Global Robotics and Hardware Startups

What it is

Robotics and hardware startups are companies building physical machines, the software that animates them, or both. The sector spans industrial automation (arms, cobots, and autonomous mobile robots for factories and logistics), humanoid robots designed to work alongside people in unstructured environments, and defense-grade autonomous systems. The unifying thesis from 2024 onward is embodied AI: foundation models trained on large, diverse datasets that give a single robot system the ability to operate across tasks, environments, and body types without per-task fine-tuning. The International Federation of Robotics (IFR) recorded a US$16.7bn global industrial robot installation market value in 2025, with 4.66 million robots in operational use worldwide and 542,000 new units installed that year, double the figure from a decade earlier.

History

Industrial robotics has been dominated since the 1970s by a small cluster of incumbents: ABB (Switzerland), KUKA (Germany, acquired by China's Midea Group in 2016), Fanuc (Japan), and Yaskawa (Japan). SoftBank Group's (Japan) 2017 acquisition of US-based Boston Dynamics and its subsequent sale to South Korea's Hyundai Motor Group in 2021 signalled that large-capital interest in mobile, non-arm-form robotics was maturing. The embodied-AI startup wave crystallised around 2022 and 2023, when Figure AI (US, founded 2022), 1X Technologies (Norway, rebranded 2023), and Skild AI (US, Pittsburgh, founded 2023) all entered with foundation-model ambitions. By September 2025, Figure AI had reached a US$39bn valuation, roughly a 15-times increase from US$2.6bn in February 2024. Skild AI followed with a US$1.4bn Series C at a US$14bn valuation, led by SoftBank, in January 2026.

Current state

As of mid-2026, robotics and hardware startups globally had raised US$18.8bn in venture and growth capital in the first half of the year alone, already past the full-year 2025 total of US$15bn and above the prior global peak of US$14.1bn set in 2021. The US leads in foundation-model robotics software: Skild AI's Skild Brain is designed to retrofit any robot without task-specific retraining; General Intuition (US$2.3bn valuation, June 2026) trains on video-game footage to generate transferable action policies; Generalist AI (US$2bn, June 2026), founded by former Google DeepMind researcher Pete Florence, closed the year's largest Series A in robotics AI with its GEN-1 model reporting 99% task success. China is running a parallel, state-aligned sprint: two Shenzhen startups, X Square Robot and AI² Robotics, both hit approximately US$2.8bn valuations on a single day (June 29, 2026), funded by Meituan, Alibaba, ByteDance, and Xiaomi rather than Western venture capital; Chinese robotics investment passed RMB 46bn (roughly US$6.4bn) in 2026 before mid-year. Germany's Neura Robotics closed up to US$1.4bn in Series C funding in June 2026, the largest European robotics round on record. Defense-grade autonomous systems overlap the category: Mach Industries (US, US$1.8bn valuation, June 2026) produces autonomous weapons platforms from a 115,000-square-foot facility in Huntington Beach, California; Saronic (US, Austin, Texas) raised US$1.75bn at a US$9.25bn valuation for autonomous naval vessels. Meta (US) entered the humanoid subsector in May 2026 by acquiring San Diego-based Assured Robot Intelligence.

Relationships

Skild AI drew strategic investment from Nvidia (which operates its own Isaac robotics platform), Samsung, LG, and Schneider Electric, evidence that hardware OEMs are treating robotics software as infrastructure rather than building it in-house. SoftBank Vision Fund (Japan) is the most active repeat backer across the US field, with positions in Boston Dynamics and Skild. China's embodied-AI capital flows from consumer-technology conglomerates, insulating the domestic sector from US export controls and pairing model development with domestic manufacturing scale. The IFR identified humanoid deployment as a structural 2026 trend, with robots moving beyond prototypes into automotive, warehousing, and manufacturing settings where cycle times and energy costs are tracked against commercial targets.

What to watch

• Whether the Skild Brain's cross-morphology claim holds at production scale, or collapses to vertical-specific fine-tuning as earlier robot-AI bets did.
• China's funding pace: if RMB 46bn in H1 2026 continues, the full year could approach US$12bn, narrowing or reversing the US capital lead.
• US export-control expansion to robotics software and training data, which would directly constrain cross-border transfer of embodied-AI models built on US compute.
• Whether defense-robotics and civilian-humanoid investment streams remain separate or converge as the underlying foundation models become applicable to both.