Crunchbase Data Shows Q3 Robotics Startup Funding Jumps 185% As US And China Lead Physical AI Push

A 185% year-over-year increase means funding during the first three quarters of 2026 was about 2.85 times the comparable period in 2025. It signals an unusually fast expansion in investor interest, not merely steady growth.

The quarterly comparison makes the surge concrete. Robotics startups raised $9.9 billion globally in Q3 2026, compared with $4.3 billion in Q3 2025. That is an increase of $5.6 billion, or more than double the earlier total.

The momentum is broad enough to span several quarters. The article says Q1 through Q3 funding was up around 185% year over year, with multiple quarters reaching nearly $10 billion or more. The surge is creating new unicorns, although the sector’s direction remains uncertain.

China and the United States dominate robotics investment in 2026. Chinese robotics startups received $13.2 billion, while U.S. companies received $12.1 billion. Germany ranked third, but its $1.7 billion total was far smaller.

China’s biggest cited rounds went to Guangzhou-based Xpeng Robotics, which raised $900 million, and Shenzhen-based AI² Robotics, which raised a $736 million Series C. Both companies are developing intelligent robots and AI capabilities for physical-world work.

The United States produced several billion-dollar rounds. Pittsburgh-based Skild AI raised $1.4 billion for its Skild Brain robotics foundation model. Los Angeles-based Atoms also raised a billion-dollar round. Germany’s Neura Robotics raised $1.4 billion for cognitive robots and supporting AI and data infrastructure.

Physical AI is artificial intelligence designed to operate in the physical world. It supports systems that sense environments, interpret conditions, and act through robots or other machines. That makes it relevant to manufacturing, healthcare, logistics, and household tasks.

Text-based AI can learn from the enormous text collections available online. Physical AI cannot rely on that resource because there is no internet of physical-AI data. Its information comes from sensors and must be converted from sensor tokens into action tokens.

This difference makes physical AI more demanding to build. Its datasets combine vision, tactile information, force, and torque. The article says companies are still in the pre-training phase, but may soon move into post-training as these systems develop.

Robotics investment is spreading across many physical-world tasks rather than one narrow product category. The article identifies heavy industry, manufacturing, logistics, healthcare, mining, transportation, services, and household uses. This breadth gives startups many possible markets to pursue.

Examples show how varied the field has become. AI² Robotics is building intelligent robots for manufacturing, healthcare, and services. Atoms includes robotics divisions for mining and transport. Other startups are developing cognitive robots, foundation models, and data systems that can support different applications.

Investors are therefore funding both robot makers and the underlying intelligence layer. Companies are building models, collecting data, and creating systems for specific industries. The article says this variety may allow many companies to win, even while the sector’s ultimate direction remains uncertain.

When large amounts of capital enter robotics, more companies can pursue different technical approaches and industries. Funding supports robot development, data collection, model building, infrastructure, and specialized applications. It also increases competition as startups race toward breakthroughs.

The scale is visible in the unicorn figures. Around 75 robotics companies on the unicorn board raised funding during the past three years. More than half, or 45 companies, joined the board in 2026. These startups span data and models, heavy industry, manufacturing, logistics, household products, and other uses.

The influx does not guarantee that every company will succeed. The article says the sector’s direction remains uncertain. Still, Ulrik Stig Hansen believes many companies may win because the physical world offers so many different applications for robotics.

Robots must understand changing physical conditions and respond with appropriate actions. Text alone cannot describe everything needed for that task. A robot also needs information about what it sees, feels, pushes, pulls, and experiences through its sensors.

The article says physical-AI datasets combine vision, tactile data, force, and torque. Developers collect this information and convert it from sensor tokens into action tokens. That creates a direct training link between a physical situation and the robot’s response.

These datasets are orders of magnitude larger than ordinary digital training collections, according to Ulrik Stig Hansen. Storing, curating, enriching, and validating them is expensive. The article says compute and storage are being built now, while physical AI remains in its pre-training phase and approaches post-training.

In general, venture capital is money invested in young companies that could grow substantially but carry significant risk. Investors provide funding in exchange for an ownership stake. The article does not define venture capital, but it shows why robotics attracts this type of financing.

Robotics companies need capital for physical systems and the intelligence behind them. They must collect multimodal sensor data, convert it into useful training material, and pay for storage, compute, curation, enrichment, and validation. These operations are especially expensive for physical AI.

Investors are funding companies before the sector’s winners are settled. The article says startups are racing toward further breakthroughs, while the direction remains uncertain. Large rounds let competing companies develop models, robots, and specialized applications at the same time, hoping one or more approaches becomes commercially successful.