Crusoe is making a much bigger bet on the infrastructure behind the AI boom, raising $3.9 billion to expand large data centers while building smaller, transportable AI facilities.
Crusoe has raised $3.9 billion in a Series F funding round, taking the company’s valuation to $30.9 billion. The round was co-led by Atreides Management, Mubadala Capital and Valor Equity Partners, with participation from Founders Fund, GIC, Nvidia, Qatar Investment Authority, Radical Ventures and TPG.
The size of the raise highlights how much investor attention is shifting toward the infrastructure needed to power AI. Building and operating AI systems requires enormous amounts of computing capacity, electricity and data center infrastructure. Crusoe is positioning itself around that entire chain.
A $3.9 billion bet on AI infrastructure
The new funding gives Crusoe significant capital to expand its existing data center projects and develop a newer class of smaller, modular facilities.
One of the company’s major projects is a large data center campus in Abilene, Texas, which is being used by OpenAI.
But Crusoe is not putting all of its resources into giant campuses.
The company is also developing Spark, a modular data center concept designed to bring AI computing capacity closer to available sources of power.
These smaller facilities can be manufactured at Crusoe’s own facilities, transported by truck and connected to large power sources.
That approach could allow Crusoe to deploy computing infrastructure faster than traditional data center construction, which often requires significant construction work, large teams and lengthy development timelines.
The idea behind “AI factories”
The term “AI factory” is becoming increasingly relevant as the AI industry moves beyond training increasingly large models.
AI infrastructure now needs to support several different workloads, including:
- Training AI models
- Running AI models for users
- GPU-based cloud computing
- Large-scale data processing
- AI inference
Crusoe’s modular facilities are designed to provide computing capacity without necessarily requiring the same footprint as a massive conventional data center.
The company sees this flexibility as particularly valuable because access to electricity has become one of the biggest constraints on AI infrastructure expansion.
In simple terms, having GPUs is not enough if there is nowhere to power them.
That is where Crusoe’s strategy becomes interesting.
From wasted gas to AI computing
Crusoe’s history is also unusual.
The company was founded in 2018 as a crypto mining operation powered by flared natural gas. Instead of allowing natural gas that would otherwise be burned off at oil production sites to go to waste, Crusoe used it to generate electricity for computing.
As demand for AI computing exploded, the company shifted its focus toward AI infrastructure.
That transformation has taken Crusoe from a relatively niche energy and crypto business into one of the more closely watched companies in AI infrastructure.
Its customers now include Meta, Microsoft and Oracle.
The company’s evolution also reflects a broader change in the technology industry. Computing is increasingly tied to questions around electricity, power availability and where data centers can actually be built.
Crusoe has several ways to make money
Crusoe’s business is not dependent on a single revenue stream.
The company generates revenue through three main areas:
- Data center leasing: Customers bring their own GPUs and lease space in Crusoe’s facilities.
- GPU rentals: Crusoe provides customers with access to GPUs.
- AI inference: The company sells computing power used to run AI models.
This gives Crusoe exposure to different parts of the growing AI computing market.
It also means the company is not simply betting on data center construction. It is building a business around the computing capacity that sits inside those facilities.
The Jane Street deal adds another layer
Crusoe’s ambitions became even clearer earlier this month when the company signed a roughly $13 billion, five-year cloud contract with Jane Street, according to Bloomberg.
The deal involves supplying GPUs and AI infrastructure to the quantitative trading firm.
A contract of that scale demonstrates the growing appetite among companies outside traditional technology businesses for dedicated AI computing capacity.
Financial firms, software companies and other enterprises are increasingly looking for reliable access to large amounts of computing power.
For infrastructure providers such as Crusoe, that creates a potentially significant market beyond the traditional hyperscalers.
Why modular data centers matter
Building a conventional data center is a complicated process.
Developers need suitable land, electricity, cooling systems, network connectivity, construction capacity and local approvals. Finding enough power can be particularly difficult as AI data centers become larger and more energy intensive.
Crusoe’s modular approach attempts to tackle part of that problem.
Instead of constructing every facility from scratch at a single location, the company can manufacture smaller units and transport them to locations where suitable power is available.
That could make deployment more flexible.
There is another potential advantage: community opposition.
Large data center projects can face resistance from local residents concerned about electricity consumption, land use, water usage, noise and the physical size of the facilities.
Smaller facilities may not eliminate those concerns, but their size and deployment model could make it easier to build computing capacity in locations where a giant data center would be difficult to develop.
Investors are putting serious money behind Crusoe
The latest round is a major jump from Crusoe’s previous financing.
Just 10 months ago, the company raised $1.38 billion at a $10 billion valuation.
Its latest round values the company at $30.9 billion.
That rapid increase reflects the enormous amount of capital currently flowing toward AI infrastructure.
The investor list is also notable.
Nvidia participated in the round, alongside major institutional investors including Mubadala Capital, GIC and Qatar Investment Authority.
Nvidia’s involvement is particularly relevant because the company’s GPUs are central to the AI computing ecosystem that Crusoe is building around.
A possible IPO could be next
Crusoe is also being watched for another reason.
The company has reportedly been meeting with investment banks, including Goldman Sachs and Morgan Stanley, about a potential IPO.
There is no announced IPO date in the information provided, but the fundraising and growing customer base put Crusoe in a stronger position if it eventually decides to enter public markets.
A public listing would also give investors another way to gain exposure to the rapidly expanding AI infrastructure market.
The bigger picture
The AI race is often described in terms of models, chips and applications.
But underneath all of that sits a less visible requirement: massive amounts of infrastructure.
AI models need GPUs. GPUs need data centers. Data centers need electricity, cooling and connectivity.
That is why companies like Crusoe are becoming increasingly important to the AI ecosystem.
The company’s strategy is essentially to control more of that infrastructure stack, from the energy that powers computing to the physical facilities and GPUs that deliver computing capacity.
Its CEO, Chase Lochmiller, described the company’s ambition as controlling the infrastructure “from electrons to tokens.”
That captures the broader direction of Crusoe’s business.
The company started by turning otherwise wasted energy into computing power. Now it is betting billions of dollars on becoming a major infrastructure provider for the next phase of AI.
What to watch next
- Expansion of the Abilene, Texas data center
- Deployment of Crusoe’s modular Spark AI factories
- Demand for GPU cloud and AI inference
- The impact of electricity availability on new AI facilities
- Crusoe’s potential IPO plans
- Whether large AI infrastructure contracts continue at this scale
The bigger question for the industry is no longer simply how quickly AI models can improve. It is how quickly the physical infrastructure supporting them can be built.