Robotics and physical AI have already drawn close to $48B in 2026

Capital keeps rushing into robotics and physical AI. According to QbitAI, in the first nine-plus months of 2026 global robotics and physical AI companies have raised close to $48 billion, more than in all of last year, with more than $160 million flowing into the sector each day.

The figure looks even more striking over a longer horizon. Citing data from the Financial Times and PitchBook, the report says annual funding for these companies sat roughly between $10 billion and $15 billion from 2022 to 2024, rose sharply in 2025, and that 2026 has already reached about 1.7 times the whole of 2025.

The money is not spread evenly; it is concentrating in a handful of leaders. The report points to a string of billion-dollar rounds and ever-rising valuations for robotics startups, with the largest deals accounting for a significant share of total funding.

In terms of where it goes, capital now spans several parts of the robotics chain: companies focused on robot bodies and hardware, companies trying to build general intelligence systems that work across different robots, and companies taking robots into factories, warehouses and homes.

A bottleneck that comes up repeatedly is training data. Large language models can pull huge amounts of text, images and video straight from the internet, but robots need information about space, motion, contact and force, data that cannot simply be harvested from the web.

That has pushed robotics companies to build their own data-collection systems. The report describes firms paying ordinary people to record everyday tasks at home, such as folding clothes or cleaning, and others sending robots into real settings to test them, while data companies that once served large-model developers now extend into robotics.

China is building similar infrastructure. Citing statistics, the report says that as of late April at least 90 humanoid robot data-collection and training centers in China were operating, under construction or planned, 64 of which were already running.

Overall, capital is spreading across the full chain from research and manufacturing to real-world deployment, but rising concentration and the unresolved data bottleneck mean that chain still has a long way to go. What to watch: whether leading companies' technical approaches pay off in real settings, and whether funding keeps clustering around a few names.