CUD negotiation tactics, the locks behind the discount
Committed use discounts reach deep rate cuts, up to 70 percent on eligible compute, and the deepest version carries the tightest lock: resource based CUDs bind to a region and a resource shape, and a workload that moves strands the commitment. The contract around the discount is negotiable, and most buyers never open it.
Prepared by Redress Compliance · August 6, 2026 · Google Cloud advisory. Based on 30 to 40 cost engagements led 2024 to 2026.
Executive summary
The rate rewards the lock. Resource based CUDs reach the deepest cuts, up to 70 percent on eligible compute at three years, because they commit to a specific amount of vCPU and memory in one region. Spend based CUDs flex across machine types at shallower rates.
The trade is explicit, and the negotiation is about how much lock the discount actually buys, because a workload that migrates region strands the resource commitment entirely.
The stranding is measurable. Across our engagements, 15 to 30 percent of resource based CUD value was stranded when workloads moved region inside the term, and the buyers who chose the resource version for its higher rate had never priced the region lock against their own migration roadmap.
The rate difference between resource and spend based coverage is the insurance premium for mobility, and it is usually cheap.
The stacking arithmetic misleads by default. Sustained use discounts apply automatically and separately, and teams that modeled sustained use and committed use as additive in the same hour overstated savings by 10 to 20 percent.
On committed resources the CUD replaces sustained use; the honest model prices the CUD against the post sustained use rate.
The EA layer is the negotiation most estates skip. An enterprise agreement with Google adds a negotiated discount layer above the public CUD rates, with the stack order, contract discount applied to the post CUD effective rate, written into the ordering document.
Estates that treated the public CUD rates as the whole game left the negotiated layer, and the flexibility clauses that travel with it, on the table.
The rate versus lock trade, priced honestly
| Resource based CUD | Spend based CUD | |
|---|---|---|
| The commitment | A quantity of vCPU and memory, in one region | An hourly dollar amount on a service |
| The rate | The deepest cuts, up to 70 percent at three years | Shallower, in exchange for flexibility |
| The lock | Region and resource shape: neither moves with the workload | Flexes across machine types within the service |
| The stranding mode | Migration, re-platforming, and region strategy changes | Spend falling below the committed hourly floor |
Price the lock against your own roadmap, not the discount table. A three year resource commitment on a workload with a two year region strategy is a decision to strand the third year at signature.
The rate premium of the resource version over the spend version is the exact price of immobility, and whether it is worth paying is answerable from your migration plan before signing.
The region lock, where value strands
Resource based CUDs cannot be moved: the commitment stays in its region billing on schedule while the workload that justified it runs somewhere else at on demand rates, the estate paying twice for one workload's compute.
The 15 to 30 percent stranding we measured came from ordinary events, region consolidations, latency driven moves, re-platforming to managed services, none of which anyone considered commitment relevant until the invoice said otherwise.
The mitigations are structural and cheap at signature: spend based coverage for any workload with mobility in its roadmap, shorter terms where the region strategy is unsettled, and the layered portfolio the CUD sizing guide works in detail, a resource based floor under the demonstrably immovable.
Spend based flexibility above it, nothing over the volatile edge.
The Google Cloud CUD negotiation playbook
The lock pricing method, the layered portfolio design, the sustained use stacking math, and the EA layer clauses that move the whole position above the public rates.
Get the white paper →The stacking truth, what adds and what replaces
Two stacking rules decide whether the savings model is honest.
Sustained use discounts apply automatically to qualifying workloads and are replaced, not joined, by CUDs on committed resources: modeling them as additive in the same hour overstated savings 10 to 20 percent in our engagements, flattering the CUD by a discount already in the bill.
The enterprise agreement layer runs the other way: it genuinely stacks, the negotiated percentage applying to the post CUD effective rate, provided the stack order is written into the ordering document rather than assumed.
That written stack order is worth several points of total spend, and it is the tactical difference between the estates that negotiate Google Cloud and the estates that merely buy it.
The wider agreement construction sits in the enterprise negotiation playbook, and the discount benchmarks show what the combined position reaches.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across cost engagements, 2024 to 2026
Across roughly 30 to 40 Google Cloud cost engagements Fredrik Filipsson led between 2024 and 2026, the recurring finding was buyers choosing resource based CUDs for the higher rate without pricing the region lock:
Commitments left billing in regions their workloads had departed, the lock never priced against the migration roadmap.
Savings models treating sustained use and committed use as additive on the same committed hour.
The third pattern was unused flexibility: estates that ignored spend based CUDs entirely, over committing on three year resource terms for the rate while their mobile workloads ran uncovered at on demand.
The corrective portfolio was never exotic, the floor and flex layering, timed against the negotiation leverage sequence, with the EA layer opened at the same table.
Your first five moves
- Price the lock against the roadmap: every resource based commitment tested against the region and re-platforming plans that outlive it.
- Model against the post sustained use rate, never list, so the CUD's contribution is real rather than double counted.
- Layer the portfolio: resource based floor under the immovable, spend based over the mobile, uncommitted over the volatile.
- Open the EA layer and write the stack order, negotiated discount applied to the post CUD rate, into the ordering document.
- Time the commitment to the leverage sequence, renewal, growth, and competitive moments, rather than the account team's quarter. The Google negotiation service runs the table with you.
Frequently asked questions
How much discount can Google Cloud CUDs reach?
Resource based CUDs reach up to 70 percent off eligible compute at three year terms, the deepest rates in exchange for the tightest lock: a commitment to specific vCPU and memory in one region. Spend based CUDs price shallower and flex across machine types, and the layered portfolio uses both.
What happens to a resource CUD if the workload moves region?
The commitment strands: it keeps billing in its region while the migrated workload pays on demand elsewhere, the estate paying twice.
Fifteen to 30 percent of resource CUD value stranded this way in our engagements, which is why every resource commitment should be priced against the migration roadmap before signature.
Do sustained use and committed use discounts stack?
No, on committed resources the CUD replaces sustained use, and modeling them as additive overstated savings 10 to 20 percent in our engagements. The honest model prices the CUD against your post sustained use effective rate, the bill you actually pay, not the list price.
Can we negotiate discounts beyond the public CUD rates?
Yes, that is the enterprise agreement layer: a negotiated discount above the public CUD rates, applied to the post CUD effective rate when the stack order is written into the ordering document. It is the negotiation most estates skip by treating the public rate card as the whole game.
Should we choose resource based or spend based CUDs?
Both, layered: resource based under workloads that demonstrably will not move, for the deeper rate, spend based over the mobile middle, for the flexibility, and nothing over the volatile edge. The single construct estates paid for the mismatch in one direction or the other in every review.
When should Google Cloud commitments be negotiated?
At the leverage moments, renewal, committed growth, and genuine competitive evaluation, with the CUD portfolio, the EA layer, and the stack order on the same table. Commitments signed off cycle, at the account team's quarter, price against your urgency instead of theirs.
