China's humanoid robot R&D continues to heat up, with expected summer production of Optimus becoming a key catalyst
AI summary card
China's humanoid robot R&D continues to heat up, with expected summer production of Optimus becoming a key catalyst
UBS believes China's humanoid robot industry is making continued progress in hardware, financing, and data collection, but true mass production still depends on repeat orders, application expansion, and delivery of the Tesla Optimus timeline.
- The Beijing humanoid robot half-marathon showed improved hardware capabilities: more than 100 teams participated, and the winning time fell from over 2 hours and 40 minutes in 2025 to about 50 minutes.
- State Grid is reportedly set to invest about RMB 6.8 billion in embodied AI in 2026 and purchase about 8,500 related devices, including 500 humanoid robots.
- China's humanoid robot index has underperformed the machinery index by 17% since October 2025 and by 10% year to date, with short-term sentiment still affected by Tesla's progress.
- UBS is watching catalysts in 2H 2026, including the release of Optimus Gen 3, potential IPOs of Chinese humanoid robot companies, and new orders, especially repeat orders.
- The report favors upstream companies with strong core businesses, high barriers to entry, or a solid overseas customer base, specifically highlighting Horizon Robotics, Kedali, and Hengli Hydraulic.
Report interpretation
Overview
This report focuses on developments in China's humanoid robot industry. UBS notes that R&D activity in China continues to heat up, with progress in hardware technology, financing activity, investment from local governments and industrial capital, and diversified data collection channels. On Tesla's 1Q 2026 earnings call, the company said that Optimus could begin slow initial production in Fremont in late July or August 2026, and Gen 3 may be released before production starts, which is seen as a reaffirmation of the industry's timeline.
Core views
The core views include: first, China's humanoid robot industry is still accelerating, but progress varies across different segments; second, current orders are mostly for data collection, entertainment applications, or preliminary testing by industrial customers, and future repeat purchase rates and exploration of new applications are more effective than one-off orders in validating mass-production potential; third, the Beijing half-marathon showed clear hardware progress in motion control, stability, batteries, heat dissipation, and energy management, but software capability for multi-task generalization still has a long way to go; fourth, short-term market sentiment may remain weak in 2Q 2026, with catalysts more concentrated in 2H 2026; fifth, from an investment perspective, the report prefers upstream companies with solid core businesses, barriers around robot exposure, or overseas customer resources.
Analysis framework
The report combines industry event tracking, observation of orders and application scenarios, Tesla earnings call commentary, relative index performance, company visits, and screening of supply chain names to assess the humanoid robot industry's technological maturity, commercialization signals, and capital market catalysts.
Methodology notes
Use repeat orders and new application exploration to judge mass-production potential
The report believes that most orders in 2025 are still used for data collection, entertainment, or preliminary testing. Therefore, repeat purchase rates, continued procurement in industrial and commercial scenarios, and rollout of new applications are more important commercialization indicators than the size of one-off orders.
Use the release and production pace of Tesla Optimus Gen 3 to gauge sentiment catalysts for the sector
The report compiles Tesla's statements on Optimus production across past earnings calls and treats the start of initial production in Fremont in late July or August 2026 and the launch of a second factory at Giga Texas in summer 2027 as important milestones.
Use the humanoid robot index's performance relative to the machinery index to measure market sentiment
The report uses the WIND AI Robot Concept Index as a proxy for the humanoid robot index and notes that it has underperformed the machinery index by 17% since October 2025 and by 10% year to date, mainly due to the lack of better-than-expected progress from Tesla Optimus.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Horizon RoboticsOne of the key China-focused names; the report positions it as a beneficiary related to AI chips.
- Strengths
- It has exposure to robot AI chips and edge intelligence, benefiting from demand for intelligent driving and robot intelligence.
- Weaknesses
- The commercialization pace of humanoid robots remains uncertain, and robot-related revenue contribution may take time to materialize.
- Comparison
- Compared with complete machine makers and lower-barrier segments, AI chips represent a direction with higher technological barriers.
- Risks
- Robot intelligence progress slower than expected, customer adoption slower than expected, and valuation volatility in the sector.
- KedaliOne of the key China-focused names; the report mentions its exposure to reducers and notes in related company reports that the target price implies upside of more than 70%.
- Strengths
- It has a strong core business foundation while also offering new market opportunities in robot components.
- Weaknesses
- It is still affected by cycles in its core NEV structural parts business, customer capacity expansion, and product quality.
- Comparison
- Compared with pure thematic companies, the report prefers its combination of core business support and optionality from new robot businesses.
- Risks
- Capacity build-out or customer capacity expansion slower than expected, loss of share with key customers, weaker-than-expected EV penetration, and rising raw material prices.
