Amazon unveils Vulcan, its first robot that can actually feel

If you drop a coin, you pick it up without looking twice. Your fingers feel the edge, gauge the pressure, and slip it into your palm. For an industrial robot, that basic task has been a stubborn wall for decades.

Most warehouse robots are completely numb. If an arm runs into something unexpectedly, it either hits an emergency stop or rams right through the obstacle and breaks it, simply because it cannot feel what it touches.

Amazon thinks it has cracked this with Vulcan. Unveiled at its Delivering the Future event in Dortmund, Germany, Vulcan is the company's first warehouse robot equipped with a physical sense of touch.

A ruler taped to a hair straightener

Amazon fulfillment centers store millions of items in fabric pods divided into one-foot-square compartments. Each bin holds around ten mismatched products. Trying to slide a new box into that crowded space without crushing anything around it has always demanded human dexterity.

Vulcan tackles this with an end-of-arm tool that looks a bit like a flat ruler mounted to a hair straightener. The flat paddle shoves existing items aside to carve out a pocket of space inside the compartment.

Then, the two arms of the straightener clamp the new item, using force-feedback sensors to squeeze just hard enough without damaging the package. Built-in mini conveyor belts gently feed the item into the slot. For picking items out, Vulcan switches to an arm with stereo cameras and suction cups, actively monitoring the bin to ensure it doesn't accidentally drag neighboring items out with it.

Taking the ladder shifts

Across a typical Amazon warehouse with one million distinct products, Vulcan can handle about 75% of inventory types. It operates at speeds comparable to human workers and can run for up to 20 hours a day.

Instead of overhauling the whole floor, Amazon is using Vulcan for the worst ergonomic spots. In facilities across Spokane, Washington, and Hamburg, Germany, the robot handles the top rows of pods eight feet off the ground, as well as the bottom bins near the floor. Workers no longer need to haul step ladders up and down all day, and if Vulcan hits an awkward item it cannot grasp, it flags a human coworker to step in.

Amazon has already used Vulcan to process hundreds of thousands of orders, and plans to roll the system out widely across the US and Europe as part of a broader ten-billion-euro European modernization push.

Why it matters

Warehouse automation has never been blocked by a lack of vision models. The bottleneck is the messy physical world, where soft socks, toothpaste tubes, and fragile electronics sit jumbled in the same fabric box.

We think Amazon made a very clever engineering bet here. Rather than getting distracted by five-fingered humanoid hands, it stuck force sensors onto a dead-simple paddle-and-belt tool. It is cheap, durable, and handles three-quarters of real warehouse inventory right now.

This also marks a shift in how industrial AI gets trained. You cannot simulate the tactile friction of random packaging inside a virtual environment. Vulcan was trained on actual physical force data and learns from its own mechanical mistakes, which looks like the blueprint most commercial robotics will have to follow.

The elephant in the room is labor. Amazon frames Vulcan as an ergonomic assistant that saves workers from climbing ladders and points to new maintenance roles. But a machine that handles 75% of items, matches human speed, and runs 20 hours a day inevitably changes the math. As Vulcan's tactile models improve, the need for frontline warehouse hands will shrink.