Figure says it built 350+ humanoids and hit a one-robot-per-hour cadence at BotQ; Tesla says it’ll convert Fremont and start Optimus production late July/August (slowly); Agility points to 100,000 totes moved at GXO. Three companies, three definitions of ‘production.’
In humanoid robotics, ‘production’ is a word with infinite elasticity. It stretches to fit whatever you did this quarter.
So let’s pin it down with something boring and useful: what each company is actually measuring, and what that implies about who’s closer to the thing everyone claims they’re already doing.
What’s happening
Figure says its BotQ facility has delivered 350+ Figure 03 robots and improved throughput from one per day to one per hour (a “24x throughput improvement”) in under 120 days. It also publishes manufacturing and test metrics: 50+ in-process inspection points, end-of-line first-pass yield “over 80%,” and 80+ functional verification tests per robot.
Tesla (via reporting on its Q1 2026 call) says Optimus production will begin at Fremont in late July or August, after converting the Model S/X line, but warns initial output will be “quite slow” given a new line and “10,000 unique parts.” Another summary claims Tesla is designing a first-gen line for 1M robots/year and preparing a second-gen line in Texas aimed at long-term capacity of 10M/year.
Agility, meanwhile, points to a different metric entirely: its Digit humanoid has moved 100,000+ totes at a GXO facility — a throughput milestone intended to demonstrate ROI in a live logistics workflow.
Three definitions of “production” (and why they’re not interchangeable)
- Units produced: how many robots came off a line (Figure’s “350+”).
- Capacity promised: what the line is designed to do someday (Tesla’s 1M/year and 10M/year claims).
- Work performed: what the robots actually did in a customer workflow (Agility’s 100,000 totes).
The problem isn’t that any one metric is “fake.” It’s that they answer different questions. Investors love capacity. Engineers love yield. Customers love throughput. Reality demands all three.
Figure vs Tesla: maturity signals to look for
Figure’s blog post reads like a manufacturing org showing its homework: supplier qualification, inspection points, yields, actuator counts, battery pack shipments, and extensive testing. That’s what you write when the constraint is making the same thing twice.
Tesla’s story, as reported, is a conversion story: dismantle an existing car line and rebuild it as a robot line in four months. That’s what you say when the constraint is building the factory that builds the robot.
The real comparison: who is closest to boring, repeatable ROI?
Agility is trying to win the unsexy prize: high-volume cycles in a logistics site, where safety, uptime, and workflow integration matter more than cinematic motion.
Figure is trying to win the scale prize: enough identical robots to generate data, learn faster, and deploy broadly.
Tesla is trying to win the narrative + supply chain prize: convert automotive manufacturing muscle into humanoid manufacturing and force an ecosystem to respond.
The Droid Brief Take
If you want to know who’s “in production,” don’t ask for a robot count. Ask for a failure rate, a service loop, and an ROI story that survives contact with fluorescent lighting.
Because the factory floor doesn’t care about your valuation. It cares about whether the robot works on the third shift, when nobody is filming.
What to Watch
- For Figure: whether “one per hour” translates into sustained shipments and real customer deployments — and how quickly yields improve.
- For Tesla: whether Fremont actually starts producing Optimus on the stated timeline, and what “slow” means in 2026 volumes.
- For Agility: whether throughput milestones expand into multi-site scaling and a repeatable deployment playbook.
Sources
Figure — Ramping Figure 03 Production
Electrek — Tesla pushes Optimus V3 reveal later this year – again
Not a Tesla App — Everything Tesla Announced During Its Q1 2026 Earnings Call: Summary/Recap
Agility Robotics — Digit Moves Over 100,000 Totes in Commercial Deployment