If you’re eyeing humanoid robots, ready to buy into the “next trillion-dollar narrative” — look at two numbers first.

In September 2025, Figure AI closed its Series C at a $39 billion post-money valuation — a 15x jump from its Series B ($2.6B) eighteen months earlier. Its annual revenue? A few million dollars — negligible against that valuation.

Around the same time, John Deere’s weeding robots were already running across more than 5 million acres — larger than New Jersey — charging farmers by the acre, and cutting herbicide use by an average of 59%.

Same “embodied AI.” One is worth $39B on almost no revenue; the other is quietly counting cash in a field, and nobody calls it a robot stock. That’s the most expensive lesson in this space: fame isn’t fortune.

The market pays for the loudest name — not the cash flow

Compress the rule into a framework — the smile curve. Value pools at the two ends: upstream (compute, models, sim) and downstream (real scenes, data flywheels, outcome-based pricing). The middle — bolting parts into a humanoid body — is the loudest and the thinnest, most easily commoditized by the Chinese supply chain.

The trap: all the spotlight is on the thinnest middle (humanoid bodies), while the actual cash flows sit downstream, in vertical leaders that don’t carry the word “robot”:

• Deere: See & Spray identifies weeds plant-by-plant and fires individual nozzles — textbook razor/blade plus data flywheel. Every connected machine is a data node; the more it’s used, the sharper and stickier it gets. Precision-ag is already its largest segment (~37% of FY2025 revenue), with a target of 10% recurring revenue by 2030.
• Aurora: the first driverless commercial freight on US public roads; 2026 revenue guidance up ~400% YoY, heading toward an ~$80M run-rate by year-end. Long-haul trucking — fixed routes, driver shortage, billable miles — has far clearer ROI than a humanoid.
• Defense autonomy: AeroVironment’s latest fiscal-year revenue was ~$1.98B, up ~141% YoY. Battlefield unmanned systems are the hardest-budgeted, already-at-scale embodied AI.

While the market cheers Figure’s $39B on near-zero revenue, Deere is already charging by the acre across 5 million of them.

“But Figure’s an option!” — Sure. Just know which kind you’re buying.

The pushback: Figure is expensive because it’s an option on a huge future.

Fair. But you must separate “delivered cash flow” from “zero-revenue narrative option” — and size them with completely different discipline. Paying 30x earnings for a machine already making money, and paying $39B for a name with negligible revenue, are two different things. The first is an asset; the second is a lottery ticket. Treating the lottery ticket as a core holding is the most common way to die in this space.

Fame rises and falls; cash flow doesn’t lie. The best already-earning embodied AI may carry no “robot” in its name at all. Next time a humanoid makes your heart race, ask one question: is this company actually getting paid today?

Do you own a name that’s “really earning embodied-AI money but nobody calls it a robot stock”? Name it below.

(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.)

— Adapted from Embodied Intelligence Investing, Deep Dive XVI: Already-Earning Embodied AI

#EmbodiedAI #Humanoids #Deere