — using a multi-prime delivery model where electrical and mechanical contractors hold contracts directly with the owner instead of subcontracting under a GC.

ByConstruction AI BriefAbout this publication

Crusoe is in talks to raise about $3 billion in a round that would nearly triple its valuation to roughly $30 billion, Bloomberg reported July 2 — up from the $10 billion-plus mark set in its October 2025 Series E. That money keeps funding a construction delivery model most GCs haven't bid against yet: multi-prime contracting, where the electrical and mechanical contractors sign directly with the owner instead of working as subs underneath a general contractor.

What is Crusoe actually building?

Crusoe pivoted from cryptocurrency mining to AI infrastructure, and now designs, finances, and builds data center campuses for hyperscale customers. The company has 4.9 gigawatts of capacity under contract and more than 40 gigawatts in its total pipeline, with named customers including Oracle, Microsoft, Google, and Meta. Its flagship campus in Abilene, Texas, built for Oracle, has grown from a two-building plan to eight buildings and roughly 4 million square feet, all still tracking toward a mid-2026 energization date.

What's different about how these jobs are contracted?

On that Abilene campus, DPR Construction runs the job — schedule, safety, site logistics — the way a CM/GC normally would. But electrical contractor Rosendin and mechanical contractor Southland aren't subcontractors reporting up through DPR. They're prime contractors with their own direct agreements with Crusoe. DPR manages roughly 60 subcontractors under its own scope; Rosendin and Southland manage another 30 collectively under theirs. Peak workforce across the campus is expected to hit nearly 5,000.

Traditional GC-led deliveryCrusoe's multi-prime model
Owner contracts withOne GCGC (as CM) + electrical prime + mechanical prime, separately
MEP relationshipSubcontractor to the GCPrime contractor to the owner
GC's roleHolds schedule risk and MEP riskHolds schedule, safety, logistics — not MEP scope risk
Payment flowOwner → GC → subsOwner pays each prime directly
Built forGeneral commercial paceCompressed hyperscale schedules (sub-year building delivery)

What should a GC do with this?

If your firm bids hyperscale data center work, don't assume the RFP looks like a normal CM-at-risk package. On Crusoe's jobs, the GC's value shifts from "I hold all the risk and coordinate everyone beneath me" to "I run the schedule and logistics while the owner carries separate direct relationships with electrical and mechanical." That's a different scope of work, a different fee structure, and a different insurance conversation — bid it as the coordination role it actually is, not a traditional at-risk GC contract with a bigger number attached.

For electrical and mechanical subs, this is the more interesting opening: a prime contract with the owner captures more of the value than a sub-tier position under a GC's markup, but it also means carrying your own schedule and payment risk directly with a customer who has never built a building before. Read the contract for what happens if the owner — not the GC — misses a decision deadline or changes scope mid-job.

What's the catch?

This structure isn't spreading to office buildings or schools next year. It exists because hyperscale AI data center owners are paying a premium for speed that a conventional GC-sub hierarchy, with its extra layer of coordination and markup, can't match at this pace. Crusoe's own project needed 25 people from four firms in one room whiteboarding the schedule before a shovel went in the ground — that's not a lighter lift than traditional delivery, just a differently distributed one. But the capital keeps flowing toward this category, and firms that understand the contract structure before they bid it will beat the ones still pricing it like a standard design-build job.

Construction AI Brief covered the credentialed labor shortage behind this same data center boom last week — this is the contract structure sitting on top of that labor question.

Friday one chart. Every week, one piece of data that should change a decision on your project. Subscribe at constructionaibrief.com.

Forward this to whoever preps your firm's hyperscale data center bids.

FAQCommon questions
How much is Crusoe raising and at what valuation?
Crusoe is in talks to raise about $3 billion in a round that would take its valuation to roughly $30 billion, up from the $10 billion-plus valuation set in its October 2025 Series E ($1.38 billion). Bloomberg reported the talks on July 2, 2026; terms and lead investors haven't been finalized.
What is Crusoe actually building?
Crusoe designs, finances, and builds AI data center campuses, with 4.9 gigawatts of capacity under contract and more than 40 gigawatts in its total development pipeline as of mid-2026. Its named customers include Oracle, Microsoft, Google, and Meta, with flagship campuses in Abilene and Amarillo, Texas.
What is a 'multi-prime' construction delivery model?
On Crusoe's Abilene campus, DPR Construction runs schedule, safety, and logistics like a traditional CM/GC, but electrical contractor Rosendin and mechanical contractor Southland hold their own prime contracts directly with the owner instead of subcontracting under DPR. Each prime carries its own scope, risk, and payment relationship with Crusoe.
How fast are these buildings actually going up?
DPR's team has delivered each 474,000-square-foot single-story building on the Abilene campus in under a year, using off-site prefabrication of skids and multi-trade racks to compress schedule. Crusoe expanded the campus from two buildings to eight mid-project and the team held the same pace.
Does multi-prime contracting apply outside AI data centers?
Not broadly yet. It's concentrated on hyperscale data center work where speed-to-energization justifies the extra coordination burden of managing multiple direct-to-owner contracts. GCs and MEP subs pursuing this category should expect it as the default structure, not an exception.
End of sheet — issue №050
Published · 2026.07.05
Project
Construction AI Brief
Dated
2026.07.05
Sheet
1 / 1
Rev
A
Published independently · constructionaibrief.com · © 2026Facebook·Privacy·About