Product & Commercial

Product surface

CoreWeave's product is GPU compute in cluster configurations. The platform surface includes:

• Virtual servers and bare metal: Direct GPU access in dedicated configurations.
• Kubernetes: Managed Kubernetes for AI workloads. CoreWeave has been notably Kubernetes-forward.
• Storage: High-performance object and block storage tuned for AI workloads.
• Networking: The InfiniBand fabric described in chapter 02.
• NVIDIA-stack integrations: Base Command, Magnum IO, NVIDIA AI Enterprise.

The product is meant for customers operating their own ML / AI workflows at scale, not for indie developers spinning up a quick experiment.

Contract shapes

CoreWeave's commercial model centers on contracted capacity rather than pure on-demand:

• Multi-year capacity reservations. The bulk of revenue. Customer commits to N GPUs for M years; CoreWeave commits to deliver the capacity.
• Annual reserved. Smaller customers; one-year commitments with optional renewal.
• On-demand. Available but secondary; pricing reflects the lack of commitment.
• Bespoke commercial structures. Large customers (the Microsoft / Meta-tier) often have custom contracts with unique pricing, capacity guarantees, and termination protections.

The revenue mix is dominated by multi-year reservations — visibility into 3+ years of revenue is a meaningful financial profile feature versus competitors with shorter contracted books.

Pricing

CoreWeave's published list prices are generally not the prices customers actually pay. Indicative list pricing:

Reserved pricing at scale can be 40-60% below the on-demand rate. Strategic accounts (Microsoft, Meta) get further substantial discounts due to volume. This is normal for enterprise compute and matches the patterns at hyperscalers.

Reserved capacity

Reserved capacity is the strategic product. The mechanics:

• Customer commits to a fixed number of GPUs for a fixed term (typically 1-5+ years).
• Customer pays for the capacity regardless of actual utilization.
• CoreWeave delivers the capacity by procuring and deploying the GPUs against the commitment.
• Pricing typically includes options for renewal, upgrade to newer hardware generations, and limited flexibility on configuration.

This is the contract shape that supports CoreWeave's debt-financed growth model. Committed customer revenue is borrowable; CoreWeave issues debt against the committed receivables to fund the GPU purchases.

Enterprise sales motion

CoreWeave has built an enterprise sales organization commensurate with the deal sizes:

• Strategic account managers focused on the very largest customers.
• Sales engineers who can architect cluster configurations.
• Customer success teams for ongoing operational engagement.
• Solution architects who help customers design their training and inference pipelines on CoreWeave infrastructure.

The sales motion looks like a hyperscaler's, not a marketplace's. The deal cycle is months-to-quarters, not minutes.

SLAs and operational guarantees

Enterprise customers expect:

• Uptime SLAs (typically 99.9%+).
• Performance guarantees on specific workload classes.
• Operational hand-holding during cluster bring-up.
• Incident response SLAs.
• Compliance attestations (SOC 2 type II at minimum; additional certifications for specific industries).

CoreWeave delivers these. The operational quality has been validated by sustained relationships with the most demanding customers in the industry.

Takeaway

CoreWeave's product and commercial model is structured around large reserved-capacity contracts, with a sales motion and operational guarantees that match enterprise expectations. The model is the inverse of a marketplace — concentrated, committed, deeply integrated. The next chapter examines the customer base.