Microsoft and Google join an open push to free enterprise data from vendor lock-in

Microsoft and Google have joined more than 60 companies backing Apache Ossie, an open, vendor-neutral format that lets enterprises move not just data but the metrics, calculations, and AI context attached to it across analytics and AI platforms. The coalition now includes Databricks, Nvidia, Oracle, Salesforce, Snowflake, Informatica, and Mistral AI, a roster broad enough that staying out has become the louder signal.

The project, formerly called the Open Semantic Interchange, entered the Apache Incubator in June. The practical pitch: when the meaning layer of enterprise data is portable, businesses can change tools without rebuilding every calculation from scratch, and AI systems can carry that context across platforms. For an enterprise that has spent years encoding its revenue or churn definitions inside a single vendor's product, that portability is the difference between switching tools and rebuilding from zero.

The format itself is hub-and-spoke. A shared JSON and YAML schema describes datasets, fields, relationships, metrics, and AI context. Each vendor writes a converter that translates between the format and their own data model. Microsoft is building a two-way converter between Power BI, the analytics platform most Fortune 500 finance and operations teams run on, and the open format and is pushing for DAX, Power BI's calculation language, to be a recognized query dialect in the spec. The DAX push matters because Power BI semantic models are where most enterprise revenue and operations logic already lives. If DAX travels with the model, an enterprise does not have to re-author its calculations in SQL or another vendor's language to move the data.

The design choice that matters most is what does not get translated. According to the converter documentation, expressions stay in their authored dialect and are never machine-translated between SQL and DAX. Metrics without a DAX expression are skipped and reported. That preserves fidelity for the source system, but it means a Snowflake customer who wants to read Power BI calculations still needs a DAX evaluator on the other end. The format ships the calculation as written, not a translated copy.

Google is joining the project and BigQuery's SQL dialect is already on the supported list, per the InfoWorld report, which attributed Google's participation to an emailed company response. Google had not specified what it would contribute. The dialect-preservation design cuts both ways: it keeps Power BI and BigQuery behavior intact on their own turf, but the burden of teaching every other platform to read them falls on whoever writes each converter. BigQuery's presence in the supported dialect list makes that work concrete: any vendor writing an Ossie converter can ship BigQuery SQL verbatim and trust the evaluation to happen downstream.

The coalition has limits analysts have already flagged. Row-level security roles and other Power BI-specific constructs are preserved for round-trips in vendor extensions but are not represented for other consumers in the vendor-neutral model. The fetched core schema is labeled DRAFT and version 0.2.0.dev0, not a finalized release. The 60+ vendor list is a backing signal, not a delivery schedule. Governance gaps remain: who arbitrates spec changes, and what happens when two vendor extensions conflict.

The hub-and-spoke architecture means interoperability still depends on who shows up to build the spokes. If a vendor's converter lags, the lock-in the format is meant to dissolve stays in place. Industry coverage has framed backing the open format as meaningful alignment among competitors without guaranteeing vendor lock-in goes away.

Watch items: a finalized 0.2 release of the core schema, Google's first committed contribution, and how Snowflake, Databricks, and the rest of the coalition actually wire their converters into production. The current format, coalition, and converters show the meaning layer of enterprise data is becoming portable in principle. Whether the spokes land on schedule is the open question.