Section B · Marquee relationships

The Microsoft Deal

CoreWeave has reported $10B+ in committed capacity sold to Microsoft over multiple years. The deal is the single most important commercial relationship in the neocloud category and shapes everything about how to evaluate CoreWeave.

Deal shape

Public disclosures and reporting describe the Microsoft relationship in broad terms:

  • Multi-year capacity reservation.
  • Aggregate contract value reported in the $10B+ range across the term.
  • Dedicated GPU capacity in CoreWeave datacenters, accessible to Microsoft's workloads.
  • Bespoke pricing and operational terms — not at list rates.
  • Tight technical integration: CoreWeave capacity appears in Microsoft's Azure AI footprint indirectly.

The exact split between training and inference workloads, the specific GPU mix, and the precise contract structure are not fully public.

Why Microsoft buys from CoreWeave

Microsoft is the parent of Azure — itself a hyperscaler with massive GPU infrastructure. Why would they buy capacity externally from CoreWeave?

  1. Speed. Building GPU capacity from scratch takes 18-24 months. Buying it from CoreWeave who already has it lets Microsoft serve customers (especially OpenAI and partners) sooner.
  2. Capital structure. CoreWeave's debt-financed GPU model lets Microsoft access capacity without the full cap-ex impact on Microsoft's balance sheet. CoreWeave absorbs the GPU depreciation risk.
  3. Allocation. NVIDIA allocations are scarce; CoreWeave's existing allocation gives Microsoft GPUs they couldn't buy themselves quickly.
  4. Strategic optionality. Buying capacity creates a relationship with another GPU-cloud player; reduces concentration on Microsoft's own datacenter build pace.
  5. OpenAI relationship. Microsoft's OpenAI commitments require massive compute; sourcing from CoreWeave was a faster path to meeting those commitments than waiting for Azure's own builds.

Strategic implications

The Microsoft deal validates several things about CoreWeave's strategic position:

  • CoreWeave's operational quality is competitive with hyperscaler-internal capacity — Microsoft wouldn't buy capacity they didn't trust.
  • The capital-intensive neocloud model has a customer base — at least one hyperscaler is willing to commit billions to it.
  • Hyperscaler-as-customer is a real business model. CoreWeave isn't a hyperscaler competitor in this lens; it's a hyperscaler supplier.
  • CoreWeave's NVIDIA relationship is meaningful enough that Microsoft can't easily replicate the allocation independently.

The framing matters. CoreWeave isn't primarily competing with Microsoft for end customers; it's selling capacity to Microsoft as one of several wholesale customers.

Financial impact

The Microsoft deal is the single largest revenue line in CoreWeave's customer mix. Industry estimates suggest it represents a high single-digit to low double-digit percentage of CoreWeave's annual revenue at the multi-year contract's average run-rate.

The financial dynamics:

  • Reserved capacity means revenue is committed regardless of actual usage. This is good for forecasting.
  • The deal supports debt financing — committed revenue can be borrowed against.
  • Margins on the deal are likely lower than CoreWeave's blended portfolio (large customers get volume discounts) but absolute dollars are very large.
  • The deal absorbed a significant fraction of CoreWeave's H100 deployment in 2023-2024.

Customer concentration risk

The Microsoft deal also creates concentration risk. If a meaningful share of CoreWeave revenue is from one customer:

  • Renewal at contract end is a binary outcome.
  • The customer has negotiating leverage at renewal.
  • Strategic shifts at the customer (Microsoft accelerating its own datacenter builds; the OpenAI relationship changing) propagate to CoreWeave.
  • Public investors discount CoreWeave's revenue multiples to reflect this risk.

CoreWeave's response has been to diversify customer commitments (the Meta deal in chapter 04 is part of this) and to pursue other strategic relationships. But the concentration is real for now.

Evolution of the relationship

The Microsoft-CoreWeave relationship has evolved through several phases:

  1. Early 2023: initial commitments related to OpenAI compute demand.
  2. 2023-2024: scaling commitments as ChatGPT and the broader AI surge created sustained demand.
  3. 2024-2025: extended terms, additional capacity, strengthening operational integration.
  4. 2025-2026: continued execution; some renegotiation as Microsoft's own capacity buildouts come online.

The relationship is mature enough that both sides know the operational rhythm. The strategic question is whether Microsoft's own internal capacity eventually displaces CoreWeave commitments, and on what timeline.

Takeaway

The Microsoft deal is the foundational commercial fact about CoreWeave. It validates the company strategically, drives a meaningful share of revenue, and creates both visibility and risk. The next chapter covers the Meta deal — the second major customer commitment that has reshaped the company's revenue base.