Cloud vendors AWS, GCP, and Azure have converged over eight years on spend-based, flexible commitment models (Savings Plans, Flex CUDs) — making FinOps knowledge largely transferable across providers. AI vendors, however, have not converged: AWS Bedrock, Azure OpenAI PTUs, Google Vertex AI, Anthropic, and OpenAI each offer structurally different commitment vehicles with different units, terms, and lock-in profiles. Four of the five AI commitment products resemble the old RI model cloud providers are phasing out. OpenAI's Guaranteed Capacity is the exception, mirroring a Savings Plan structure. The piece argues that existing FinOps tooling is unprepared for AI commitment complexity, that cost-visibility gaps are the next major FinOps challenge, and that AI commitment optimization requires an entirely new skill set distinct from cloud compute optimization.
Table of contents
1. AWS, GCP, Azure: The Slow Convergence2. Then AI Showed Up3. The Comparison Nobody Has Made Yet4. Your FinOps Stack Wasn't Built for This5. What This Means For The Next 12 MonthsSort: