The robot knows when you're coming and it's programmed to play dead—which tells you everything about where humanoid deployment actually is.
The Summary
- Agility Robotics unveiled Digit 5, a humanoid designed to detect nearby humans, slow down, and power down completely if approached too closely
- The safety protocol reveals the gap between "humanoids in warehouses" and "humanoids working alongside people"—Agility's previous robots operated behind safety barriers
- The company reports $300M in Digit 5 orders but $140M operating loss on $1.8M revenue in 2025, as it seeks a $2.5B public debut
The Signal
Agility's pitch for Digit 5 sounds like progress until you realize what it's actually saying. The robot shuts itself down when humans get too close. Not because it's being polite. Because the alternative is liability no insurance company will touch and no factory floor manager will risk.
This is the humanoid gap nobody wants to talk about. The demos show robots picking boxes and walking up stairs. The reality is they still can't be trusted in the same room as a person without a kill switch that activates on proximity. Cofounder Jonathan Hurst calls this solving "major, absolute blockers" they didn't know existed until customer deployments. Translation: the robots worked in labs, then met actual warehouses.
"The first machine that is the full, formal product that has come from customer requirements."
Agility has real deployment numbers that competitors don't. Nine customer facilities running Digit across North America, including Amazon, Toyota, and GXO Logistics. But the financial picture tells a different story:
- $1.8M in sales for 2025
- $140M operating loss same period
- $300M in Digit 5 orders (multiyear, unprofitable at current margins)
- Seeking $2.5B valuation for public offering
The math doesn't math yet. Each deployed robot is a cost center Agility is subsidizing to prove the category exists. They're not selling robots at scale. They're selling the promise that robots at scale will eventually be sellable. That's a VC pitch, not a business model. The IPO is the product.
Meanwhile Figure AI and Tesla are burning even more capital on the same bet, with Tesla claiming it'll have Optimus in its own factories by end of 2026. The race isn't to profitability. It's to whoever can deploy enough units to make the unit economics work before the funding runs out. Agility's head start matters less if Figure ships a robot that doesn't need to power down every time a forklift driver walks past.
The Implication
Watch what Agility does with that $300M in orders. If those are conditional purchase agreements tied to performance benchmarks, the IPO story gets shakier. If they're firm commitments from enterprise buyers willing to pay for robots that crouch and power down, that's signal that someone believes the category timing is real even if the margins aren't.
The humanoid buildout isn't going to look like software scale. It's going to look like the early auto industry, capital-intensive and bloody, with most companies dying before the survivors reach breakeven. Agility has deployment lead and customer relationships. Whether that's enough depends on how fast they can make robots that don't need to play dead around humans.