1X Robotics vs. Figure AI: The Business Model Battle Behind the Robot Hands Race

The Robot That Moves Like It Means Business

1X’s Neo robot just demonstrated finger dexterity that made the internet uncomfortable — and that’s exactly the point. But while the tech press obsesses over how fast those fingers move, the more important question is: what business model sits behind the hardware? Because in the humanoid robot race, the company that survives won’t necessarily be the one with the best hands. It’ll be the one with the smartest monetization architecture.

1X’s Business Model: The Quiet Norwegian Outlier

1X Robotics — backed by OpenAI and headquartered in Norway — is pursuing a model that looks less like a hardware company and more like a robot-as-a-service infrastructure play. Rather than selling Neo units outright, 1X has positioned itself around deployment contracts with industrial and logistics partners. That’s a recurring revenue bet, not a one-time transaction bet.

This matters enormously. Hardware companies die from margin compression. Service companies die from churn. 1X is betting it can thread the needle: build robots that are good enough to deploy at scale, lock in operators through training data loops, and monetize the ongoing relationship — not just the unit sale.

The “freaky fast fingers” demo isn’t just an engineering flex. It’s a sales pitch to warehouse operators who need manipulation speed to justify the deployment cost. Every viral video is a B2B sales funnel entry point.

Figure AI Is Playing a Different Game — and So Is Tesla

Compare 1X’s approach to Figure AI, which signed a deal with BMW and has been explicit about its intent to sell robots into manufacturing lines at scale. Figure’s model leans harder into the enterprise hardware sale — get the unit placed, then sell software upgrades and support contracts on top. It’s closer to how Kuka or ABB built industrial robotics empires: hardware first, ecosystem second.

Tesla’s Optimus is the wildcard. Tesla doesn’t need Optimus to be profitable as a standalone product — it needs Optimus to reduce Tesla’s own labor costs in its factories first, then potentially externalize that capability as a product. That’s a vertical integration play, not a robotics startup play. Tesla can afford to subsidize the hardware because the savings accrue internally. 1X and Figure cannot.

This creates a structural asymmetry: Tesla is building a cost center that becomes a profit center. Startups like 1X are building profit centers that currently operate as cost centers. The cash burn math is very different.

The Prompt Injection Problem Hiding Inside Every Robot

Here’s the angle almost nobody is discussing: humanoid robots running on large language model instruction layers are prompt injection attack surfaces. This week, Ars Technica reported that defenders are now actively using prompt injection as a security tool — essentially fighting fire with fire in AI systems. That dynamic applies directly to robots like Neo.

If Neo’s dexterity is driven by natural language instruction parsing, then its physical capabilities are only as secure as its prompt layer. An attacker who can inject a malicious instruction into a warehouse robot’s task queue doesn’t steal data — they cause physical damage or shut down operations. The liability architecture of humanoid robotics companies becomes a core business model question, not just a legal footnote.

This is why 1X’s OpenAI backing is strategically significant beyond just capital. OpenAI has direct financial incentive to solve the prompt security layer — because if it doesn’t, 1X’s robots become a public relations catastrophe that damages OpenAI’s entire enterprise positioning. The security model and the business model are fused.

The Framework: Hardware as Distribution, Data as the Real Asset

The humanoid robot companies that survive the next five years will follow a consistent pattern: use hardware as a distribution mechanism to collect proprietary manipulation and task-completion data, then monetize that data through model licensing and software subscriptions. The robot is the Trojan horse. The training data is the moat.

This mirrors how platform business models work in digital markets — the product that looks like the business is actually just the on-ramp to the business. Understanding how companies engineer these flywheel structures is central to the business model canvas analysis that separates real competitive advantages from hardware demos.

1X’s Neo demo is impressive. But the real story isn’t the fingers. It’s whether 1X can deploy enough robots, in enough facilities, to build a manipulation dataset that no competitor can replicate. That dataset — not the hardware — is what a future acquirer or IPO investor will actually pay for.

Bold Prediction

Within 24 months, at least one major humanoid robotics company will pivot away from hardware sales entirely and re-launch as a “robot intelligence” software licensor — selling its trained manipulation models to third-party hardware manufacturers. The company most likely to make that pivot first is not Tesla. It’s one of the underfunded startups that runs out of hardware margin before it runs out of data value.

If you want frameworks like this applied to the biggest business model shifts in tech every week, subscribe to Business Engineer here.

FourWeekMBA AI Business Intelligence — strategic analysis of the moves that matter.

91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.