1. The Inspection Event: What Happened and Why

On 1 April 2026, the China Securities Association of China (CSAC) published its second batch of 2026 on-site inspection targets for IPO applicants. Under the direct observation of regulators, self-regulatory organisations, industry representatives, and media, two companies were randomly selected: Unitree Robotics and CASIC Space. Both had filed for STAR Market listings within the preceding two weeks.

Unitree's application — seeking to raise RMB 4.2 billion (approximately USD 610 million) — was formally accepted by the Shanghai Stock Exchange on 20 March 2026. It was among the first to undergo the exchange's new "preliminary review" mechanism. CASIC Space, a commercial rocket company, was accepted on 31 March and selected the very next day. Together, the two represent a statistically expected outcome: of the six STAR Market applications received in Q1 2026, two — or roughly 33% — were drawn under the CSAC's standard 20–33% random sampling rate.

Unitree's selection is therefore not inherently anomalous. However, given the company's profile — global humanoid robot shipment leader, profitability claims rare in its sector, a Spring Festival Gala performance watched by hundreds of millions — the selection has attracted outsized attention from media, investors, and peer companies alike.

Table 1: Key Facts at a Glance

2. Regulatory Context: From 'Death Warrant' to Routine Compliance Check

The market reaction to Unitree's selection cannot be understood without acknowledging the historical connotation of IPO on-site inspections in China. Between 2021 and 2024, companies selected for inspection withdrew at alarming rates — 71.74% in 2021, rising to 82.35% in 2023. The colloquial label "death warrant" was apt: being drawn was widely interpreted as a signal of pre-existing regulatory concern, and companies often withdrew preemptively rather than face scrutiny.

That paradigm has materially changed. The 2025 regulatory reform package introduced several structural shifts. The "file and be accountable" principle now prohibits companies from withdrawing applications after receiving an inspection notice without facing regulatory consequences. A tiered-response framework now distinguishes between minor disclosure deficiencies — which trigger remediation — and substantive violations, which lead to direct rejection. Crucially, frontier sectors including humanoid robotics and commercial aerospace now receive differentiated treatment that balances innovation-sector tolerance with financial disclosure rigour.

The results bear out the shift. Of 31 companies selected for inspection between March 2024 and April 2026, only one — Xinmi Technology, a semiconductor materials firm — ultimately withdrew its application, and that withdrawal occurred nine months after selection. Eight of the 31 selected companies have successfully listed. As of April 2026, all 13 companies drawn in the current calendar year remain in normal review.

The statistical logic also matters. In Q1 2026, only 11 new IPO applications were received across all A-share boards, six of which were for the STAR Market. Applying a 33% draw rate to that pool mathematically yields approximately two selections. Unitree and CASIC Space being drawn simultaneously is therefore consistent with standard process, not indicative of targeted regulatory concern about either company specifically.

3. Unitree's Financial and Operational Profile

3.1 Headline Metrics

Unitree's 363-page prospectus presents a financial narrative that stands apart from most peers in the humanoid robotics sector. The majority of competitors globally are loss-making, capital-intensive, and dependent on equity fundraising to fund operations. Unitree's reported figures are exceptional by comparison.

Table 2: Unitree Financial Summary (RMB millions)

3.2 Product and Market Position

Founded in 2016 in Hangzhou by CEO Wang Xingxing, Unitree built its initial business around quadruped robots. By 2024, it held approximately 70% of global quadruped unit sales, shipping over 23,700 units that year across the Go, A, and B series — a range spanning from the USD 1,600 Go2 consumer model to the industrial B2-W variant. Boston Dynamics' Spot is priced at USD 74,500; Unitree's closest comparable performs roughly 80% of Spot's capability at approximately 2% of the cost.

On humanoids, Unitree shipped over 5,500 units in 2025, capturing 32.4% of the global humanoid market — more units than any other manufacturer, including Tesla Optimus (whose shipment data remains undisclosed). The product range spans the R1 (USD 5,900, entry-level), G1 (USD 13,500–27,000, mid-range research), and H2 (USD 29,900, full-size industrial-grade flagship). Humanoid revenue surpassed quadruped revenue for the first time in 2025, reaching 51.5% of core business revenue.

