Family offices are buying robotics while the rest of the market argues about an AI bubble

Investment firms belonging to billionaires in the United States, Europe and Asia have been putting money into robotics businesses built on machine learning, according to Bloomberg reporting published on 4 August 2026. The buying has carried on through a period in which a good part of the market has been arguing about whether AI valuations have outrun the companies underneath them.

Bloomberg cites three cases. The venture arm of Bernard Arnault’s family office took part in the Series A of Humanoid, a London maker of industrial robots. Jeff Bezos’s investment firm increased its allocation to Generalist AI, a US robotics software company. The family office of the Indian technology founder Azim Premji is in talks to lead a funding round for a maker of general-purpose robots with operations in Paris and Silicon Valley, which would be at least its second commitment to the sector this year.

The Humanoid round gives a sense of the numbers involved. The company announced $152 million on 21 July at a post-money valuation of $1.35 billion, led by the deep-tech firm Prime Movers Lab, with Bosch, Schaeffler and Fubon Financial Holding Venture Capital also participating alongside Aglaé Ventures, the Arnault family’s investment vehicle. Humanoid was founded in 2024 and has now raised $270 million in total. A Bosch robotics subsidiary will act as its contract manufacturer.

Bloomberg puts the combined net worth of the billionaires behind these vehicles at $628.5 billion on its Billionaires Index, and makes the point that none of those fortunes were made in artificial intelligence. The money came from luxury, retail and IT services.

Why hardware rather than models

Barclays expects the market for humanoid robots to grow by more than 1,000 per cent over the next decade, a forecast that comes with all the usual caveats attached to a decade-long projection of a market that barely exists yet.

Our own view, offered as an assessment and not as reported fact: robotics gives an investor things a foundation model cannot. A physical product, a named industrial buyer, a factory that will build it, a price per unit. Families whose wealth came out of manufacturing and retail have somewhere to stand when they look at that. Whether that footing holds when most of the value turns out to sit in the software rather than the arm is the open question, and nobody investing at Series A knows the answer yet.

The part that matters for the family office

The interesting part is the method rather than the theme. These are direct commitments at Series A, made by in-house teams, into companies with almost no revenue history to underwrite. CNBC reported in early July that the Bezos office alone backed five AI startups during the month of June.

That model asks a lot of the office. Early-stage rounds carry no reliable marks for years, which makes reporting to the family a matter of judgement rather than arithmetic. Follow-on capacity has to be reserved before it is needed, or the office watches its position dilute in the rounds that matter most. And a cluster of separate investments in one theme is a single position, however many company names appear on the schedule.

Then there is the governance question, which tends to be the one families answer last. A direct deal at this stage needs someone with the authority to say no, a written mandate on concentration by theme rather than by asset class, and a record of who decided what and on what basis. Offices that built their direct-investing capability around real estate and operating businesses do not automatically have any of that.

Arnault, Bezos and Premji can be wrong about robotics and barely feel it. A family office running a few hundred million cannot, and that difference gets forgotten quickly once a peer group’s allocation starts being read as a benchmark.

Sources: Bloomberg, 4 August 2026; Forbes on the Humanoid round, 21 July 2026; CNBC, 2 July 2026.