Humanoid robotics 2026: Figure, Apptronik, 1X and Tesla compared

The most valuable humanoid robot company has not shipped a thousand robots. Figure is worth $39 billion. Its fleet is in the hundreds. That gap, between what the market prices and what the factories actually produce, is the real story of humanoid robotics in 2026.

After $7 billion flowed into humanoid startups in 2025, four names separated from the pack. None of them ships at scale yet. Each is spending its war chest on a different answer to the same question: which route leads to a robot that works 24/7 without a human on a headset?

What is accelerating: paid pilots at BMW, Mercedes and GXO

The 2024 field ran on prototype videos. The 2026 field runs on purchase orders. The shift happened quietly, inside factory gates.

Figure operates paid pilots at BMW's Spartanburg plant in South Carolina. Its BotQ factory now produces a robot roughly every 90 minutes, and monthly shipments were doubling as of April, reaching about 240 units. The commercial case is no longer hypothetical. BMW pays for the units it runs.

The capital structure changed with it. Figure's Series C was led by Parkway Venture Capital with participation from Brookfield Asset Management, NVIDIA, Intel Capital, Salesforce, LG Technology Ventures, T-Mobile Ventures and Qualcomm Ventures. That is a strategic coalition of infrastructure money and chip money, not a speculative seed round. It values the company at $39 billion, up about 15 times from the $2.6 billion Series B valuation in early 2024.

Apptronik took the same route through a different door. Its Apollo robot runs pilots at Mercedes-Benz plants in Germany and Hungary and at GXO logistics sites. The company closed a $520 million extension to its Series A in February 2026, led by Google, Mercedes-Benz and B Capital, with AT&T Ventures, John Deere and the Qatar Investment Authority joining. The round pushed Apptronik past $935 million in Series A financing and above a $5 billion valuation.

The investor roster reads like a map of industrial buyers, not a venture fund list. That is the point. These are companies placing strategic bets on a robot they might operate.

What is decelerating: fleets counted in hundreds

The uncomfortable number sits on the factory floor, not the balance sheet.

Every leader's deployed fleet is measured in hundreds, not thousands. Figure has talked about 100,000 robots over four years. The delivered total so far is a small fraction of that. Apptronik's Apollo and 1X's Neo are in pilots and preorders, not volume production.

Tesla's Optimus is the clearest case of promise ahead of product. The robot works mostly inside Tesla's own factories rather than in paying third-party plants. Production targets have slipped repeatedly, and the long-run $20,000 to $30,000 price is a projection, not a price list. The company brings unmatched manufacturing scale and the data engine behind Full Self-Driving. It has yet to convert either into external paid deployments.

Autonomy is the bottleneck. 1X has been candid that early Neo units rely on remote human teleoperation for edge cases while the AI learns. That honesty is rare in a sector built on staged videos, and it shows how far the shipped capability sits from the marketing.

As we wrote in July, humanoid robots reached European factory floors at BMW's Leipzig plant, and the economics of production humanoids remain the open question.

What is new: the $20,000 home robot

The new development of 2026 is not industrial. It is domestic.

1X opened direct consumer preorders for Neo at $20,000, or roughly $500 per month as a subscription. That makes the home the first mass market for humanoid robotics, decades after industrial automation went mainstream.

Tesla guides Optimus toward the same $20,000 to $30,000 band, but as a long-run target at scale, not a current offering. The price convergence matters more than the rivalry. Two of the four leaders are pointing at the same number, which signals where unit costs have to land for the category to work.

Consumer units now price below $25,000 while industrial platforms run roughly $150,000 to $320,000. The split is not cosmetic. The economics are different. A home robot sells on safety, noise and software updates. An industrial robot sells on uptime, payload and total cost per task.

The home bet is riskier and bigger. If a robot works in a living room, it can work almost anywhere. That is the logic, and it is why 1X took it while the other three stayed in warehouses and factories.

The four leaders, side by side

Figure has the most capital and owns its AI stack. Apptronik has the deepest robotics IP and the most credible industrial pilots. 1X owns the only serious home-market entry. Tesla owns manufacturing scale. None of them has solved the autonomy problem that separates demos from reliable 24/7 work.

That is why the valuations look identical from the outside and feel different from the inside. The market is not pricing four robots. It is pricing four bets on which model of progress wins.

What to check before underwriting a humanoid bet

The four-way comparison collapses into three data points when capital is at stake.

First, delivered units versus stated targets. Figure talks about 100,000 robots over four years; the audit trail is monthly shipment disclosures from its own factory. Tesla talks about millions and publishes none of the numbers. When a company publishes its own production data, you can verify it. When it does not, you are underwriting a slide deck.

Second, who pays. A paid pilot at BMW or Mercedes-Benz is revenue with a name attached. A pilot that is still described as an evaluation, or a preorder that has not converted, is not. The difference shows up in the same quarter you are trying to value.

Third, the autonomy floor. Teleoperation hides the gap between the demo and the product. 1X says its early Neo units use human backup for edge cases. That is honest, and it means the real capability level is not what the marketing implies. Ask how much of the workload still runs on a human in a headset.

The unit economics matter more than the robot: the profit or loss on each deployed unit. A $39 billion company shipping hundreds of units implies a revenue multiple that would be extreme even for a public software business. If the multiple is to compress, the fleet must compound. That is the only variable that resolves the paradox in the opening line.

Valuations in this sector are priced for thousands of robots. The fleets are in the hundreds. The gap is not a scandal. It is the thesis, and it either closes or it collapses.

Hardware is close to solved. The bottleneck is AI generalization and unit economics. Whoever turns a $39 billion valuation into thousands of reliably deployed robots wins the decade.

The fleet gap is closing, but slowly. For an investor, the four leaders are four different risk profiles wearing the same label. Read the deployment data, not the demo reel.