CNBC's The China Connection newsletter: Humanoid robots are great, but they need buyers too

China's humanoid robotics industry is scaling quickly, with a Beijing factory saying it has already produced 300 robots and is targeting 10,000 this year, while Lingyi iTech aims for 500,000 annual units by 2030. Pricing could fall from about $30,000 per humanoid as production scales, but demand remains unproven with most orders still just one or two units and regulators warning about a bubble. The article also highlights government-backed robot showrooms, Nvidia's planned robotics system with Unitree, and expanding overseas ambitions for Chinese robotics firms.

Analysis

The near-term market is likely overestimating the revenue quality of the humanoid build-out while underestimating the strategic value of the manufacturing stack. The first-order beneficiary is not the robot brand owner, but the contract manufacturer and component ecosystem that can scale units, drive down BOM cost, and own the data flywheel; that favors suppliers with PCB, motor-control, power, and assembly exposure more than pure-play AI narratives. The second-order effect is price compression: if scale cuts humanoid cost by ~50% over a multi-year horizon, the winners will be the firms that can monetize installation, service, and training rather than capture one-time hardware gross margin.

For NVDA, the bullish case is less about direct humanoid unit sales and more about preserving platform control as robotics moves from chips to software-defined stacks. The risk is that Chinese developers treat Nvidia as a development layer rather than a moat, then localize the full stack once training data and middleware mature; that is a 12-36 month threat, not a quarterly one. The bigger medium-term upside may come from inference and edge AI demand tied to robot fleets, but only if deployment shifts from demos to repeat commercial orders.

TTMI screens as a cleaner expression of the policy-driven supply-chain shift. If U.S. industrial policy expands credits for domestic AI hardware inputs, PCB demand could re-rate before volume visibly accelerates, because customers will pre-source capacity to qualify for incentives and de-risk tariffs. The contrarian takeaway is that the strongest trade may be in the “boring” enablers: if humanoids stay politically favored and supply-chain localization intensifies, component makers can outperform while robot OEM valuations mean-revert on bubble concerns.

Key Decisions for Investors

• Long TTMI over the next 3-6 months as a policy- and localization-beneficiary; use any tariff/credit headlines as entry points. Risk/reward is attractive because multiple expansion can occur before revenue inflects, while downside is cushioned by diversified PCB demand.
• Buy NVDA on weakness only, but size it as a platform-preservation trade rather than a pure humanoid call. Favor 6-12 month call spreads to express upside from robotics software lock-in while limiting valuation risk if Chinese stacks localize faster than expected.
• Short a basket of recently hot Hong Kong-listed AI/hardware IPOs on 1-2 quarter horizon where post-listing momentum has already broken. The thesis is repeat-order disappointment and liquidity air pockets after initial enthusiasm fades.
• Pair trade: long U.S. PCB / electronics manufacturing exposure vs. short a basket of China humanoid OEMs if accessible. This captures the policy and supply-chain beneficiaries while fading speculative hardware valuations that depend on future utilization.
• For GS, avoid assuming the Hong Kong weakness is purely sentiment-driven; use it as a signal to reduce exposure to China-linked capital markets businesses if mainland preference persists for another quarter. A renewed mainland-vs-HK allocation gap would pressure fee pools and secondary activity.