A factory in Liaoning rolls 13,676 robots out the door in a single year — a 65% jump over the year before. The same company's profit statement, however, tells the opposite story: it lost more money than ever. China's most established robot maker has run straight into a paradox the whole industry would rather not discuss.
Siasun Robotics (新松机器人), founded in 2000 and listed on the Shenzhen ChiNext board under ticker 300024, is the kind of company China's robotics story is supposed to celebrate: two decades old, deeply embedded in automotive, lithium-battery, semiconductor and transit supply chains, and a holder of more than 1,900 patents. Its 2025 annual report, filed with the exchange, lays out the contradiction in plain numbers.
The paradox in the numbers
Siasun's 2025 revenue came to RMB 4.122 billion (about $581 million / HK$4.49 billion), down a marginal 0.40% year-on-year. Its net profit attributable to shareholders was -RMB 398 million (roughly -$56 million / -HK$434 million) — an exact figure of -397,880,934.47 yuan disclosed in the board resolution. The loss widened 105% versus 2024.
And yet unit volume surged: about 13,676 robots sold in 2025, up 65%. Gross margin sat at 12.48%, down about 2.14 percentage points. In other words, Siasun sold far more machines and earned less money doing it — the "sell more, lose more" trap that haunts capital-intensive manufacturing when prices compress faster than costs.
Where the money goes
The report breaks revenue into segments, and only one grew. Automation assembly, inspection lines and system integration rose 16.27% to RMB 1.493 billion (around $210 million / HK$1.63 billion) and was the single pillar (支撑) of 2025 revenue. Everything else slipped: industrial robots and integration fell 8.77% to RMB 1.097 billion (about $155 million / HK$1.20 billion); logistics and warehouse automation dropped 6.66% to RMB 734 million (roughly $103 million / HK$800 million); transit automation fell 7.27% to RMB 261 million (about $37 million / HK$284 million); and semiconductor equipment declined 8.58% to RMB 525 million (around $74 million / HK$572 million).
The company attributed the squeeze to three forces: intensifying domestic competition, a weaker dollar late in the year that trimmed margins on U.S.-settled overseas work, and losses at associated companies counted by the equity method. Research spend held at about RMB 364 million (around $51 million / HK$397 million), 8.84% of revenue — a level Siasun argues is necessary but which, combined with falling margins, deepened the red ink.
Why 'selling more, losing more' happens
Siasun is not a startup burning venture cash; it is a profitable-era company sliding backward. The 2023–2025 arc is stark: revenue of RMB 3.967 billion, RMB 4.138 billion and RMB 4.122 billion, against net profit of RMB 49 million, -RMB 194 million and -RMB 398 million. The first quarter of 2026 offered no relief — revenue RMB 630 million (about $89 million / HK$687 million), down 8.42%, with a -RMB 93 million (roughly -$13 million / -HK$101 million) net loss.
The root issue is structural. Large automation projects carry long build, install and acceptance cycles, so revenue recognition lags. High inventory — about RMB 3.09 billion on the books — ties up cash and invites write-downs. And as new entrants flood robotics, price competition erodes the very margins that once made scale pay. Siasun's own line is that it is shifting from project-based integration toward standardized, productized robots; the 65% volume jump suggests the product push is working on quantity, not yet on margin.
The pivot to embodied AI
Even a company this rooted in industrial automation is now chasing embodied AI (具身智能). Siasun set up an embodied-intelligence institute in April 2025, showed a wheeled humanoid called SongYi (松羿) at the Global Industrial Internet Conference in July 2025, and by December 2025 had a half-body dual-arm humanoid in production-line verification. At the World Robot Conference in August 2026 it released OneHub (羿枢); the SongYi "power" version carries a 15 kg dual-arm load with a 0–1.8 meter working height.
Whether a legacy integrator can win the embodied-AI race against venture-backed specialists is an open question. Siasun's edge is claimed to be its decades of industrial-site data and field understanding — but turning that into a shippable humanoid is a different discipline from shipping an AGV.
Honest limitations
Every financial figure here — RMB 4.122 billion revenue, -RMB 398 million net loss (exact -397,880,934.47 yuan), 12.48% gross margin, 13,676 units (+65%), segment revenues, RMB 364 million R&D, and the Q1 2026 numbers — is drawn directly from Siasun's 2025 annual report and board resolutions filed with the Shenzhen Stock Exchange (szse.cn / cninfo.com.cn) and reported by China Business News and others. They are audited company filings, not estimates. The overseas share (~14.19%), 40-plus export countries, 4,000-plus customers and 45% Fortune 500 claim, and the patent count (>1,900, >1,400 inventions), are from a compiled corporate profile and should be read as company-stated. The embodied-AI product timeline (institute April 2025, SongYi July 2025, OneHub August 2026) is also company-disclosed. The "sell more, lose more" framing is an editorial reading of those audited numbers, not the company's own wording.
What readers can do now
- When a robotics company touts "unit shipments up 65%," check the margin line before cheering — Siasun shows volume growth and profit decline can happen at once.
- For investors, favor robotics firms that disclose audited gross margin and net profit, not just funding or demo videos; exchange filings (like Siasun's) are the harder, more honest source.
- Track whether legacy robot makers can actually ship embodied-AI humanoids at scale — the 2026 product launches are promises, and the 2025 financials are the reality check.