Robotics funding Oct 19 to 28: USD 1.63bn in 10 days, 88% of all of Q3

Ten days produced almost a full quarter of disclosed robotics funding, and the cheques went to companies that sell outcomes on heavy machines, aircraft, factory lines and labs. That changes the reference points for every robotics founder raising in the next six months: a USD 30 to 90 million seed is now on the record several times over.

The concentration also matters. With 80% of the money in six rounds, the period says more about a handful of growth investors (CapitalG, Lightspeed, Andreessen Horowitz, M&G, Nimble) than about the whole market. A smaller startup should not read the total as easier money for itself.

Bedrock now has more disclosed capital than any other construction-autonomy company in our database, and its round lands two weeks after ICON's USD 200 million Series D on Oct 13. Reliable and Kodama show the same thesis in aviation and forestry: retrofit autonomy onto machines customers already own.

In models, Eka (vision-force-action, reinforcement learning in simulation), Dream Labs (a robot brain sold via API), Machenta (learning from one demonstration) and Enact (force and touch) all position against Physical Intelligence and NVIDIA GR00T. None has published a third-party benchmark that separates it from those incumbents yet.

Only two valuations in the period come from the companies: Tractian at USD 1.4 billion and Reliable above USD 1 billion. Bedrock's USD 4.1 billion and Eka's USD 1.45 billion are outlet reports that the companies have not published, so we do not use them in any ratio.

For scale, Tractian's valuation is 10.8 times its USD 130 million round, a typical late-stage multiple; Reliable reached a billion while its Series D was only USD 275 million in total, which suggests investors are paying for its FAA certification lead rather than revenue.

Most of the money buys deployment, not invention: excavator retrofit kits, aircraft integration, sensors on industrial motors, factory floor space (Foundry's 130,000 sq ft Emeryville plant), and lab buildings (Warp). That spending lands with US integrators, machine OEMs and contract manufacturers rather than with Asian component makers.

The model companies spend on compute and data collection. Maxinsights (egocentric video) and Enact (force data) are early examples of a data supply chain for robot learning that did not exist as a funded category a year ago.

1) Monday, Nov 2: MEKAWIRE weekly funding data; a second week above USD 100 million a day would confirm a new pace. 2) Mid-November: Form D filings for the US rounds on EDGAR, which give share classes and first-sale dates. 3) Any confirmed Physical Intelligence round; at the reported USD 1 billion it would be the largest robotics round in our October data.

Thesis: the late-October cluster rewards robots with contracted revenue and the model layer that could serve them, not general-purpose humanoids. Confidence: medium. The mix is clear in our data, but the count before October is incomplete and one cheque is 37% of the total.

We would lower confidence if November rounds swing back to humanoid hardware or if the large rounds turn out to be mostly debt.

How much of Bedrock's USD 600 million is primary capital versus secondary? What share of Foundry's Series A is equity, and who led it? Will Eka, Dream Labs and Machenta publish results on a shared benchmark? How many of the 16 rounds were committed before the summer and only announced now?