Before discussing commitments, understand the meter. Claude usage in Microsoft Foundry is billed in Claude Consumption Units (CCU) through Azure Marketplace — a single billing dimension that replaced the earlier per-model token meters. Anthropic's pricing documentation states the CCU rate: $0.01 per CCU, so 100 CCU equals $1.00 of fees at standard per-model token rates after discounts. Token usage is converted to CCU using Anthropic's published per-model rates; Microsoft's own docs deliberately publish no per-token dollar prices and defer to Anthropic's rates and the Marketplace offer. CCU is metered hourly and invoiced monthly in arrears — pay-as-you-go, with no prepaid CCU credits or balances.
The commitment vehicle that clearly applies: MACC
The one interaction Microsoft documents explicitly is the Microsoft Azure Consumption Commitment (MACC). The CCU meter is MACC-eligible: Claude spend on Foundry decrements your MACC like other Azure Marketplace consumption. (CCU billed on other clouds does not.) For an enterprise that has signed a multi-year Azure consumption commitment, this matters twice over: Claude spend helps burn down a commitment you already owe Microsoft, and it does so without any extra negotiation. If your organization has a MACC, Foundry Claude spend counts toward it by default — confirm the current eligibility status of the meter with your account team, since eligibility is a property of the Marketplace offer.
Negotiated discounts: Azure Marketplace private offers
Committed-use discounts for Claude on cloud marketplaces are negotiated with the cloud provider, not Anthropic. On Foundry, the documented mechanism is the Azure Marketplace private offer: negotiated discounts are applied at the token-to-CCU conversion step and can carry different rates per model. In practice this means your discount shows up as fewer CCUs metered for the same tokens, not as a changed CCU price. Private offers are deal-specific — no public percentages exist, so treat any "typical discount" figure you hear as rumor and negotiate from your own volume data.
What about Azure Savings Plans and reservations?
Azure Savings Plans for compute and Azure reservations are commitment vehicles for Azure's own services. Claude on Foundry is a third-party Azure Marketplace offering from Anthropic, billed on a Marketplace meter — and neither Microsoft's Foundry billing documentation nor Anthropic's docs describe savings plans or reservations applying to the CCU meter. Do not assume marketplace spend gets savings-plan treatment; the documented levers for Claude specifically are MACC eligibility and private offers. If your FinOps model depends on a different answer, put the question to your Microsoft account team in writing before you commit the budget.
What to ask your Microsoft account team
1. Is the Claude CCU meter currently MACC-eligible under our agreement, and how will the spend appear in our commitment burn-down reporting?
2. At our projected volume (bring per-model token forecasts — the conversion to CCU uses per-model rates), what private-offer terms are available, and can rates differ by model to match our routing mix?
3. How do existing deployments transition? Deployments created before CCU became generally available continue on their previous per-model plan, while new deployments bill in CCU automatically — ask which of your deployments are on which plan.
4. What cost visibility should we build? Azure Cost Management shows a single CCU line; per-model token and request detail lives in the Foundry portal's Monitoring tab. Agree up front on how you will reconcile the two for chargeback.
5. If we split traffic between "Hosted on Azure" and "Hosted on Anthropic" deployment versions, does the commercial treatment differ?
Come armed with a baseline built from Anthropic's list prices — for example, a workload projected at 500 million input and 100 million output tokens per month on Claude Sonnet 5 at standard rates ($3/$15 per million) is $1,500 + $1,500 = $3,000 a month before any negotiated terms. A concrete number anchors the private-offer conversation far better than a growth story does.
Where to go next
For the billing plumbing in detail, see Foundry billing mechanics. The cross-platform negotiation picture is in committed-use discounts, and the reserve-or-not framework in on-demand vs reserved.