Is the robot fleet actually funded? In Uber's case the 2020 answer was: funded, then divested. UATC, LLC's May 19, 2020 grant US10656657B2 claims object-motion prediction and autonomous control — a core, hard, expensive piece of the self-driving stack.
Read the burn behind the patent. Motion prediction sits at the center of an AV's planning loop, and the CPC tags confirm the depth: G05D 1/0246, G06K 9/00335 actor classification, G06K 9/6256 and G06K 9/6288 machine-learning detection, across a spread of G05D control classes. This is not a feature; it is a research program with payroll attached.
The cash-flow read is what makes this grant a Burn & Runway artifact. Uber's Advanced Technologies Group was consuming hundreds of millions per year, and patents like this are the capitalized residue of that spend. Months after this grant issued, Uber offloaded ATG to Aurora — converting an ongoing cash drain into an equity stake.
For a public-equities reader the lesson is that a divestiture does not erase the spend; it relocates it. The IP Uber generated, including this prediction patent, traveled with the unit. The burn was real; the exit was an admission that Uber could not fund the runway alone.
The honest limit: a single patent cannot quantify a unit's burn. It documents capability, not cost. But capability of this depth does not appear without a balance sheet behind it, and that is the inference the filing supports.
The takeaway for the money desk: when an autonomy unit is generating dense, central IP like motion prediction, ask who is funding the runway — and whether the answer is still the company whose name is on the patent.