Summary
Zornitsa Todorova, Head of Thematic Research at Barclays, presents the bank’s long-term forecast for the humanoid robotics market on Bloomberg Tech, projecting it could reach $200 billion by 2035 as part of a broader $1 trillion physical AI economy. In 2026 alone, approximately 60,000 humanoid robots are expected to be deployed globally — a four-fold jump from the 15,000 units shipped in 2025 — with China accounting for 85 to 90 percent of near-term production. By 2035, Barclays models roughly 13 million new units deployed per year, with 30 million humanoids potentially in active operation.
The central bottleneck Todorova identifies is not demand or capital, but real-world training data. Unlike the text and image data that powered the current wave of large language models, training a general-purpose humanoid requires physical-world data — humans performing everyday tasks like laundry, household chores, and deliveries — that simply does not exist at scale on the internet. Simulation offers a cost-effective partial solution but is likely to hit fundamental limits in replicating the messiness of real physical environments.
On hardware, Todorova notes the humanoid supply chain is nascent compared to EVs or traditional industrial robots, and argues the ecosystem will crystallize around a killer application that justifies the scale-up. The US is narrowing China’s lead in robot capability if not unit volume, and the digital twin approach is highlighted as a promising path to grounding humanoid perception in real-world physics.
📺 Source: Bloomberg Tech · Published September 02, 2026
🏷️ Format: Interview
