The fastest way to turn cloud enthusiasm into executive skepticism is simple: ship something impressive in Data or AI…and then hand Finance a bill no one can explain.
That’s not a tooling problem. It’s a structure problem.
In this post, I’m going to make the case for strong FinOps structures that actively engage Data, AI/ML, and the broader cloud stack—not as a “cost police” function, but as an operating model for technology value. We’ll look at why the scope of FinOps has expanded, what makes Data and AI spend uniquely tricky, and what “strong” actually looks like when it’s working.
FinOps has expanded beyond cloud cost optimization
FinOps is often introduced as cloud cost management, but the current definition is broader and more useful: it’s an operational framework and cultural practice designed to maximize business value from cloud and technology, enabling timely, data-driven decisions through collaboration between engineering, finance, and business teams.
That “and technology” matters. The FinOps Foundation’s 2025 framework revisions explicitly reflect a Cloud+ reality, adding Scopes as a core element to capture the different types of cost and usage data organizations now manage (SaaS, licensing, private cloud, data center, and more).
If you want a snapshot of how quickly this has shifted, the State of FinOps 2026 survey describes FinOps accelerating into a proactive, technology-wide discipline—moving “up and left” into future decisions, not just past reporting. It also reports that FinOps for AI is the top forward-looking priority, and that teams with stronger executive engagement have dramatically more influence over technology selection decisions.
This is the pivot: strong FinOps structures aren’t primarily about cheaper infrastructure. They’re about better decisions—earlier—across more technology categories.
Data and AI don’t behave like “normal cloud spend”
Traditional cloud optimization patterns (rightsizing VMs, commitments, storage tiering) still matter. But Data platforms and AI workloads introduce cost behaviors that expose weak FinOps operating models fast:
Data platforms create shared, multi-tenant cost puzzles. Warehouses, lakehouses, streaming, orchestration, and feature stores rarely map neatly to one app team. A single “bad” query pattern can become an organizational tax. Egress and replication can turn architecture choices into recurring spend.
AI adds volatility, speed, and new units of measure. GPU consumption, training runs, embedding pipelines, retrieval, evaluation, and inference don’t fit cleanly into monthly budget rhythms. Even when the underlying services are “managed,” the economics are still shaped by engineering choices: batch sizes, context windows, caching strategies, model selection, and when you run workloads.
This is why the FinOps Foundation treats FinOps for AI as a distinct scope: it calls out cost complexity, faster development cycles, spend unpredictability, and the need for a greater degree of policy and governance to support innovation—through allocation, forecasting, and optimization aligned to business value.
And crucially, AI spend doesn’t stay inside one bucket. The FinOps for AI guidance emphasizes that AI investments often transcend technology category boundaries—spanning data center, SaaS, enterprise agreements, specialized vendors, and multiple cloud providers.
If your FinOps structure only engages the “cloud infrastructure” team, you’ll always be late to the real conversation.
Strong FinOps starts with a data foundation—because governance is a data problem
When leaders say, “We need better cost control,” what they’re often missing is: we need better cost and usage data—timely, accurate, consistently modeled, and usable by the teams making the decisions.
The FinOps Framework updates even reflect this directly in its principles language, emphasizing that FinOps data must be accessible, timely, and accurate, and that FinOps should be enabled centrally (not controlled centrally).
This is also where standards are starting to matter more than dashboards.
The FinOps Open Cost and Usage Specification (FOCUS) is an open technical specification intended to create a common format and terminology for technology billing datasets produced by vendors and consumed by FinOps practitioners and tooling. The FOCUS project began in 2023, and the latest version is 1.3 (ratified in December 2025).
FOCUS is not just “nice to have” for multi-cloud reporting. It’s the kind of foundational layer you need if you want FinOps to scale into Data, AI, and SaaS without multiplying one-off pipelines and spreadsheets.
And the spec is evolving toward real operational pain points. FOCUS 1.3, for example, targets three persistent challenges: splitting shared resource costs, tracking contract commitments, and verifying data freshness/completeness—exactly the issues that tend to break allocation, forecasting, and month-end trust.
The practical implication is straightforward: if your FinOps structure doesn’t have a strong data engineering partnership (or capability), it will struggle to earn credibility with AI and Data teams, because the inputs won’t be trusted.
What “strong FinOps structure” looks like when Data and AI are in scope
“Strong” doesn’t mean “more meetings.” It means a clear operating model that makes good financial decisions easier for technical teams.
A strong FinOps structure tends to show up as:
- Central enablement with embedded engagement. A small core FinOps team sets standards, models, and governance—but Data Engineering, Platform, and ML teams have named counterparts (FinOps champions) who translate action into the backlog. This aligns with the shift toward enabling centrally rather than trying to enforce centrally.
- Scopes-based thinking. Instead of forcing every cost into the same playbook, you define what success looks like per scope (cloud platforms, data platforms, AI, SaaS) and then apply the right capabilities. The 2025 framework explicitly positions Scopes this way—as a roadmap for applying FinOps to segments of technology spend.
- Unit economics that technical teams recognize. Data and AI teams don’t think in “monthly spend by account.” They think in outcomes. FinOps should help translate cost into units such as cost per pipeline run, cost per dashboard refresh, cost per model training cycle, or cost per 1K tokens—then connect those units back to value. (The goal isn’t perfection; it’s decision-grade clarity.)
- Shift-left cost decisions. The State of FinOps 2026 highlights FinOps influencing future technology decisions “before commitments are made,” including interest in pre-deployment architecture costing. This is where FinOps becomes a product partner, not a postmortem narrator.
- A shared source of truth for cost and usage data. Whether you adopt FOCUS directly or use it as a target schema, the principle stands: normalize and govern billing/usage datasets so teams can allocate, forecast, and optimize with confidence across vendors and platforms. The FinOps Foundation explicitly positions FOCUS as a unifying format to reduce complexity and enable capabilities like allocation and unit economic reporting.
When these pieces are in place, FinOps engagement with Data and AI becomes tangible. You start seeing decisions like:
Data teams proactively tuning workloads because they can quantify the unit impact, not because they got yelled at after month-end.
ML teams choosing model architectures with an explicit cost/value trade-off, rather than treating GPU spend as “R&D overhead.”
Platform teams designing shared services (Kubernetes, databases, feature stores, vector databases) with cost allocation and transparency built in, because the FinOps structure made it part of “how we build,” not “how we explain.”
Closing thought: FinOps is how Cloud+ stays strategic
Here’s what we covered: FinOps has expanded into a Cloud+ discipline; Data and AI introduce cost behaviors that punish weak operating models; and strong FinOps structures are built on trustworthy cost/usage data, clear scopes, and real engagement with the teams shipping platforms, pipelines, and models.
If your FinOps practice still lives at the end of the month—after the spend has already happened—Data and AI will continue to feel like runaway line items. But when FinOps is structured to engage those domains early, you get something better than savings: you get control, predictability, and the ability to invest in the right things with confidence.
If you’re building or evolving a FinOps function right now, the next step isn’t “more reports.” It’s tightening the structure: define scopes, normalize the data, embed partnerships with Data and AI teams, and measure outcomes in units that connect technology choices to business value.
