Section C · Operations

Product & Commercial

Crusoe Cloud is the customer-facing product. The commercial model is heavily weighted toward reserved capacity rather than on-demand. Pricing reflects the energy cost advantage but customer terms are typically negotiated, not list-priced.

Crusoe Cloud

Crusoe Cloud is the company's GPU-compute offering. The surface includes:

  • Bare-metal and virtual machine GPU compute on H100, H200, and Blackwell-generation hardware.
  • Managed Kubernetes for AI workloads.
  • Storage (object storage, parallel filesystem options for high-throughput training).
  • InfiniBand networking for distributed training.
  • Integration with NVIDIA's AI Enterprise stack and other software.

The product is enterprise-oriented. Self-serve is available but the bulk of commercial activity goes through sales conversations.

Contract shapes

Like CoreWeave, Crusoe's commercial model centers on reserved capacity:

  • Multi-year reservations for large customers (Stargate-scale and similar).
  • Annual reservations for mid-market customers.
  • On-demand pricing as a secondary offering.
  • Hybrid arrangements: dedicated capacity with overflow on-demand.

The reserved-capacity heavy mix matches the financial model — committed customer revenue supports the cap-ex needed for buildouts.

Pricing

Crusoe's pricing isn't fully public. Indicative pricing for current-generation hardware:

  • H100 on-demand: around $3-5/hour list, with reserved pricing meaningfully below.
  • H200: $4-6/hour list with reservations available.
  • Blackwell tier: bespoke pricing reflecting allocation scarcity.

The energy cost advantage doesn't fully pass through to list pricing — Crusoe captures some of the savings as margin. Customer-specific reserved contracts may bring effective pricing closer to or below CoreWeave's comparable offerings.

Customer segments

Who buys from Crusoe:

  • Frontier AI labs (Stargate-linked). The flagship relationships.
  • Other major AI companies. Companies doing meaningful training without being in the Stargate orbit.
  • Enterprise AI adopters. Mid-to-large enterprises with significant compute needs.
  • Some research and academic users. Smaller but present.

Crusoe doesn't actively serve the indie ML market. The product, sales motion, and contract minimums don't fit that customer profile.

Commercial partnerships

Several strategic partnerships shape Crusoe's commercial reach:

  • Oracle: Crusoe capacity surfaces through Oracle Cloud Infrastructure for some workloads. Particularly relevant for Stargate.
  • NVIDIA: Direct relationship for allocation, technical integration.
  • Energy sector partnerships: With oil-and-gas operators on site access; with renewable producers on power-purchase agreements.
  • Stargate consortium partners: OpenAI, SoftBank, MGX.

Takeaway

Crusoe's commercial model is concentrated, reserved-capacity-heavy, and structured around a small number of very large customers (Stargate-anchored). The energy advantage funds the commercial position. The next chapter looks at the underlying hardware infrastructure.