By CrusoeLearning Objectives
- Understand what Crusoe is and how its "energy-first" model differs from a typical GPU cloud
- See where Crusoe Cloud fits among AI-infrastructure providers like CoreWeave, Baseten, and Together AI
- Recognize why power — not chips — has become the binding constraint on the AI buildout
What Is Crusoe Cloud?
Crusoe is a US AI-infrastructure company that builds and operates AI-optimized data centers and a managed GPU cloud, vertically integrated all the way down to the power that runs them. Crusoe Cloud is the product most teams interact with: on-demand and reserved access to NVIDIA accelerators for training and running large AI models, without having to build or lease data-center capacity yourself.
Founded in 2018 in Denver, Colorado by Chase Lochmiller (CEO) and Cully Cavness (COO), Crusoe started life capturing wasted flare gas at remote oil wells to power mobile data centers. It has since pivoted hard into AI, becoming one of the largest AI-native data-center developers in the United States and describing itself as an "energy-first AI factory company."
💡Key Concept
Energy-first infrastructure: Most cloud providers rent space and power from someone else and layer GPUs on top. Crusoe works the other way around — it secures the electricity first (through generation, substations, and long-term power deals), builds the data center around that power, then runs its cloud on top. In 2026, power availability, not chip supply, is often the real bottleneck for new AI capacity.
The Energy-First Model
Crusoe's pitch is that it owns the full stack of an AI data center:
- Power — procurement, on-site generation, and substation buildout, so a site's electricity is locked in before construction starts
- Data centers — high-density, liquid-cooled buildings designed specifically for large GPU clusters rather than general-purpose cloud
- Crusoe Cloud — the managed GPU platform that customers actually use, with NVIDIA hardware, high-speed networking, and storage tuned for AI workloads
Owning every layer lets Crusoe move faster on the thing that now gates the whole industry — getting gigawatts of power connected to the grid — and capture margin at each step instead of paying it to a landlord or a utility middleman.
Stargate and Major Deployments
Crusoe is the lead developer behind the Stargate campus in Abilene, Texas — a 1.2-gigawatt AI data center built for OpenAI alongside Oracle and SoftBank, and one of the most powerful AI facilities in the world. The company supplies compute capacity to customers including OpenAI, Oracle, Meta, and Microsoft, and says its total development pipeline now exceeds 40 gigawatts across current and planned sites.
That customer list is the clearest signal of Crusoe's position: the same frontier labs and hyperscalers that design their own chips still lean on specialist builders like Crusoe to get physical capacity online quickly. In mid-2026 Crusoe was reported to be in talks to raise about 3 billion dollars at a valuation near 30 billion dollars — roughly triple its 2025 level — one of several signs of how much private capital is flowing into the physical backbone of AI.
Crusoe Cloud vs. Competitors
| Provider | Model | Best For |
|---|---|---|
| Crusoe Cloud | Energy-first, owns power-to-cloud stack | Large reserved GPU capacity, new-build AI campuses |
| CoreWeave | GPU-specialized neocloud (public) | Large-scale training and inference contracts |
| Baseten | Inference-focused platform | Deploying and auto-scaling models in production |
| Together AI | Open-model cloud | Fine-tuning and serving open-weight models |
The "neocloud" category — GPU-first clouds built specifically for AI rather than general computing — has become one of the most contested layers of the AI market. Crusoe's differentiator is going furthest upstream, into power generation and data-center construction, rather than renting capacity and reselling GPUs.
Pricing
- Per-hour NVIDIA GPU instances
- No long-term commitment
- Spin up and down as needed
- Discounted committed GPU clusters
- Guaranteed availability
- For sustained training and inference
- Purpose-built data-center capacity
- Multi-megawatt to multi-gigawatt
- For hyperscalers and frontier labs
Crusoe Cloud is priced like enterprise infrastructure: on-demand rates for smaller workloads, discounted reserved contracts for sustained use, and bespoke arrangements for the largest customers who effectively co-develop entire campuses. Exact rates are quoted per deal, so treat any headline per-hour figure as a starting point rather than a fixed price.
Company Details
| Attribute | Detail |
|---|---|
| Founded | 2018, Denver, Colorado |
| Leadership | Chase Lochmiller (CEO), Cully Cavness (COO) |
| Type | Private |
| Key backers | Founders Fund, NVIDIA, and others |
| Flagship project | Stargate — 1.2-gigawatt OpenAI campus, Abilene, Texas |
| Notable customers | OpenAI, Oracle, Meta, Microsoft |
Strengths
- Owns the binding constraint. By securing power first, Crusoe can bring AI capacity online where GPU-only clouds stall waiting for grid connections.
- Blue-chip demand. Contracts with OpenAI, Oracle, Meta, and Microsoft validate both the engineering and the scale.
- Vertical integration. Controlling generation-to-cloud captures margin at every layer and speeds up delivery.
Limitations and Considerations
- Capital intensity. Building gigawatt-scale data centers requires enormous, ongoing project financing; the business is exposed to interest rates and to any slowdown in AI compute demand.
- Customer concentration. A pipeline anchored by a handful of frontier labs and hyperscalers means a single customer's roadmap shift can move the whole plan.
- Not a self-serve developer tool. Unlike a chat app or an API you sign up for in minutes, Crusoe Cloud is enterprise infrastructure — most value shows up at large, contracted scale.
Key Takeaways
- Crusoe is an energy-first AI-infrastructure company; Crusoe Cloud is its managed NVIDIA GPU platform, built on data centers it develops from the power up.
- It is the lead developer of Stargate, OpenAI's 1.2-gigawatt Abilene, Texas campus, and supplies compute to OpenAI, Oracle, Meta, and Microsoft, with a stated pipeline exceeding 40 gigawatts.
- Its rise — and a reported mid-2026 round of about 3 billion dollars near a 30 billion dollar valuation — reflects a broader truth of the AI boom: the scarce resource is increasingly power and physical data centers, not just chips or models.
