Robotics Funding Boom Rests on a Few Mega-Rounds and Data Bets

A big number with a narrow base

Robotics and physical-AI startups have raised roughly $48 billion so far this year, according to PitchBook data cited by the Financial Times 3. The figure suggests a sector in full bloom. A closer look shows something narrower. Most of the money is flowing into a small number of very large rounds, and a growing share is going to a less visible part of the stack: the companies that supply training data to robot-learning systems.

The number also needs careful handling. Crunchbase separately counted $47.4 billion in "physical AI" funding for the first half of the year. That category includes Waymo and companies in defense, aerospace and autonomous transportation, so treating it as a pure robotics figure would badly overstate the market 2. The two totals are close enough to be confused, but they measure different things.

Concentration, not breadth

The stronger evidence on how the money is distributed comes from F-Prime's latest State of Robotics analysis, which uses PitchBook data. It found that the funding growth of the past two years came from rounds above $50 million, while early- and mid-stage robotics funding has stayed essentially flat for five years 2.

The headline deals fit that pattern. Skild AI raised $1.4 billion at a valuation above $14 billion for a general-purpose "brain" meant to run across different robot bodies 2. Generalist followed with $400 million, Physical Intelligence previously raised $600 million, and FieldAI has raised $405 million for foundation models aimed at harsh real-world settings 2. A handful of rounds like these can account for much of an annual total.

This matters because a rising aggregate can look like a healthy funnel. If seed and Series A activity is not growing, the sector is not adding many new entrants. It is concentrating capital in perceived winners. That leaves less room for the next set of challengers, and it ties the overall numbers to the fortunes of a few heavily priced companies.

The data layer gets repriced

The clearest sign of where investors think value sits is the competition over robot training data. XDOF, a robotics data startup, is in late-stage talks for a Series B that would value it at $1.2 billion, about three months after it emerged from stealth 1. According to people familiar with the deal, the company was not trying to raise. Venture firms approached it after seeing how fast it was growing, following a $70 million Series A in June backed by Thrive Capital, Andreessen Horowitz, Lux and Spark Capital 1. XDOF reports roughly $50 million in annualized revenue and 20 customers, several of them frontier AI labs 1. Its products are data pipelines, collection tools and labeling systems that labs find hard to build themselves 1. TechCrunch could not confirm the size of the raise or whether the new money is included in the $1.2 billion valuation, and terms could still change 1.

Rival Mecka AI has closed a $60 million Series B led by Sequoia at around a $500 million valuation 3. Together with an $8 million seed and a $60 million Series A split into two tranches, that brings its disclosed funding to $128 million in just over a year 3. TechCrunch frames this as capital moving from model builders toward suppliers of scarce real-world data, with XDOF's reported price setting a benchmark well above Mecka's 3.

How to read it

The sources mostly agree on the story but stress different parts of it. The funding overview focuses on concentration and the risk of inflated totals 2. The deal reporting focuses on momentum and investor urgency 13. Both views can hold. The data layer is attracting money quickly because it is a real bottleneck. Robot foundation models need large amounts of physical-world data, unlike text models, which could draw on the web. Companies that can collect and label that data at scale have revenue now, which is more than many robotics startups can show.

Still, repricing a company several times over in one quarter, on about $50 million in run-rate revenue, is a bet on how big the category will become, not on how big it is today. Mecka's fast fundraising points to the same thing: investors want to secure positions before the market consolidates 3.

So the $48 billion figure is best read as a measure of conviction concentrated in a few places, not as broad growth across the sector. The base of the funnel is flat while the top is getting crowded and expensive. If the large robot-brain bets and data suppliers deliver, that concentration will look smart. If they stumble, there may be few younger companies coming up behind them, given how little early-stage funding has grown.