Key investors include Alibaba, Tencent, China Mobile, Geely Capital, Ant Group, Jinqiu Capital (ByteDance), and HongShan Capital (formerly Sequoia China), reflecting both strategic and financial validation from China's largest technology ecosystem participants.

4. Four Controversies the Inspection Will Likely Examine

Regardless of the inspection outcome, four structural issues in Unitree's prospectus have generated substantive debate among analysts, investors, and competing IPO candidates. These questions are not unique to Unitree — they represent the core set of issues that any humanoid robot company pursuing a public listing will need to address credibly.

4.1 R&D Intensity vs. Hard-Tech Classification

The STAR Market's "hard-tech" classification is its foundational listing criterion, designed to attract companies with genuine scientific innovation rather than commercial scale alone. R&D expenditure as a share of revenue is among the most scrutinised indicators. Unitree's R&D-to-revenue ratio has declined from approximately 21% in 2022 to 7.7% in the first nine months of 2025 — a significant trajectory mismatch with its hard-tech positioning claims.

For context, UBTech's R&D expenditure exceeded RMB 500 million in 2025, representing 25.4% of revenue, with cumulative four-year R&D investment of RMB 1.9 billion. Some commentators have stated bluntly that Unitree's R&D profile is difficult to reconcile with a hard-tech IPO narrative.

The counterargument — and it deserves serious consideration — is that Unitree's technology model is fundamentally different from a pure research company. The firm claims self-developed component ratios exceeding 70%, spanning motors, reducers, and controllers. Its technology pathway is one of engineering iteration at scale: using mass-production feedback loops to drive capability improvement rather than basic research investment. If regulators accept this model, it opens a clearer listing pathway for hardware-first robot manufacturers. If they do not, companies with lower R&D intensity relative to revenue will face a materially higher compliance burden on the path to listing.

4.2 Revenue Structure: Research and Education vs. Industrial Deployment

This is perhaps the most consequential structural question in Unitree's prospectus, and it extends well beyond a single company. Of Unitree's humanoid robot revenue in the first nine months of 2025, approximately 73.6% was classified as research and education sales. True industrial deployment — what investors most value as a proof of commercialisation readiness — represented approximately 9% of humanoid revenue, with actual intelligent-manufacturing sales of just RMB 15.7 million.

Further, within the industry-application segment, Unitree's prospectus acknowledges that enterprise reception and tour-guide deployments account for roughly 50–70% of income — uses that represent early-stage or novelty-driven procurement rather than genuine production-line integration.

This structural reality reflects an industry-wide problem, not a company-specific failure. The gap between unit shipment volume and downstream industrial penetration characterises the entire embodied intelligence sector at this stage of development. Inspectors will likely probe the sustainability of research-and-education revenue as a long-term demand driver, and whether Unitree's production scale translates into a demonstrable pathway to industrial adoption.

The contrast with UBTech is instructive. UBTech delivered 1,079 full-size humanoid units in 2025 generating RMB 820 million in revenue, with over 80% deployed in automotive manufacturing, smart logistics, semiconductor fabrication, and 3C electronics — genuine industrial use cases accumulated over 22 months of field validation. UBTech achieved this through a methodical, deployment-first model at the cost of slower initial scale. Unitree's strategy has prioritised volume and price competitiveness. Both approaches have legitimate strategic logic; their capital markets implications differ significantly.

Table 3: Unitree vs. Peer Commercialisation Comparison

4.3 Valuation Mechanics and Earnings Quality

Unitree's IPO implies a market capitalisation of no less than RMB 40 billion — a figure that represents more than three times its most recent private-market valuation of RMB 12.7 billion (USD 1.7 billion post-Series C). The targeted IPO valuation of up to RMB 50 billion (approximately USD 7 billion) circulated by some sources represents a nearly four-fold step-up.

