Goldman Sachs On Spring Festival Gala Robots: Hardware Advances Drive Adoption, Future Hinges On Underlying AI

Goldman Sachs assesses that the humanoid robot performance at the 2026 CCTV Spring Festival Gala demonstrated notable advances in China’s robotics hardware engineering, while the demonstration did not fully validate the robots’ underlying AI capabilities. In a report dated February 21, the firm observed that performances by Unitree, Noetrix, Magic Lab and Galbot highlighted substantial improvements in mechanical design and whole‑body control, yet the highly choreographed nature of the show limits conclusions about real‑time autonomous cognition.

Goldman Sachs retains its shipment forecasts of 51,000 global humanoid robots in 2026 and 76,000 in 2027, representing a multiple‑fold increase from the 15,000–20,000 units estimated for 2025. The bank expects near‑term upside for supply‑chain equities but cautions investors about potential first‑quarter earnings pressure and inflationary raw‑material costs. Over the longer term, the pace at which general AI capabilities mature will determine the sector’s upside, and Goldman Sachs singles out “world model” approaches as a technology pathway deserving close attention.

The Spring Festival Gala provided a concentrated platform to display hardware capabilities. Unitree’s systems illustrated coordinated multi‑robot choreography, fluid motion, dynamic balance and human‑like fault recovery in several scenarios. Magic Lab and Galbot demonstrated high levels of dexterity and precision in complex tasks, while Noetrix emphasized bionic designs that enable seamless human‑robot interaction and expressive movement. Analyst Jacqueline Du noted that the coordination and stability evident in these routines reflect rapid progress in actuators, sensor integration and advanced control algorithms.

Despite the impressive physical execution, Goldman Sachs underscores that gala performances are typically preprogrammed and rehearsed to showcase peak performance under controlled conditions, relying on predefined trajectories rather than on autonomous decision‑making in unpredictable environments. Consequently, the robots’ cognitive reasoning, adaptability and capacity for independent learning in unstructured settings remain to be proven.

Goldman Sachs also highlights the Gala’s role in raising public awareness and acceptance. The event’s large audience offered many viewers a first close encounter with advanced humanoid machines, helping demystify the technology and potentially reducing public apprehension. By presenting robots in an engaging entertainment context, the Gala may broaden social acceptance and facilitate expansion of applications beyond traditional industrial uses into services, entertainment, education and personal assistance.

Commercial deployments are expected to drive the near‑term shipment growth Goldman Sachs projects. Use cases such as security patrols and customer service in public venues—hotels, banks, museums, exhibition centers, car dealerships and supermarkets—leverage existing planning, mobility and interaction capabilities while avoiding the complexities of highly dexterous manipulation. The bank anticipates that markets may react positively to key supply‑chain stocks in the coming days as applications scale.

Goldman Sachs warns investors to remain vigilant about short‑term risks. First‑quarter results, to be reported in March–April, could present challenges for companies with heavy exposure to China’s electric‑vehicle market. Rising costs for copper, silver, gold and memory chips also pose margin risks that should be factored into investment decisions.

Looking further ahead, Goldman Sachs stresses that the industry’s long‑term trajectory depends on the development speed of general AI capabilities, which in turn relies on effective data and model strategies. The firm references its 2035 baseline of 1.38 million units and notes that the realization of additional upside will be driven by advances in AI generality. In particular, the “world model” paradigm—where agents learn internal representations or simulations of their environment to predict future states and plan actions without continuous physical interaction—could enable robots to reason about causality and support the emergence of more general, capable and safer machines.