Figure AI closed a Series C exceeding $1B in September 2025 at a $39B post-money valuation. Skild AI raised a $1.4B Series C in January 2026 — pushing cumulative to $2.2B — at over $14B, with SoftBank leading. Saronic Technologies stacked $2.58B through a Series D on March 31, 2026 to build autonomous warships. Anduril is at $2.8B raised through Series E with the Pentagon as anchor customer. The 41 companies in this category split cleanly into three story arcs running in parallel — and only one of them has unit economics that close.
Humanoids hit the unicorn threshold, not the factory floor
Figure AI, 1X, Apptronik, Agility Robotics, and Sanctuary AI all have humanoid platforms shipping, demoing, or pilot-deployed. Figure's $39B valuation is the headline number; Agility's Digit is the only humanoid running paid commercial pilots in actual logistics environments at scale. Underneath the hardware, Physical Intelligence ($735M Series B) and Skild AI are racing to build the foundation model that controls the next generation of humanoids — Skild Brain explicitly markets as a single model that runs any robot, any task. The capital has crossed the unicorn line. The hardware unit economics — bill of materials, manufacturing yield, service costs over a multi-year deployment — have not.
Defense and dual-use is where the contracts live
Anduril ($2.8B, Series E) and Saronic ($2.58B, Series D, March 2026) scale against Pentagon procurement budgets that pure-software startups never see. Skydio ($340M Series D) ships autonomous drones for inspection and public safety. Helsing in Europe has the equivalent posture against EU and NATO buyers. Defense is the only buyer category in robotics today that absorbs premium pricing on day one — which is why every major round in 2025-2026 has at least a partial dual-use story attached.
Self-driving consolidates, warehouse robots quietly ship
Waymo ($31B raised, Alphabet-owned) and Aurora ($2.5B growth) anchor self-driving as a long-cycle infrastructure play. Wayve ($2.25B Series D, February 2026) is the UK end-to-end-learning entrant. The smaller AV pure-plays have largely been absorbed or wound down. Meanwhile, the warehouse and manufacturing robot plays — Covariant ($222M Series C), Sereact ($110M Series B), Standard Bots, Path Robotics — are the ones quietly shipping working systems with revenue, while the humanoid headlines absorb the press cycle. The 2026 question is whether foundation-model-driven control collapses the per-task data cycles fast enough for general-purpose hardware to catch the specialized incumbents.