There are two distinct concerns here. The first is methodological: humanoid robot hardware companies scale linearly, not exponentially. Unlike software-driven AI companies where each additional customer adds near-zero marginal cost, each additional robot unit requires incremental manufacturing, supply chain capacity, and after-sales support. The non-linear valuation frameworks applied to software businesses are structurally inapplicable to hardware-first robotics. Regulators and investors alike are being asked to develop new valuation frameworks for a category that does not neatly fit existing precedents.

The second concern is revenue recognition specificity. Unitree's prospectus flags potential audit risk around its JD.com fulfilment channel, where a "ship now, settle later" model creates timing ambiguity in revenue recognition. Additionally, reliance on automated system confirmation of delivery introduces subjective elasticity into when revenue is formally booked. These are not allegations of misconduct — they are disclosed risks — but they are precisely the kind of issues on-site inspectors are mandated to evaluate.

4.4 Founder Pre-IPO Share Transfer

Prior to filing, Unitree's founder Wang Xingxing transferred shares via secondary sale — a form of partial liquidity realisation ahead of a public listing. Under the Chinext Fourth Listing Standard introduced in April 2026, controlling shareholders, actual controllers, directors, and senior management are restricted from reducing holdings of pre-IPO shares for three complete fiscal years post-listing.

The regulatory relevance of pre-IPO share transfers is a nuanced question that inspectors may or may not prioritise. More broadly, it raises a governance question that is sector-wide: how should capital markets balance legitimate founder liquidity needs against the expectation that leadership remains economically aligned with long-term company performance? The regulatory treatment of Wang Xingxing's transfer will set a precedent for other founder-led robotics IPO candidates currently in the pipeline.

5. Sector-Wide Implications

5.1 Regulatory Standard-Setting for Hard-Tech Listings

The STAR Market's "1+6" reform package (2025) extended the Board's fifth listing standard — previously applicable only to biotech — to commercial aerospace companies, creating a framework for pre-revenue or loss-making innovators to list on verified technology merit. A parallel accommodation for humanoid robotics is under active consideration.

The Chinext Fourth Listing Standard (published 11 April 2026) establishes quantitative hurdles: minimum estimated market capitalisation of RMB 3 billion plus annual revenue of at least RMB 200 million plus a three-year revenue CAGR of at least 30% (Track A); or minimum market capitalisation of RMB 4 billion with three-year cumulative R&D of at least RMB 100 million comprising no less than 15% of revenue (Track B). These standards create a structured pathway for mid-tier embodied intelligence companies that do not meet STAR Market thresholds.

The inspection outcome for Unitree will serve as an interpretive precedent. Regulatory acceptance of its engineering-iteration model as consistent with hard-tech classification will clarify the definition of science and technology innovation attributes in a direction favourable to hardware manufacturing companies. A contrary finding would tighten the standard, increasing compliance costs for companies prioritising production scale over basic research expenditure.

5.2 Valuation Recalibration Across the Sector

The humanoid robot equity index (PE TTM) stood at approximately 31.5x as of 20 March 2026, placing it near the 7th historical percentile over five years — a significant derating from the frenzied 2024–2025 period. While a one-year recovery has brought it to approximately the 45th percentile, the market is materially more selective than at its peak.

Market concentration is accelerating. TrendForce projects that Unitree and AgiBot will together account for approximately 80% of domestic humanoid robot shipments in 2026, with the top five players capturing close to 60% of total market volume. This Mattheian dynamic — increasing advantages for scale leaders, compression for the long tail — is reshaping how capital allocators think about sector exposure.

If Unitree's inspection proceeds cleanly and its listing is validated, a RMB 40–50 billion valuation establishes a pricing anchor for subsequent listings by UBTech (Hong Kong), AgiBot, Galaxy General, and Zhiyuan Robotics. If the inspection surfaces material concerns, a broader re-rating of sector valuations is likely, making the state-backed investment position — including the National Integrated Circuit Industry Investment Fund ("Big Fund III"), which made its first embodied intelligence investment in 2025 — a potential stabilising variable.