- Hengli HydraulicOne of the key China-focused names; the report links it to upstream high-barrier segments such as planetary roller screws.
- Strengths
- It has strong capabilities in hydraulics and mechanical components, and if it enters overseas brand supply chains or sees stronger-than-expected demand for non-standard products, it has upside optionality.
- Weaknesses
- It is still affected by excavator and industrial cycles, and the contribution from robot business still needs to be validated.
- Comparison
- Compared with lower-barrier manufacturing segments, key components such as planetary roller screws have higher barriers to entry.
- Risks
- Domestic excavator sales weaker than expected, weaker-than-expected demand for pumps and valves, and overseas business hindered by trade friction.
- Tesla OptimusA core sentiment and industry timeline catalyst for the global humanoid robot sector.
- Strengths
- If the release of Gen 3 and Fremont production materialize, they will strengthen expectations for industry mass production and supply chain validation.
- Weaknesses
- Initial production is expected to be slow, the supply chain is entirely new, and ramp-up may take a long time.
- Comparison
- Compared with local Chinese orders, Tesla's timeline has a more direct impact on global investor sentiment.
- Risks
- Delay in the release of Gen 3, slower-than-expected production ramp-up, and slower-than-expected supplier updates.
Key data
- TARS financingUS$455mTARS completed a pre-A financing round, reflecting accelerating fundraising for some robot companies.
- State Grid embodied AI investment planabout RMB 6.8 billionAccording to Jiemian News, around 8,500 embodied AI devices may be procured in 2026.
- State Grid planned purchase of humanoid robots500 units, budget of about RMB 2.5 billionApplications focus on dangerous or complex scenarios such as power inspection, maintenance, emergency rescue, and logistics.
- Scale of participants in the Beijing humanoid robot half-marathonmore than 100 teamsIncluding 5 international teams, about 5 times the previous year.
- Change in winning time for the Beijing eventshortened from over 2 hours and 40 minutes in 2025 to about 50 minutes in 2026Showing improvement in hardware capabilities such as motion control, stability, and energy management.
- 2026 event control methods60% remote control, 40% autonomous navigationThe event introduced separate autonomous navigation and remote control groups, reflecting that autonomous capability has become a key area of assessment.
- Relative performance of the humanoid robot indexunderperformed the machinery index by 17% since October 2025 and by 10% year to dateShort-term market sentiment has been affected by Tesla Optimus progress falling short of expectations.
- April index reboundabout 9%As Middle East tensions eased, the humanoid robot index rebounded in April and roughly matched the machinery index's monthly return.
- Optimus production timelineslow initial production in Fremont to begin in late July or August 2026Stated on Tesla's 1Q 2026 earnings call; Gen 3 may be released before production starts.
- Second Optimus factoryGiga Texas to start production around summer 2027The report says meaningful ramp-up is expected from late 2026 to 2027.
Impact & implications
For investors, the industry is moving from proof of concept into a more structured phase of technological and commercial validation. Hardware progress and confirmation of Tesla's timeline help support sector catalysts in 2H 2026, but valuations and sentiment still require validation from real orders, repeat purchases, mass-production pace, and improvements in robot intelligence. Upstream segments with high barriers to entry, such as components, AI chips, planetary roller screws, and reducers, may offer more selective opportunities.
Risks
- A reduction in investment at the macro level may weigh on demand for Chinese industrial goods and import/export volumes.
- If policies such as tax incentives for high-tech enterprises are canceled, corporate profitability may be affected.
- Intensifying domestic and international competition may lead to market share loss.
- If humanoid robot orders fail to transition from pilots to repeat purchases, mass-production expectations may be revised down.
- Robot software generalization capability remains insufficient, and there is uncertainty around the rollout of multi-task applications.
- If the release or production ramp-up of Tesla Optimus Gen 3 falls short of expectations, sector sentiment may continue to be pressured.
- EV shipments, customer capacity expansion, raw material prices, and trade friction may affect the fundamentals of related component companies.
What to watch
- Whether Tesla Optimus Gen 3 will be released around late July or August 2026.
- Whether initial Optimus production in Fremont starts on schedule, and whether suppliers provide incremental updates.
- Potential IPO progress of Chinese humanoid robot companies.
- New orders in industrial and commercial scenarios, especially the proportion of repeat orders.
- Whether embodied AI procurement by State Grid and other power-sector players is implemented, and the specific procurement pace.
- The real-world application effectiveness of humanoid robots in power inspection, maintenance, emergency rescue, and logistics.
- Development of autonomous navigation, multi-task generalization, data collection channels, and robot rental models.