If you’re in public markets, itching to get on the “humanoid robot” wave but haven’t pulled the trigger yet — this is for you. Don’t rush to buy the body.

Start with one abnormal number.

In September 2025, humanoid company Figure closed a $1B Series C at a $39B post-money valuation — a 15x jump from its prior round a year earlier.

Let me clear up the quotes in the title: Figure isn’t literally “zero-revenue.” It’s had paying customers since 2025 — the BMW line pays roughly $25 per robot-hour (not per vehicle; those 30,000 X3s are the line’s output, not money to Figure). But at that price, a few dozen robots running flat-out for a year likely annualizes to only a few million dollars.

Put that against the scale of the industry: one Goldman Sachs estimate for the entire global humanoid market by 2035 is about $38B.

A company with revenue still in the low millions carries a private valuation approaching a top bank’s forecast for the whole industry a decade out — a price-to-sales in the thousands-of-times. “Zero-revenue” never meant literally not a cent; it means revenue is negligible against valuation.

This isn’t a question of “expensive or not.” It’s a signal: the capital pricing of embodied AI runs far ahead of its financial delivery. The market isn’t paying for cash flow — it’s paying for the option on a paradigm: the unproven-but-enormous possibility that “physical-world labor can be redefined by AI.”

And here’s the most expensive mistake retail makes: equating “embodied AI” with “the humanoid body.”

Value isn’t spread evenly across the chain. It’s a smile curve: thick upstream (compute, foundation models, sim), thick downstream (real scenes, data flywheels, per-unit paid operations), thin in the middle (body assembly, standard parts).

The “humanoid body” sits exactly at the thinnest, most crowded bottom of that curve — compute and models pulling value from above, scenes and data pulling value from below, and the Chinese supply chain crushing hardware prices from the side. Attacked on three fronts.

Compare who’s actually earning:

• The automation systems moving totes in warehouses (Symbotic) — a backlog north of $20B, revenue truly ramping;
• The da Vinci in the operating room (Intuitive Surgical) — the “install base + consumables” model proven for decades: that is already-delivered embodied AI;
• And most humanoid-body companies: huge raises + stunning demos + revenue negligible against valuation.

Per BofA, dedicated humanoid funding was ~$4.3B in 2025, up from ~$700M in 2018. Capital is pouring in far faster than revenue.

So the disciplined play isn’t “bet on which body company wins.” It’s a barbell:

One end, take certainty via the shovel sellers — don’t bet on a winner, just benefit as total robot count rises;
The other end, take moats via downstream names with proprietary scenes and real cash flow;
The bodies in the middle — only a tightly-capped small position, as an option on the still-unproven technical assumption: whether VLA models show a predictable “scaling law” the way LLMs did.

Embodied AI is a real decade-long industrial trend — I don’t doubt that.
But the price the market pays for it today is half real industrial opportunity, half an overdraft of capital sentiment.
Telling those two apart, and betting with a barbell at the border of iron and silicon — that’s the whole craft.

So next time a “robot company raises a fortune, drops a stunning demo” headline appears, which question will you ask first: did it change the industry’s value distribution — or just push the narrative premium higher again?

(Independent industry and educational research, not investment advice. Companies named are illustrative examples, not recommendations. Data is from public sources and may change. The author and affiliates may hold positions in securities mentioned.)

— Excerpted from the preface of Embodied Intelligence Investing: The Industrial Map and Public-Market Framework of the Physical AI Era

#EmbodiedAI #Humanoids #Investing