Power is the new real estate. In the race to supply computing for artificial intelligence, the limiting factor is often not capital or chips but the ability to find electricity and put it to work quickly. On 17 September 2026, Denver-based Crusoe announced the initial closing of a $3.9 billion Series F that values the company at $30.9 billion.
The money is aimed at industrializing a different kind of data center: one built in a factory, loaded onto a truck, and delivered to wherever spare power already exists.
From stranded energy to AI campuses
Crusoe began by converting stranded natural gas into computing power. Over time, the company expanded into large fixed campuses and, more recently, into modular units it calls Spark. The Series F round, co-led by Atreides Management, Mubadala Capital and Valor Equity Partners, brings in a long list of additional backers, including Founders Fund, GIC, Nvidia, the Qatar Investment Authority, Radical Ventures and TPG. The company describes the round as oversubscribed.
The capital will support existing large projects, the expansion of Crusoe Cloud, and the scaling of Spark production. Crusoe reports more than $140 billion in total contracted value across its platform and more than 6 gigawatts of gross contracted capacity, of which roughly 1 gigawatt is already operational.
Crusoe’s $3.9 Billion Bet: The Spark idea
Traditional hyperscale data centers take years to permit, finance and construct. Crusoe’s Spark modules are designed to compress that timeline. Manufactured in the company’s own facilities, the units can be transported by road and connected to power sources in a matter of weeks rather than years. A plant outside Denver is planned to produce up to a gigawatt of Spark capacity annually. A second manufacturing site in Tulsa, Oklahoma, opened in mid-September 2026, adding hundreds of jobs and focusing on electrical components such as switchgear, controls, and busbars.
The modular approach also offers a possible answer to local resistance. Large fixed campuses often face community pushback over noise, water use, and visual impact. Smaller, relocatable units may prove easier to site, at least in some markets, because they can be placed closer to existing generation or industrial land rather than requiring entirely new power infrastructure.
Large campuses still matter
Spark does not replace the company’s big projects. Crusoe’s 1.2-gigawatt campus in Abilene, Texas, has already hosted training runs for OpenAI’s Astra model through Oracle. Nearby, the company is developing additional capacity for Microsoft. Reports also link Crusoe to a Google campus in Texas.
These fixed sites remain the backbone of contracted gigawatts. The modular line is intended to complement them, giving customers a faster path to incremental capacity when large campuses are delayed or fully booked.
Vertical integration as strategy
Crusoe’s pitch is “electrons to tokens.” The company develops power resources, builds the data centers that consume them, and operates cloud and managed-inference services on top. By making many of the electrical components in-house, it aims to control cost, quality and schedule more tightly than a pure developer that relies on external suppliers. The manufacturing facilities in Colorado and Oklahoma are concrete expressions of that ambition.
Cloud bookings have grown more than twenty-fold year on year, according to the company. Managed inference, launched late last year, has already accumulated more than $100 million in contracted annual recurring revenue. Named cloud customers include Cognition, Figure and Perplexity.
The numbers in context
A $30.9 billion valuation roughly triples the mark Crusoe reached in late 2025. The jump reflects both the intensity of demand for AI infrastructure and investor appetite for companies that can deliver capacity faster than the traditional construction cycle allows. At the same time, the gap between contracted and operational gigawatts remains wide. Delivering on the $140 billion contracted-value figure will require sustained execution across power procurement, manufacturing, logistics and customer delivery.
New board members announced alongside the round include Cloudflare CFO Thomas Seifert, infrastructure investor Bill Stein, and Redwood Materials founder JB Straubel. Straubel’s presence underscores the energy-storage and recycling dimension of Crusoe’s supply chain.
Speed as competitive advantage
In a market where GPU clusters can be ordered faster than the buildings and substations needed to power them, any method that shortens the path from capital to compute has strategic value. Factory-built modules that travel by truck are one such method. They do not eliminate the need for large campuses, but they create an additional tool for matching supply to demand in real time.
Crusoe is not the only company exploring modular or containerized computing. What distinguishes its current position is the combination of large contracted campuses, in-house manufacturing scale, and a cloud layer that turns physical capacity into a service. The Series F capital is intended to accelerate all three.
Looking forward
The next phase will test whether modular AI factories can move from promising prototypes to reliable, high-volume products. Success will depend on manufacturing yields, transportation logistics, interconnection timelines, and the willingness of customers to adopt a more distributed model of capacity. Failure modes are equally clear: delays in factory ramp-up, community resistance even to smaller sites, or an inability to secure power at the locations where modules can be delivered.
For now the funding gives Crusoe runway to pursue both paths at once: the multi-gigawatt campuses that define today’s AI map, and the truckable units that may define tomorrow’s. In an industry where the distance between an idea and a working cluster is measured in years of concrete and permits, the ability to ship computing capacity by road is a material advantage. The $3.9 billion is a bet that the industry will pay for that speed.