Table 4: Robot Sector Valuation Context (as of April 2026)

5.3 Consolidation Timeline and Exit Pressure

The inspection event occurs against a backdrop of accelerating capital market activity across the robotics sector. UBTech listed on HKEX in late 2023 as the first humanoid robot public company. Zhipu AI and MiniMax listed on HKEX in January 2026. Unitree is pursuing an A-share listing. Behind them, AgiBot, Leju Robot, Deep Robotics, and others are progressing through IPO preparation stages.

The urgency is partly driven by window dynamics: early movers benefit from novelty premiums and benchmark-setting advantages that erode as the listing queue lengthens. It is also driven by VC exit pressure — backers with 2021–2023 vintage positions are approaching fund lifecycle constraints. Unitree's Series C investors, including Alibaba and HongShan, have limited natural liquidity options other than a public market listing.

Parallel to the direct-listing pathway, a parallel M&A and backdoor listing track has opened. AgiBot (Zhiyuan) acquired a controlling stake in Swancor Advanced Materials (listed on STAR Market). Qiteng Robotics acquired control of Shengtong Energy. These moves suggest that for companies that cannot clear the direct listing bar — on financial metrics, R&D intensity, or timing — alternative capital market access routes are being actively constructed. Jaka Robotics, despite substantial IP and backing from SoftBank Vision Fund and Temasek, withdrew its A-share application in 2025 after failing to adequately satisfy regulatory standards. The message from that case: patent volume alone is no longer sufficient as a hard-tech qualifier.

6. Two Scenarios and Their Implications

The inspection process will take weeks to months to conclude. Its outcome remains uncertain at the time of publication. We present two base scenarios and their respective industry implications:

Scenario A: Inspection Passes, IPO Proceeds

If Unitree's on-site review reveals no material deficiencies and its listing application continues through normal review channels, the following outcomes are most probable:

• A RMB 40–50 billion valuation receives implicit regulatory endorsement, establishing a reference price point for competitor listings in 2026–2027.

• The engineering-iteration pathway to hard-tech classification is validated, reducing listing friction for hardware-first manufacturers in the sector.

• Capital concentration around proven leaders accelerates, with primary market resources disproportionately flowing to companies with verified shipment scale and financial credibility.

• Unitree's prospectus becomes the de facto disclosure template for subsequent humanoid robot IPO candidates, shaping how others frame R&D, revenue structure, and commercialisation narratives.

Scenario B: Inspection Surfaces Material Issues

If the inspection identifies substantive deficiencies requiring remediation, correction, or rejection:

• The inspection outcome may cause a broader reassessment of how R&D ratios, revenue mix, and valuation assumptions are evaluated across all pending robot IPO applications.

• The IPO window for research-and-education-dominated revenue models narrows materially, pushing peer companies to demonstrate industrial deployment depth before filing.

• State-backed investment activity — particularly Big Fund III — would likely serve as a market floor, mitigating but not fully offsetting the valuation correction.

• Alternative listing routes — Hong Kong primary listings, backdoor acquisitions — become relatively more attractive for companies with uncertain A-share eligibility.

7. Editorial Assessment

The framing of Unitree's on-site inspection as a regulatory crisis misreads both the mechanism and the context. The selection was random, the sampling rate was expected, and the track record of companies inspected since 2024 is strongly positive. Unitree's fundamental financial performance — profitability, gross margins approaching 60%, and revenue growth that is exceptional even by global standards — is not in dispute.

The genuine questions are more nuanced. They centre on whether a business whose humanoid revenue remains majority research-and-education, whose R&D ratio has declined substantially as scale has grown, and whose pre-IPO valuation step-up is substantial, can satisfactorily address regulatory scrutiny on each dimension. None of these questions is fatal to the listing thesis. All of them are legitimate subjects for the inspection process to examine.

More consequentially, the outcome of this inspection will be one of the most closely watched regulatory signals in China's technology capital markets in 2026. Unitree is not simply filing for an IPO. It is, whether it intended to or not, filing a test case that will define what it means to be a humanoid robot company worthy of public market access in China. The answer to that question will shape the sector's development trajectory for years.