Crusoe began with an awkward contradiction.
In oil fields, natural gas was being burned because there was nowhere useful for it to go. Elsewhere, computers were consuming more and more electricity. Chase Lochmiller and Cully Cavness wondered what would happen if the computers went to the energy instead.
They founded Crusoe in 2018 and began converting stranded natural gas into electricity for mobile computing systems. Bitcoin mining provided an early workload; GPU computing followed. Years later, Crusoe sold the bitcoin-mining business and committed itself fully to AI infrastructure. The technology changed, but the original instinct survived: find the energy first, then bring compute to it.
That instinct has become considerably more valuable.
Crusoe has closed the initial $3.9 billion of a Series F at a $30.9 billion post-money valuation, with Atreides Management, Mubadala Capital and Valor Equity Partners co-leading the round. NVIDIA, Founders Fund, GIC, Qatar Investment Authority and others also participated. The company now reports more than $140 billion in total contracted value, over 6GW of gross contracted capacity, and 1GW already operational.
For Lochmiller, the strategy comes down to a remarkably physical description of an industry usually discussed in software terms:
“Getting there means controlling the infrastructure from electrons to tokens.”
That journey begins long before a model sees a prompt. It begins with land, substations, power plants, cooling equipment, switchgear and the months or years, it can take to connect them.
At Crusoe’s Abilene operation, the company says it delivered an on-site power plant and energized its first substation in under six months. Rather than treating electricity as a utility problem to solve after choosing a site, Crusoe develops power and compute together. Its current energy strategy spans renewables, batteries, natural gas, hydro, geothermal and newer technologies.
The same impatience with waiting shows up in manufacturing.
Crusoe makes power-distribution systems, electrical enclosures and industrial controls in-house. Its Crusoe Spark platform packages racks, power, cooling, fire suppression and monitoring into prefabricated modular AI facilities. The company says this approach can shrink parts of the construction process from years to weeks, while allowing customers to add capacity in smaller increments instead of waiting for an entire hyperscale campus.
Only then does Crusoe arrive at the layer most people associate with AI infrastructure: the cloud.
Its platform offers GPU capacity, training, fine-tuning and inference. Crusoe says Cloud bookings have grown more than 20-fold year over year so far in 2026, while Managed Inference has contracted more than $100 million in annual recurring revenue. Customers named by the company include Cognition, Figure and Perplexity.
Seen separately, those businesses can look unusually broad: an energy developer, a manufacturer, a data-center builder and a cloud provider under one roof.
Seen through Crusoe’s history, they are variations of the same problem.
Every handoff creates somewhere AI infrastructure can wait: for power approval, for electrical equipment, for construction, for GPU capacity. Crusoe is spending heavily to bring more of those dependencies inside the company.
Its board is beginning to reflect that ambition too. On the same day as the funding announcement, Crusoe added JB Straubel, founder and CEO of Redwood Materials and former Tesla CTO; Bill Stein, former Digital Realty CEO; and Thomas Seifert, Cloudflare’s CFO. The additions bring energy, data-center and cloud experience into the same governance structure.
JB Straubel framed the pressure plainly: “Grid demand is growing faster than infrastructure can keep up.”
That may be the simplest way to understand Crusoe’s $3.9 billion round.
Eight years ago, its founders went searching for computing workloads that could use stranded energy.
Now the direction has reversed.
AI has become the workload searching for energy and Crusoe is building as much of the road between the two as it can own.
