There's a new line on your Snowflake bill, and it's growing faster than anything else on it.
Actually, there are two: AI Services and AI Inference. If your data team has adopted Snowflake Cortex AI, that's where every Cortex Analyst query, every Cortex Search call, every fine-tuning run, and every token your agents burn ends up. Two service types. No teams, no models, no use cases. Just a number that was small last quarter and isn't anymore.
We built full cost visibility for Cortex because we kept seeing the same conversation play out: data teams love it, finance can't explain it, and nobody can answer the only question that matters- which team, model, or use case is driving this?
Cortex Isn't a Side Feature Anymore. It's The Strategy.
If you think of Cortex as an experiment running in some corner of your data platform, Snowflake's own numbers say otherwise. On their most recent earnings call, Snowflake reported more than 7,100 accounts using Cortex Code- one of the fastest-adopted products in the company's histor, and accounts on Snowflake Intelligence more than doubling quarter over quarter. A quarter earlier, they counted over 9,100 accounts using their AI offerings.
Their CEO isn't subtle about where this is going. Sridhar Ramaswamy called the last quarter "a clear inflection point" and describes Snowflake's ambition as becoming "the control plane for the Agentic Enterprise"- a message he repeats everywhere from earnings calls to his LinkedIn feed.
Here's what that means for anyone who owns a cloud budget: AI on the data platform is no longer a rounding error. It's the fastest-growing, least-explained line on the data bill. And unlike warehouse compute — which the FinOps world has spent years learning to attribute and optimize — Cortex spend arrives with almost no built-in explainability. Snowflake meters it in credits, rolls it into two service types, and moves on.
How Finout Breaks It Down
Finout decodes those two line items completely, down to the level FinOps teams actually work at.
Every Cortex service is covered: Cortex AI Functions (formerly Cortex AISQL), Cortex Analyst, Cortex Search, Cortex Agents, Snowflake CoCo (Cortex Code), Document AI, Cortex Fine-tuning, and Provisioned Throughput. You can group and filter Cortex cost by service, by model (yes, you can see exactly what claude-sonnet is costing you), by function type, by warehouse, and by query type. Credits are converted to real dollars.
And this is the part we're proudest of: the breakdown reconciles exactly to your Snowflake invoice. Every enriched row sums back to the AI Services and AI Inference totals in Snowflake's own billing data. No dropped rows, no double counting, no "directionally correct" estimates. That's possible because it's built on Finout's org-level Snowflake integration, which sources the dollars Snowflake actually invoices you- not credits multiplied by a rate someone typed into a settings page.
This matters because most AI cost visibility today stops at the category level: the bill stops saying "compute" and starts saying "Agents" or "Search." That's labeling, not answers. A category tells you what kind of workload spent the money- it can't tell you which model, which function, or which team, and it makes no promise of matching what Snowflake actually charges you. Finout's Cortex breakdown does both: model- and function-level granularity across every Cortex service, reconciled to the invoice cent for cent. It's the difference between a dashboard you glance at and a number you can put in front of your CFO.
Because Cortex costs behave like any other cost in Finout, everything else just works on top: anomaly detection catches a model or use case that spikes overnight, Virtual Tags allocate Cortex spend to the teams driving it, and showback and chargeback flow through your existing dashboards. No new tooling, no extra permissions beyond the standard Finout Snowflake role.
Why This Matters Now
Every platform vendor is racing to embed AI, and every one of them is metering it in their own units- credits, tokens, capacity hours. The bills arrive aggregated; the questions arrive specific. The gap between those two is where budgets quietly break.
Snowflake is simply the most important place this is happening right now, because Cortex sits directly on top of the data most enterprises care about most. Adoption is compounding, the CEO is calling it an inflection point, and the spend is following. The organizations that build attribution muscle for AI services today will be the ones that get to say yes to more AI tomorrow, because they can explain what it costs.
Your team turned on Cortex. Now you can attribute every dollar of it, by service, by model, by team- and reconcile it back to the invoice.
Check out our documentation to learn more, or book a demo with our team to see it live in the platform.
cloud & AI spend
.svg)
