Snowflake has embedded AtScale as the only semantic layer in its platform, extending Semantic Views to Power BI and Excel, so they get the same accurate metrics your AI gets.
Snowflake, whose thousands of customers include Capital One, Toyota, and Uber, doesn’t add software to its platform casually. Snowflake chose AtScale as the only semantic layer in its platform. It’s a bet that business definitions are becoming as foundational to an enterprise data platform as storage or compute.
That means AtScale extends Snowflake Semantic Views to Power BI and Excel, through an XMLA endpoint, the tools where critical enterprise decisions get made, while embracing the Open Semantic Interchange (OSI) standard to keep those definitions portable and vendor-neutral. One governed semantic layer lives across every tool in the Snowflake stack.
Owning your semantic definitions is the difference between AI you can trust and AI that guesses
The value of a semantic layer starts with accuracy. According to the BIRD-SQL benchmark, the leading standard for text-to-SQL evaluation, AI agents querying raw data answer correctly only about 70% of the time. Ground those same agents in governed semantic definitions, and accuracy approaches 100%. This isn’t a smarter LLM. It’s better context.
Most enterprises attempt to harness semantics with yesterday’s enterprise data architecture. For example, when Power BI can’t connect to Snowflake, teams extract and mirror data, build caches, marts, and cubes, maintained by hand or automated ETL jobs. Each copy can introduce definition drift, until revenue means something different in every dashboard. And when an AI agent runs a query, it works from whichever copy it can reach, with no way to know which one is right.
Mirroring is how Microsoft moves your semantics, compute, and governance off the warehouse you own and into Microsoft Fabric while marketing that replication as “free.” It’s a Trojan Horse that hijacks your compute. Recently, Microsoft blocked Power BI from accessing Databricks semantics, as Kevin McLaughlin reported in The Information. We’ve written about this as Microsoft holding customers hostage.
A universal, standards-based approach to semantics gives you the freedom to choose your tools and own your semantics. But that’s just half of the story.
Accuracy and freedom matter. So does cost.
That ownership is exactly why metadata and catalog vendors are suddenly marketing themselves as semantic layers, or have rebranded themselves as “context layers.” But semantics is only half of the solution; the payoff comes when LLMs use those semantics to compute answers
AtScale helps you capture semantics and compute answers. Once you define the definition of “revenue,” AtScale optimizes how to calculate it, and does the job for any LLM, BI tool, or Excel that needs that definition. One Tier 1 multinational bank measured the impact of the AtScale computation optimization layer. In a benchmark of five production business question workloads, the AtScale semantic layer resulted in 21,903x fewer bytes billed, 11,425x less compute, and query costs dropped from $17.93 to $0.0008, as compared to unguided or metadata-guided queries.
The waste comes when AI re-derives metrics your organization settled years ago. Those answers should live in the semantic layer, not get recalculated on every query, and not get inferred from the abstract map a metadata or context layer provides.
The bank projected the math: unguided and metadata-only approaches cost $9 million a year in wasteful LLM thrashing and guessing. A semantic layer that takes charge of computing the answer not only increases accuracy, it reduces cost.
Snowflake + AtScale: where 1+1=5
The partnership math is simple: 1+1=5. The new XMLA endpoint that’s built into Semantic Views is a Snowflake native way for Snowflake’s customers to make their business semantics directly accessible to millions of Power BI and Excel users without surrendering ownership of their semantic definitions.
It’s called Snowflake Semantic Views XMLA Endpoint, Powered by AtScale. This product connects Power BI and Excel directly to Snowflake Semantic Views, so governed metric definitions reach every tool without a copy in between. Autonomous in-warehouse aggregates handle the heavy lifting, so dashboards that used to time out on hundred-million-row fact tables now return in under a second.
The XMLA endpoint is just the start. As your semantics grow, the AtScale / Snowflake solution grows with you. AtScale’s MCP tooling helps any LLM use the same semantic definitions. AtScale Enterprise connects Tableau to Snowflake and reaches other Iceberg compute loads. The Snowflake / AtScale partnership delivers. For more, visit snowflake.atscale.com.
Your AI agents, your BI tools, and your CFO’s spreadsheet all get the same numbers, with no changes to how you work.
AtScale extends Snowflake Semantic Views to Power BI and Excel. But more than the technical connection, this partnership is based on a shared belief that customers should have full agency over how they define and manage the semantics of their business. You should have the freedom to choose, control your semantics, control your costs, and control the accuracy of your analytics.
SHARE
Guide: How to Choose a Semantic Layer