Venture investors put $47.4 billion into physical AI companies in the first half of 2026 — more than the $41.9 billion the sector raised across the entire three years from 2022 through 2024, according to Crunchbase data published Tuesday. The haul came across 521 deals, up nearly 80% from the $26.4 billion across 436 deals in H1 2025 and close to four times the $12 billion the sector took in during the second half of last year. That last comparison is the tell: H2 2025 had more deals, 470 of them. Roughly the same number of companies are now raising almost four times the money.
By Crunchbase's criteria, physical AI spans robotics, autonomous vehicles, aerospace, drones, industrial automation and sensors — broad enough that a single check can move the whole number, and one did. Waymo's $16 billion Series D in February, co-led by Alphabet, Dragoneer, DST Global and Sequoia Capital at a $126 billion valuation, accounted for nearly a third of every venture dollar in the category. Strip it out and the remaining $31.4 billion across 520 deals is still a record, but growth falls from 80% to roughly 19%. The rest of the megadeal list skews defense: Anduril raised another $5 billion in May at a $61 billion valuation, double its mark from less than a year earlier; Shield AI took a $2 billion Series G in March at $12.7 billion; and Saronic, which builds autonomous sea vessels, raised $1.75 billion in a Kleiner Perkins-led round at $9.25 billion.
Humanoids draw the attention but not the biggest American checks. They are where China's money goes. Robotics alone has pulled $18.8 billion globally in 2026, against $15 billion for all of 2025 and $14.1 billion in the 2021 peak. China accounts for more than 43% of global robotics venture investment, per Crunchbase, and its rounds skew earlier and denser — Shanghai's TARS Robotics raised a $513 million seed at a $1.9 billion valuation about a year after founding; Shenzhen's EngineAI took $200 million. Through mid-May, China-based robotics companies had raised $5.6 billion across 176 deals, already past the $4.3 billion they raised in all of 2025.
Exits have followed, concentrated in aerospace and defense: SpaceX raised $75 billion in a June IPO at a $1.77 trillion valuation, and Mobileye paid roughly $900 million for Israeli humanoid startup Mentee Robotics. Then came Wednesday. Unitree Robotics, priced at 50.8 yuan a share for a roughly $9 billion valuation, surged as much as 620% in its Shanghai STAR Market debut — the first humanoid robotics listing in mainland China — briefly touching a 445 billion yuan ($66 billion) market cap before closing up 460%. The same morning in Beijing, the 11th World Robot Conference opened with more than 300 exhibitors and 150 product launches, and Xiaomi unveiling its first humanoid.
Unitree shipped more than 5,500 humanoid robots last year, a number both remarkable and tiny. "It is one of the few companies in humanoid robotics with genuine revenue, profitability and meaningful production volume," Rinat Mirzaitov, founder of Humanoid Analytics, told Al Jazeera. "Its main challenge is turning hardware and manufacturing leadership into application and workflow leadership."
Why It Matters
The $47.4 billion is best read as a rotation, not an expansion. Crunchbase points to Wall Street Journal reporting that firms built on software, internet services and social media are now writing checks into physical technologies tied to the AI boom — same funds, same limited partners, different asset shape. Software AI is underwritten on gross margins near 80% and near-zero marginal cost. Physical AI is not. For this capital to clear its hurdle rate, three things have to become true at once: hardware costs keep falling, deployments convert into recurring revenue, and reliability reaches the point where a customer will run a fleet without a human minder.
Backers argue the first is already done. Ryan Ziegler, a general partner at Edison Partners, told Crunchbase News that cheap sensing changed the math — "Even our mobile phones now have LIDAR scanners on them" — and that the strongest of these businesses look like vertical software, with "attractive unit economics, large deal values and multi-year deployments."
But Joe Fath, partner and head of growth at Eclipse, supplied the line that should give underwriters pause: "customers value operational efficiency, reliability, and revenue, not technical sophistication alone." That is the gap. JP Morgan puts 2025 global humanoid robot sales at roughly $2 billion, against a projected $300 billion by 2035. Investors just committed more than twenty times that revenue base in six months, on a ten-year forecast. Waymo at $126 billion is at least a metered business with rider-miles behind it; much of the rest is funded on the premise that generalization arrives on schedule.
What To Watch
Second-half comparables are the first test. Without a Waymo-scale anchor, H2 2026 has to clear roughly $31 billion on organic deal flow for the trend to look broad rather than like one financing. Watch whether deal count rises with the dollars, or concentration deepens.
Watch Unitree's aftermath and the policy split around it. A 460% first-day close sets a comparable that AgiBot, Galaxea AI, EngineAI and Robot Era will be measured against. Meanwhile the Trump administration banned imports of Chinese-manufactured humanoid robots in July, and JP Morgan estimates China already controls roughly three-quarters of the global market for humanoids and autonomous vehicles — a bifurcation that pushes US deployment economics further out.
And watch what these companies disclose. The number that settles this argument is not another round. It is units shipped, revenue per unit and gross margin — figures Unitree now has to publish on a schedule, and that almost none of its private peers do.
“Customers value operational efficiency, reliability, and revenue, not technical sophistication alone.”— Joe Fath, Partner and head of growth, Eclipse