If your proof of a production ramp is a chart with no Y-axis posted on Threads, you don’t have a ramp. You have a narrative. Narratives are easy. Manufacturing is where the narrative goes to die.
Myth to retire: “Doubling every month” means “we’ve solved scale”
Forbes reports that Figure CEO Brett Adcock shared a chart implying Figure is doubling Figure 03 shipments month-over-month — but the chart reportedly has no Y-axis. Forbes then sketches back-of-the-envelope numbers (e.g., ~60, ~120, ~240 units across Feb–Apr) by anchoring to an Omdia estimate of ~150 shipped in 2025.
That’s not useless. It suggests momentum. But it’s not a production ramp. It’s an interpretation of a visual posted on social media, turned into numbers by a journalist doing math in public. That’s not “bad.” It’s just not the kind of evidence that survives procurement.
The definition trap: delivered, deployed, or doing work?
In robotics, “delivered” can mean anything from “it arrived at a customer site” to “it’s running a workflow and generating value.” The gap between those two is where most startups go to quietly suffer. It’s also why shipment charts can be misleading: you can scale boxes faster than you can scale installation, safety sign-off, training, maintenance, and support.
So when you hear “doubling,” ask the adult questions: how many of those units are active? What tasks? What intervention rate? What does “autonomous” mean here — and who gets woken up when it fails?
Reality: the metric that matters is boring and it’s missing
Shipping units is not the same as shipping reliable units. If you’re building humanoids, your real scale number is not “deliveries,” it’s: operating hours in the field per intervention. Put another way: how often does a human have to save the robot from physics, edge cases, or its own confidence?
For contrast, another Forbes report cites Shanghai-based Agibot claiming 5,000 humanoids shipped in three months. That’s the other end of the spectrum: volume claims. The industry is now split between “slow, high-capability, prove-it” and “ship a lot, learn fast, pray about support.” Both can work. Both can also explode.
The Droid Brief Take
Humanoid companies keep asking you to infer reality from a curve. I want the ledger. Unit count without failure rate is just cardio for the hype machine.
If Figure is genuinely accelerating shipments, good. But the next adult step is to publish deployment-shaped numbers: tasks, uptime, intervention rate, and how many “deliveries” turned into “works on Tuesday.” Until then, the chart is a vibe with a battery.
Also: charts without axes are not “secret sauce.” They’re “please don’t ask follow-ups.”
What to Watch
- Field evidence: named deployments, task scope, and any hard metrics beyond unit counts.
- Support reality: spares, repair turnaround, and what happens when the robot breaks in a customer facility.
- Definition hygiene: what counts as a “delivery,” and what counts as “autonomous” work.