Yushu lays bare ugly quarter before IPO, bets on AI over hardware

The numbers

Yushu (宇树) has chosen candor over gloss. It reported Q1 revenue of ¥423 million, up 68.49% year‑on‑year, while adjusted net profit fell to ¥40.25 million, down 52.55%. R&D spend jumped by ¥38.32 million year‑on‑year and sales and brand‑promotion expenses surged after a Spring Festival Gala appearance — costs the company openly highlighted in its prospectus rather than burying. Wang Xingxing (王兴兴) reportedly controls 68.78% of voting rights, giving management latitude to court long‑term investors rather than short‑term momentum chasers.

Strategy: sacrifice margins to buy time for AI

Yushu is reframing itself from a robot seller to an AI infrastructure builder. The firm says the extra R&D is being directed to embodied intelligence large models, motion‑control algorithms and data loops — areas management argues are harder to copy than hardware. That is a deliberate bet: single‑unit margins on humanoid robots have been sliding as mid‑range models and price cuts flood the market, and Yushu’s own hardware gross margin pressure has accelerated despite group gross margins rising from 44.22% in 2023 to 60.13% through Q3 2025.

Sector and capital markets context

This is not just one company’s pivot — it signals an industry shift from spectacle to sustained operations. The humanoid robot field, fuelled by gala appearances, VC heat and subsidy money, is moving from scarcity pricing to scale competition. Yushu shipped roughly 5,500 humanoid units in 2025 and sits atop a crowded field that includes UBTECH (优必选), Xpeng’s (小鹏) unit Pengxing (鹏行) and, looming internationally, Tesla’s Optimus. Reportedly, Yushu’s IPO plan values the company at about ¥42 billion with a ¥4.2 billion raise; using disclosed 2025 adjusted profits implies a price/earnings multiple in the dozens — a tall ask if profits slip further.

Why it matters — and what to watch

Why lay out a “bad” quarter before going public? Because the company wants investors to reprice risk around building AI moats, not near‑term margin optics. Investors should watch order stability, cost control and real‑world deployments — can robots work eight hours reliably on a factory floor or autonomously navigate warehouses? Geopolitics matters too: trade tensions and export controls on advanced chips could constrain Chinese AI hardware timelines, making software and data advantages even more pivotal. Yushu’s gamble is clear: concede hardware profitability now to try to buy an AI lead that others cannot easily replicate.