A Verdict on Software, Delivered in Robots
Two "physical AI" rounds closed 24 hours apart. One company has delivered nothing. It raised four and a half times more.
Two "physical AI" rounds closed 24 hours apart. One company has delivered nothing. It raised four and a half times more. What happened 24 August — XPENG Robotics raised over $900 million at a post-money valuation above $6.3 billion , described by the company as the largest single-round private financing in China's embodied-AI industry. Led by IDG Capital, with Gaorong Ventures, and Tencent and Alibaba as strategic investors. 25 August — Gatik raised a $200 million Series D , led by the Qatar Investment Authority and Koch Disruptive Technologies , with Millennium Management, ARK Invest and Intact Private Capital participating. Both were reported under the same heading. They are not the same kind of company. XPENG Robotics Gatik --- --- --- Raised $900M $200M Valuation $6.3B post-money Undisclosed Customer deliveries Zero 85,000 driverless orders Operating today Nothing shipped 41 trucks , 250 stores Reliability — 99% on-time Revenue — $600M contracted Named customers — Walmart, Dollar General, Frito-Lay Next milestone Mass production expected end 2026 100+ trucks targeted end 2026 The part the headline leaves out Of that $900 million : - $600 million from external investors - $200 million from XPeng itself - $100 million from top executives (Round structure as reported.) A third of the largest embodied-AI round in Chinese history came from inside the company. There are two honest readings and we cannot separate them from outside. Conviction: carve-outs routinely involve parent participation, and executives committing $100M of personal capital is the alignment signal every governance textbook asks for. Price support: if external demand had cleared comfortably at $6.3 billion, a third of the round would not have needed to come from inside the building. What is not in dispute is the arithmetic. External capital bought roughly 9.5% of a company that has delivered nothing to a customer. Why Gatik's numbers mean something Fixed, repeated routes. That constraint is the technology. A route driven a thousand times has a bounded perception problem, enumerable edge cases, and an operational design domain small enough to genuinely validate. Gatik reports eighty-five thousand completed driverless orders at 99% on-time — company-disclosed and unaudited — across 41 trucks serving around 250 retail locations. That number is meaningful precisely because the routes repeat. And the 99% reliability figure means something because of its denominator — 99% over 85,000 is a claim; 99% over 40 would be a rounding artefact. A humanoid robot in unstructured environments has no such bound. General-purpose embodiment is a substantially harder problem and nobody has solved it at commercial scale. XPENG's own plan concedes this honestly: initial deployment at XPENG's own stores and campuses — controlled environments — before external delivery in 2027. These two problems are separated by years, plausibly a decade. The category label carries them in the same sentence. The quote t