UBTECH's factory-humanoid revenue jumps 1,445% — the money is finally real, and so are the losses.

Hong Kong-listed UBTECH released its 2026 interim results on August 28, and for the first time the company's headline business is humanoids rather than the education and consumer robots it was built on. Group revenue for the first half doubled, rising 104.2% year on year to 1.269 billion yuan (about $188 million). The standout was the full-size embodied humanoid division: its revenue surged 1,445% to 590.3 million yuan (about $87.6 million) from just 38.2 million yuan a year earlier, and its share of total revenue leapt from 6.1% to 46.5% — formally eclipsing every other line to become UBTECH's top source of sales. The company sold 921 full-size humanoids in the half, up 1,946.7%, and 16,123 robots across all its humanoid categories. Because those big industrial machines carry higher margins, gross margin widened 9.7 points to 44.7%.

This is the counterweight to a summer dominated by valuations and sprint times. Where Unitree's listing and XPeng's $900 million round measured investor appetite, UBTECH's filing measures something harder — audited revenue booked from machines actually sold into factories. Its Walker S2 industrial humanoid and wheeled Cruzr Y1 have been tested or deployed in automotive manufacturing, logistics and aviation; Airbus acquired a Walker S2 this year as part of a partnership exploring humanoids in aircraft production. UBTECH has set a 2026 target of 10,000 full-size humanoids and is pushing to make the units repeatable, pairing them with an in-house embodied-AI stack it calls Thinker, plus deals with Siemens on manufacturing software and a chip joint venture with GPU developer MetaX. The direction is unambiguous: a Chinese maker turning humanoid demonstrations into an actual, fast-growing revenue stream.

The same filing is candid about how far that has to go. UBTECH is still losing money — its net loss narrowed 23% to 338.8 million yuan (about $50 million), and its adjusted EBITDA loss nearly halved to 174.1 million yuan, but profitability is not here yet — while research spending rose 38.9% to 303.1 million yuan, nearly a quarter of revenue. And the deployments remain early: UBTECH describes its industrial work as validating solutions for tasks like material handling, sorting and palletizing, and says it is only now trying to move 'from individual task validation toward larger deployments' in which several robots run longer sequences together on a line. In other words, the robots are selling, but they are still being introduced job by job rather than dropped in and trusted to run a shift. Its newer lines stretch the story further from the factory: a bipedal Walker C1 for commercial service and a U1 companion robot, whose consumer series has drawn more than 13,000 cumulative orders, are bets on markets UBTECH has barely begun to prove.

For a procurement team, UBTECH's earnings are the most useful kind of humanoid evidence — numbers filed under an exchange's rules, not a demo reel — and they cut both ways. They confirm that full-size industrial humanoids can now generate real, rapidly compounding revenue, which de-risks the category as a whole and gives buyers a shipping, publicly reporting vendor to benchmark against. But they also confirm the honest picture: a company that sold roughly 900 units in six months, still burns cash, and is deploying its robots task by task rather than at line scale. If you are weighing the Walker S2, press for exactly what the results do not disclose — a per-unit price, the autonomous success rate on your own tasks, how many robots one operator supervises, and dated multi-shift output at a named customer — and weigh the vendor's financial durability, because buying a humanoid from a loss-making maker is also a bet that it survives to profitability. One more constraint for US buyers: UBTECH is Chinese-made, so its robots are shut out of new US sale by the FCC's July import ban regardless of how fast the revenue climbs.