人形机器人：真正的护城河

Humanoid Robotics: The True Moats In our initiation (here), we believed that it was still too early to identify the ultimate winnersin humanoid robotics. We continue to see meaningfulsecond-mover advantagesin thisspace (Exhibit 1), particularly as first-movers have yet to establish durable moats. Therefore,we would like to assess what will ultimately emerge as the industry’s true moats. Dien Wang, Ph.D.+852 2123 2622dien.wang@bernsteinsg.com Stacy A. Rasgon, Ph.D.+1 213 559 5917stacy.rasgon@bernsteinsg.com Technology can confer an advantage, but not a moat. In China, tech leads are short-lived (with a typical lead window of 1~2 years) and quickly eroded without sustainedinnovation, particularly when companies misjudge technology paths or hit innovationceilings. This pattern, seen in solar and new energy vehicles, is now emerging in humanoidrobotics: strong locomotion capability has rapidly shifted from a differentiator to a baseline,with late entrants even overtaking first movers (here). Robin Zhu+852 2123 2659robin.zhu@bernsteinsg.com David Dai, CFA+852 2918 5704david.dai@bernsteinsg.com Moats for humanoid robot OEMsare likely to rest on four pillars:high-value data,ecosystem,IP & brand, andcost leadership. 1) High-value data (Exhibit 9), e.g. corner-cases, is key to pushing robotic performance from 95% toward near-perfect reliability,which in turn enables broader robotic deployment. 2) Ecosystem is a classic moat, with self-reinforcing loop driving user growth and retention (Exhibit 11). However, it remains unclearwhether it will be led by humanoid pure-plays or by incumbents integrating robots into theirexisting ecosystems. 3) IP & brand (e.g.Tesla,Sony,Nintendo,Disney,Hybe) will becomeincreasingly important differentiators when robots enter consumer markets (Exhibit 16to Exhibit 20). 4) Cost leadership will remain critical for mass adoption and competitiveadvantage (Exhibit 15). So far, however, no company appears to have built a truly defensiblemoat in humanoid robotics. Laurent Yoon+1 917 344 8502laurent.yoon@bernsteinsg.com Min-Joo Kang+852 2123 2644minjoo.kang@bernsteinsg.com Jay Huang, Ph.D.+852 2123 2631jay.huang@bernsteinsg.com Component suppliershave a clearer path to building moats. The advantages that drovesuccess in their existing end-markets are likely to carry over into humanoid robotics. Wesee three core moats: superiorquality controlin mass production,cost leadership,andrapid-response R&D. While best-in-class quality and cost control remain traditionallean manufacturing advantages, rapid-response R&D is increasingly important in meetingcustomers’ evolving needs and deepening partnerships.Shuanghuan, the global leaderin gears & reducers, stands out as a rare player that has built deep moats and continuesto gain share across multiple verticals. Among semi components suppliers,InfineonandRenesasare key beneficiaries given their established industry leadership and broadproduct portfolios spanning power semiconductors, analog, sensing, connectivity, etc. Weibin Liang, Ph.D.+852 2123 2666weibin.liang@bernsteinsg.com Hyrum Caesar+81 3 6777 6979hyrum.caesar@bernsteinsg.com Jack Lin+852 2123 2683jack.lin@bernsteinsg.com Carmine Milano, CFA+44 20 7762 1857carmine.milano@bernsteinsg.com Compute and software platform providersalso have a clear path to building moats withNVDAandQCOMat the forefront of developing processors that function as the “brain”of the robot, allowing it to process sensor data, reason, plan, and execute actions quickly.These are highly complicated system solutions that are difficult develop without expertise insemiconductor design and deep engagements with developers and customers, somethingboth NVDA and QCOM have cultivated over the years. Furthermore, NVDA offers full-stackrobotics, with an ecosystem spanning across training, models, simulation and on-robotinference, which further ensures continued engagement, design wins and deployment.Though not quite as extensive as NVIDIA’s ecosystem, QCOM’s workflow is also end-to-end, with the company building hardware, generating data and models, and deploying intocustomer environments, also ensuring continued customer engagement. Andrew Chung+1 917 344 8302andrew.chung@bernsteinsg.com Alrick Shaw+1 917 344 8454alrick.shaw@bernsteinsg.com Table Of Contents Humanoid Robot OEMs......................................................................................................................................................................................................... 2Components Suppliers........................................................................................................................................................................................................ 14Compute and Software Platform Providers.................................................................................................................................................................22Investment Implications....