Microsoft’s About-Face on Semantic Layer Standards

When Snowflake launched the Open Semantic Interchange in September 2025, dozens of companies signed on. Microsoft wasn’t one of them. Power BI’s semantic models stayed in DAX, built for Power BI, while the rest of the industry worked on a shared way to describe a metric. 

This month, that changed. Microsoft is now listed on the Apache Ossie ecosystem page alongside Snowflake, Databricks, Salesforce, and AtScale. There’s a Microsoft converter in the Apache Ossie repository on GitHub that translates Power BI and Fabric semantic models into Ossie and back. And the Power BI team affirmed its “commitment to Apache Ossie“, promising to help customers “define semantic context once and reuse it across ecosystems without duplicating data or business logic.” 

The Information called it an about-face. When they asked me about it, I said Microsoft’s reversal is the latest sign that proprietary approaches to data management aren’t feasible.

“The future is going to require a heterogeneous stack that lets any database connect to any LLM or SaaS application.” 

Microsoft’s customers made this decision long before Microsoft did. 

Microsoft wants the standard to speak DAX 

Welcome to the club. The Power BI team also says it wants to establish “DAX as an Ossie-recognized query language.” Microsoft is joining the standard and bringing its own language along. 

AtScale already speaks DAX. Power BI users send DAX queries to AtScale today, and ACE, our AI Computation Engine, computes the answer on whatever data platform the customer picked. That matters most when the data lives in Snowflake or Databricks instead of Fabric. A DAX measure shouldn’t force you onto Microsoft’s compute to get the right number. 

Customers want freedom of choice 

Few businesses run on a single vendor anymore. Their data sits in Snowflake, Databricks, BigQuery, and Redshift. Their questions come from Claude, ChatGPT, and Cursor as often as from Power BI and Excel. A semantic model that only works inside one BI tool is invisible to every AI agent that never opens that tool. 

That’s what the open standard is for. It moved into the Apache Incubator this summer as Apache Ossie, and more than 50 organizations have signed on. Once the industry agrees on how to describe a metric, the holdout’s format is the one no other tool can read. Microsoft did that math. 

The hard part is doing the math under Ossie 

Ossie gives the industry a common way to write down what a metric means. Someone still has to compute it, the same way, every time, for every tool that asks. 

AtScale is a calculator for AI. ACE, our AI Computation Engine, takes a metric defined once and computes it identically whether the question comes from Claude, Cursor, Power BI, Excel, or Tableau, running on Snowflake, Databricks, BigQuery, or Redshift. 

We’re an Apache Ossie member because we want the standard to win, and we’ve served Power BI users for years. Neither will change. What’s gone is the argument that an open standard isn’t practical. 

Questions we’re hearing 

Holding out has an expiration date 

Microsoft spent a year on the sidelines of the open standard, then joined it. Every vendor still betting on a semantic model that only works inside its own walls should ask how long it can hold out. Customers have already decided: any database, any LLM, any application. Build for that, or explain to your customers why you didn’t.