FinOps for Data: The Bill Was Always Going to Arrive

For years the warehouse was sold as elastic, pay-for-what-you-use, gloriously cheap at rest. The 2026 State of FinOps survey is the part where the meter looks back.

The FinOps Foundation's 2026 State of FinOps report, published February 19, found that 98% of the 1,192 surveyed practitioners now manage AI spend, up from 31% two years earlier, across companies representing more than $83 billion in annual cloud spend. The report's quieter line is the one to sit with: many organizations say they are being asked to self-fund their AI investments through efficiency gains. The bill, in other words, is now the budget.

This matters because consumption pricing quietly inverted who holds the risk. In the seat-license era, you over-bought and the vendor's upside was capped. In the credit era, a forgotten join scales the vendor's revenue for you. Snowflake's own documentation notes that a statement runs by default for up to two days before it times out, and that resource monitors are not enabled until someone deliberately creates one. Nothing stops the meter except a person who remembered to.

What it reveals is that 'cheap to run' was never a property of the platform; it was a property of discipline you hadn't yet been billed for. The elasticity that absorbs a traffic spike also absorbs a clustering key nobody set, a dashboard refreshing every five minutes for an audience of zero, and a backfill that ran twice. Efficiency was always the actual product. The platform just unbundled it and sold it back as a quarterly initiative.

Watch whether 'FinOps for AI' becomes a standing function with a name on it or another dashboard that reports the spend without the authority to stop it. The tell is ownership: ask who can suspend a warehouse at 2 a.m. and whether they will be thanked or questioned for it. If the answer is a tool rather than a person, the bill is still in charge.