Morgan Stanley published its latest "Humanoid Horizons" dispatch on February 26, 2026, and the timing couldn't be more charged. With a Trump-Xi summit potentially on the horizon, the report — co-authored by Adam Jonas in New York and Sheng Zhong in Hong Kong — zeroes in on a question that is rapidly moving from theoretical to operational: can the US and China actually collaborate on building the robots that may define the next industrial era, even as both governments race to dominate it?

The answer, at least for now, appears to be a cautious and commercially driven yes — though the fault lines are unmistakable.

Chinese Suppliers Plant a Flag in America

The most concrete development in the report is the joint venture announced earlier this month between Leaderdrive and Minth Group, two China-based manufacturers, to design and produce humanoid robot joint module assemblies for the North American market. Minth's California subsidiary holds 60% of the venture; Leaderdrive holds 40%.

Morgan Stanley frames this as a potential "win-win" — but the strategic calculus cuts both ways. For Chinese suppliers, establishing US production capacity reduces geopolitical exposure and locks in their position in a supply chain they already dominate. China currently supplies the vast majority of robotic actuator components globally, a fact the bank's channel checks confirm extends even to the largest American robotics firms. For the US, the argument runs that permitting advanced Chinese manufacturers to localize creates an opportunity to absorb critical process knowledge and, over time, rebuild a competitive domestic supply base.

Whether Washington views it that way is another matter. Congress has introduced bipartisan legislation to establish a National Commission on Robotics and restrict federal procurement of certain foreign-made humanoid robots on national security grounds. The Commerce Department, under Howard Lutnick, has reportedly been exploring executive orders to accelerate American robotics capacity. The regulatory environment around this JV is, to put it plainly, unsettled.

Beijing Makes It Political — Literally

If anyone doubted how seriously China's leadership takes humanoid robotics, the Lunar New Year broadcast on state television CCTV settled it. Robots from four startups — Unitree, MagicLab, Noetix Robotics, and Galbot — performed sword-fighting, kung-fu, and trampoline routines in front of one of the largest domestic television audiences of the year.

Morgan Stanley is careful to temper the spectacle: the routines were almost certainly pre-programmed, not evidence of genuine real-time AI decision-making. The real bottlenecks — training data scarcity, compute constraints, model architecture limitations — remain very much intact. But the bank's analysts note that the Chinese government's deliberate choice to spotlight humanoids during a flagship national broadcast "underscores the strategic importance it places on robotics as a pillar of its broader ambition to lead in global AI development." That is a policy signal, not a product demo.

Optimus 3 and the Tesla Wildcard

On the US side, the most closely watched near-term catalyst is Tesla's Optimus Gen 3 unveil, expected before the end of the first quarter of 2026. Tesla has already announced it is winding down production of the Model S and Model X — roughly 2% of total 2025 vehicle sales — to free up factory space and engineering resources for the Optimus program.

Morgan Stanley expects Gen 3 to be a substantial redesign rather than an incremental update, with Elon Musk's recent comments pointing toward improved dexterity and a simplified, more manufacturable architecture. The bank's auto team, led by Andrew Percoco, expects Optimus Gen 3 to remain largely confined to Tesla's own factories in 2026, serving as a controlled environment for data collection and hardware iteration — a sensible, if unglamorous, first step toward the mass-market ambitions Musk has telegraphed.

OpenAI, Microsoft, and the Depth of US Supply

OpenAI's late-January request for proposal seeking US-based supply of actuators, harmonic drives, precision bearings, permanent magnets, and power electronics is being read by Morgan Stanley as a pivotal demand signal. The argument: legitimate procurement interest from a company of OpenAI's scale may finally convince major US industrial suppliers — currently focused on legacy markets like automotive, aerospace, and machinery — to build advanced robotics capacity and compete directly with Chinese counterparts.

Microsoft, meanwhile, released its first robotics-focused vision-language-action model, "Rho-Alpha," derived from its Phi series, designed to help robots perceive and act in less structured environments. Qualcomm unveiled its Dragonwing IQ10 robotics processor platform at CES 2026. The breadth of big-tech engagement is widening.

The Numbers: China Doubles Shipments, Global Market Heads to $7.5 Trillion

Morgan Stanley estimates China humanoid shipments will roughly double to 28,000 units in 2026, up from approximately 12,000 in 2025. Critically, the composition is shifting: last year, 42% of deployments went to R&D and education, 19% to data collection centers backed by local governments — including RMB 829 million (US$120 million) in orders received by UBTECH from six municipal governments — and only 4% to industrial and logistics applications. In 2026, the bank sees the center of gravity moving toward business-led commercialization, with integrators running trials across commercial use cases in pursuit of scalable, specialized deployments.

The bank's AlphaWise survey found 62% of respondents likely to adopt humanoid robots within three years — a strong underlying demand signal even if near-term products remain premature for most applications.

Longer term, Morgan Stanley projects the global humanoid market reaching $7.5 trillion in annual revenue by 2050, with cumulative global adoptions hitting approximately 1 billion units. China alone is projected to account for roughly 302 million of those units; the US, approximately 78 million. East Asia and the Pacific region collectively would represent the largest share, at around 43% of global cumulative adoptions.

Morgan Stanley's equal-weighted Humanoid 100 index is up 37% since its inception in February 2025, outpacing the S&P 500, MSCI Europe, and MSCI China — though trailing MSCI Korea and MSCI Taiwan. Top performers include Hyundai, Teradyne, Samsung Electronics, and Rainbow Robotics. The message from the market, at least so far, is that the robotics supercycle is real — even if the geopolitics of who builds it, and where, remain very much in play.