Agility's Digit 5 drops the safety cage — read the 2027 timeline

The headline number in humanoid robotics has never been the robot. It is the fence around it. On September 15 Agility unveiled Digit 5, the fifth generation of its industrial humanoid, and the pitch is that it does not need one — the company calls it "cooperatively safe" work at scale. That is a real engineering claim with a real delivery date attached, and the date is the part most coverage buries.

What actually happened

Digit 5 is 284 lb, 5'11", reaches 7.2 feet, and lifts up to 50 lb repeatedly — a 40% payload increase over the prior generation. The battery runs 90 minutes and fast-charges in 9, a 10:1 run-to-charge ratio that Agility says supports more than 20 working hours in a 24-hour day. The Robot Report's teardown covers the new cycloidal leg actuators and the swappable ISO-mount grippers.

The safety architecture is the actual product. Digit 5 adds 360° human detection using visual and auditory sensors, an independent safety controller that can trigger without the main stack agreeing, and a graceful-degradation behavior where the robot sits down and disables its motors rather than risk contact. It has padded knees for that. Agility built it on NVIDIA IGX Thor with the Halos for Robotics platform. Co-founder Jonathan Hurst described the approach as touching "every feature of the robot — electrically, mechanically, sensors, computing."

Now the dates. Early access is the first half of 2027. General availability is the end of 2027, for manufacturing, warehouse and distribution operators, expanding to the EU and UK. Agility reports more than $300 million in multi-year orders as of May 2026, against $1.8 million in 2025 net sales, and is going public via a $2.5 billion SPAC merger with Churchill Capital Corp XI. RoboFab in Salem, Oregon is designed for up to 10,000 units a year. Named deployments so far include GXO, Schaeffler, Amazon and Toyota Motor Manufacturing Canada.

Why this matters for your business

You are not buying a humanoid. Your 3PL might. If you ship physical product through a contract warehouse, this is a 2027–2028 question about your partner's cost structure, not a purchase order you write. The right move now is asking your 3PL what they have piloted and on which SKU profiles — Digit handles totes, not irregular hand-packs.

The fence was the cost, and that generalizes. Caged industrial robots need floor space, guarding, and a re-layout of the aisle. Removing the cage is what turns a robot from a capital project into something you drop into an existing building. The same logic runs through software automation: the integration and the guardrails cost more than the capability, every time.

Read the order book against the revenue. $300M contracted versus $1.8M recognized is a bet on a machine that will not ship in volume until late 2027, with milestones attached. That gap is not a scandal — it is what a pre-revenue hardware ramp looks like. It is also a reason to treat vendor roadmaps as forecasts and to avoid designing your 2027 operations around a robot in early access.

Narrow and boring still wins. 50 lb totes, moved repeatedly, in a lit warehouse, for 20 hours. That is the shape of automation that pays — one task, high volume, tolerant of a machine that sits down when confused. Pick your own automation targets the same way.

Sources: Agility, The Robot Report.