The arms-dealer thesis for robotics is usually argued with rhetoric. It is cleaner to argue it with the XBRL numbers in the filing, retrieved through EdgarBeast and anchored to the FY2026 10-K at sec.gov. Nvidia's reported full-year revenue runs $26.97B (FY2023), $60.92B (FY2024), $130.50B (FY2025), and $215.94B (FY2026).
Now put R&D beside it. Research-and-development expense over the same fiscal years was $7.34B, $8.68B, $12.91B, and $18.50B. Revenue grew roughly eightfold; R&D grew about two-and-a-half-fold. That divergence is the entire operating-leverage story — the company is funding a widening research front out of a far faster-growing top line.
Why this matters for robotics specifically: the compute platform underneath AI training, autonomous-vehicle stacks, and embedded robotics is largely shared R&D. The same architecture work that serves datacenter AI also feeds DRIVE for vehicles and Jetson for embedded robots. Nvidia does not have to fund three separate moonshots; it funds one compute platform and points it at three markets.
That is why a robotics startup cannot easily out-spend Nvidia on the underlying compute. An $18.5B annual R&D budget, sized against $216B of revenue, is a structural advantage no single-application robot company can match. The boring incumbent here is Nvidia's income statement, and it quietly out-resources the field it equips.
The falsifiable version of the thesis is simple: if Nvidia's robotics ambitions were a distraction, you would expect R&D to balloon faster than revenue as it chased a new market. The sec.gov XBRL shows the opposite — revenue is outrunning R&D, which means robotics is being funded off operating leverage, not at its expense. Numbers indexed by EdgarBeast.