Barclays: General-Purpose Humanoid Robots Face Long Delay to 2035

The optimism surrounding general-purpose humanoid robots has hit a significant reality check. Barclays, in a research note published in September, argues that large-scale deployment of fully autonomous humanoids is unlikely to occur until around 2035. This pushes back the timeline by approximately five years from the 2030 target that many in the market had anticipated. The bank’s assessment suggests that the current excitement is outpacing the actual technological readiness of the sector.

Crucially, the firm indicates that the first substantial wave of value in physical AI will not come from human-shaped machines at all. Instead, specialized robots designed for narrow, specific tasks are expected to drive initial commercial gains. The core issue identified is not a lack of mechanical parts, but a fundamental gap in intelligence. Current humanoid capabilities often rely on pre-programmed sequences or remote human control, meaning they have not yet achieved the true autonomy required for broad economic utility.

Intelligence remains the primary bottleneck

The industry is currently stuck in a complex cycle involving scale, cost, and capability. Without large-scale deployment, manufacturing costs remain high. However, without advanced AI, there is no proof of commercial value to justify that scale. This creates a catch-22 that mirrors the long, expensive journey autonomous driving vehicles have undergone over the past decade. The hardware exists, but the software required for perception, reasoning, and real-time action is still immature.

Barclays notes that this technological gap is the fundamental obstacle preventing the industry from moving from demonstration to deployment. While mechanical arms and joints are becoming more refined, the ability to navigate unpredictable physical environments without human intervention remains elusive. This means that investors and businesses should not expect a sudden explosion of general-purpose labor replacement in the immediate future.

Compute demand drives early infrastructure

Even before robots are widely sold, the demand for computing power is already surging. This is because training the AI models that will eventually control these machines requires massive data center resources. Barclays highlights that this compute demand precedes hardware deployment. Developers need to simulate millions of scenarios and train foundation models long before a single robot is deployed in a factory or warehouse.

A prime example is the partnership between Figure and Nscale, which involves a multi-year agreement worth billions of dollars to support tens of thousands of high-end GPUs. This illustrates a key economic shift: the early winners in this space may be the infrastructure providers and AI model developers rather than the robot manufacturers themselves. The data center market is already tight, and this new demand from robotics adds further pressure to the system.

Specialized robots lead the near-term market

While general-purpose humanoids wait for a breakthrough in AI, specialized robots are already moving ahead. These machines are designed for specific tasks, such as sorting packages or welding car parts, where the environment is controlled and the actions are predictable. Because they do not require the same level of general reasoning, they can be deployed profitably today.

For the reader, this means the near-term impact of robotics will be felt in narrow industrial applications rather than in broad household or general labor roles. The trade-off is clear: we get efficiency in specific areas now, but we must wait for a decade or more for the versatile, general-purpose workers that were originally promised. The gap between current capabilities and future potential remains the defining feature of this market.