Unitree's 6x first-day pop split from China's robotics sector, which fell 8%

Unitree's first trading day on August 19, 2026, closed near 340 billion yuan in market cap (~$47 billion at prevailing 2026 rates), about six times the 61 billion yuan (~$8.5 billion) IPO issue valuation, per Sina Finance and the "What's Next 科技早知道" S10E26 podcast. During the session, intraday market cap touched around 440 billion yuan (~$61 billion) before pulling back. On the same day, a broad index of Chinese robotics stocks fell nearly 8%.

The company trades on Shanghai's STAR Market, the technology board of mainland China's main stock exchange, under the ticker 688836.SH. It was founded in Hangzhou in August 2016 by Wang Xingxing and built its early reputation on quadruped (four-legged) robots before moving into humanoids, per Huxiu.

Peers did not ride along. The 219x P/E at issue and the 440 billion yuan intraday cap put Unitree several multiples above its listed peers. The same session that lifted the leader dropped the index nearly 8% (Sina Finance).

Unitree's 2025 financials gave the bid a foundation. Revenue hit 16.99 billion yuan (~$236 million) with a 60.13% gross margin, up from 44.22% two years earlier. The first quarter of 2026 turned mixed: revenue grew 68% year over year to 4.23 billion yuan (~$59 million), while adjusted net profit fell 52.55% to 40.25 million yuan (~$5.6 million). R&D in the first half of 2026 hit 136 million yuan (~$19 million), 1.52x the prior-year period, with the spend steered toward what the company calls the robot's "brain" (embodied AI models, vision, and task planning). Sales expense in the first half of 2026 reached 164 million yuan (~$23 million), already past the full-year level for 2025 (Huxiu).

Episode 26 of "What's Next 科技早知道" returned repeatedly to a single question: who actually pays for a humanoid at scale. Host Diane (Ding) and guest Aaron (Zhou Jiuzhou), a former CICC and China Asset Management investor who now runs the "不止金钱" and "涉市未深" podcasts, walked through four paths: government and research procurement, already validated but with limited repeat orders; industrial automation and warehousing, the only one with proven unit economics in some sites; IP-driven experiences à la Disney's Imagineering and the Project Kiwi character, where the robot is a delivery vehicle rather than the product; and household companions, where form, safety, and regulation are still undefined.

Aaron reserved his sharpest argument for the form-factor question. He argued that Tesla's combination of in-house motor design, factory-floor training data, and a frontier model trained on its own production line is a closed loop that few Chinese players can replicate. Without a comparable training ground, the humanoid bet becomes a bet on whether a different loop (cheaper hardware, a broader customer base, or a non-Tesla platform) emerges first.

Unitree's capital structure tilts the bet toward one founder. Wang held 23.82% of shares before the IPO and controls 65.31% of voting power afterward through a dual-class share structure, in which founders hold shares with multiple votes per share. The setup lets the company keep funding long-horizon R&D; it also means public shareholders are underwriting a thesis whose execution depends on one set of calls (Sina Finance).

The 2024 funding round closed near 1 billion yuan (~$139 million), and a 2025 pre-IPO round valued the company at 12.7 billion yuan (~$1.76 billion) post-money. The IPO issue valuation near 61 billion yuan (~$8.5 billion) reset the reference point for the rest of the sector within five quarters (Huxiu).

The first post-IPO earnings report (Q3 2026, due in late October under A-share disclosure rules) will show whether the 60% gross margin can hold while R&D spend keeps doubling, and whether the 164 million yuan (~$23 million) in first-half sales expense reflects a demand pull or a one-time channel build. Until then, the market is pricing a category leader and a sector separately. The spread between them (6x on the leader, minus 8% on the field, on the same day) is the closest thing yet to a verdict on who actually pays for a humanoid.