On September 24, 2026, Google Cloud updated the documentation for a billing feature most finance teams have never opened: the Incentives page. It gathers every conditional credit your organization has negotiated into one tracker, with progress bars, milestone dates, and a status on each one. On the surface it is a welcome piece of transparency. Read the eligibility rules behind it, though, and the page tells a different story: the credits reward the spend you were going to grow anyway, and they exclude the line item growing fastest on almost every enterprise bill.
If you sign Google Cloud contracts, this is worth ten minutes before your next renewal.
What the page actually tracks
The Incentives page surfaces three kinds of conditional credit, according to Google’s own documentation. Spend-based milestone credits pay out after you cross a threshold, the example Google gives is 25% credit for every dollar spent past a $100 milestone on a service. Incremental Workload Credits track your spend as it climbs across multiple milestones over time. And Rapid Migration and Modernization Program credits reward workloads you tag and move under a migration agreement.
The tracker itself is straightforward. Each program shows a milestone index, an end date, a spend target, live progress, and the reward attached. Milestones carry a status: active, achieved, upcoming, or missed. When the credit lands, it appears in your billing reports as a PROMOTION type credit, typically within 45 days of hitting the milestone.
That is genuinely useful. Before this existed, conditional credits lived in contract PDFs and account-team spreadsheets, and finance often could not tell whether a milestone had been met until the credit either showed up or didn’t. Putting it in the console is the right move.
The math on incremental workload credits
The published numbers on the migration side are specific enough to model. For General Workload Types, Google’s RaMP user guide sets Google Cloud Service Credits at 25% of incremental eligible spend, capped at the lesser of 30% of your Projected Annual Run Rate and $3M USD. Partner Services Funds run separately at the lesser of 20% of Projected Annual Run Rate and $2M. Credits disburse quarterly, within 45 days of each calendar quarter-end, starting with the quarter that ended June 30, 2026. The January 2026 launch blog adds a layer of Advanced Workload Credits for higher-effort migrations such as SAP, Oracle, VMware, and data analytics.
Twenty-five percent back on incremental spend reads like a strong number. The words doing the work are “incremental” and “eligible.” Incremental means the credit is calculated against growth over a baseline, so the program only pays when your bill goes up. Eligible is where the real filtering happens.
The exclusion list is the whole story
Read the eligibility rules and the same document lists what does not count toward incremental spend: Google Cloud Marketplace offerings, Google Cloud Support including Technical Support Services, subscriptions, Archival Storage, Google Distributed Cloud, and GPUs.
That last exclusion is the one to sit with. GPU spend is the fastest-growing line on enterprise cloud bills right now. In the State of FinOps 2026 report, 98% of the 1,192 respondents said they now manage AI spend, up from 31% two years earlier, and AI cost management ranked as the single most desired skill across organizations of every size. The workloads finance is most worried about are precisely the ones the credit program will not reward you for growing.
Stack those two facts together and the incentive structure points somewhere specific. The program pays you to grow general compute, storage that isn’t archival, databases, and networking. It pays you nothing to grow the GPU inference and training spend that is actually inflating your bill. So the credit nudges you to expand the predictable, already-discounted part of your estate, while the volatile part runs uncapped and unrewarded. You can hit every milestone on the tracker and still watch your real cost problem grow untouched.
There is also the commitment underneath. A credit calculated against incremental spend over a baseline is, functionally, a growth commitment. It pays best when you spend more with one provider, which is the definition of a lock-in lever dressed as a rebate. That is not a reason to refuse the credit. It is a reason to price it as what it is.
What to do before you count the credit
Treat the tracker as an input, not a scoreboard. A few practical moves:
Model the credit net of exclusions. Take your projected 12-month growth, strip out GPUs, Marketplace, support, subscriptions, archival storage, and Distributed Cloud, and apply the 25% only to what remains under the cap. That is the number to put in the forecast, not the headline rate. The gap between the two is usually large on an AI-heavy account.
Watch the baseline. Because credits pay on incremental spend, the Projected Annual Run Rate you agree to becomes the reference point for everything. A higher baseline quietly shrinks what counts as incremental. This is a negotiation term, not a fixed input.
Fold it into the commitment picture, not against it. Conditional credits sit on top of committed-use discounts and reservations, and the two interact. If you are already weighing reserved instances against savings plans or treating commitments as a capacity hedge, the incentive program is another variable in the same decision, not a separate win to bank.
Keep the exit math honest. A credit that rewards growth with one provider raises the cost of leaving it, the same dynamic that makes egress and switching terms matter more than their monthly line suggests. Value the flexibility you trade away, not just the credit you collect.
The Incentives page is a good thing. Any finance team managing a large Google Cloud contract should open it and reconcile it against what the account team promised. Just read it as a map of where the provider wants your spend to go, because that is what it is. The credits point away from the workloads that are actually growing, and a program you understand beats a rebate you assumed was free.