Read the segment language, not the keynote. Nvidia's FY2021 annual report, filed February 26, 2021 and surfaced through SEC filings against the SEC 10-K, lists automotive AI cockpit, autonomous-driving development agreements, and autonomous-vehicle solutions as distinct components. The middle one is a window into the business model's current stage.

Development agreements are how a platform vendor monetizes autonomy before the cars carrying its chips reach mass production. Automakers pay Nvidia to co-develop and integrate; the volume silicon comes later, if and when those programs ship. For a money desk, that means today's autonomy line is partly a forward indicator — contracts that may or may not convert to scaled product revenue.

That is both reassuring and limiting. Reassuring, because design and development engagements are a real commercial signal that automakers are committing to Nvidia's stack. Limiting, because engagement revenue is lumpier and lower-volume than the production silicon the bull case ultimately depends on.

The arms-dealer thesis still holds: Nvidia profits whether or not any single automaker's autonomy program succeeds, as long as enough of them build on DRIVE. But the FY2021 disclosure is a useful corrective to anyone modeling automotive as if it were already a high-volume autonomy chip business. As of this filing, the development-agreement framing says it is earlier than that.

What to track: the migration from development agreements to shipped autonomous-vehicle solutions in production volumes. That transition — from co-development fees to per-vehicle silicon — is when the automotive line stops being a pipeline indicator and becomes a scaled revenue stream.

The honest limit: the filing names the categories but does not size them, so an outsider cannot say how much of automotive is development fees versus shipped product. What it does establish is the stage of the business — and the word the company chose, “agreements,” is the tell.