Nvidia expands Halos safety system to robotics as physical AI push grows
The chipmaker's safety platform moves from self-driving cars to warehouse robots and humanoids, with Agility Robotics as the first partner
Nvidia’s stock story is mostly about data centers. Its side quest is about robots that don’t run into people.
On June 22, 2026, the chipmaker unveiled Halos for Robotics at the Automate conference in Chicago. The full-stack safety platform is built to help companies deploy autonomous machines in places full of humans, such as warehouses and factory floors.
What Halos for Robotics actually includes
Think of Halos as a safety kit rather than a single gadget. It bundles AI compute, functional safety hardware and software, sensor integration, simulation and validation tools, and help with certification.
On the hardware side, the core pieces are the NVIDIA IGX Thor, which comes with functional safety built in, and the Holoscan Sensor Bridge. The second one handles getting sensor data where it needs to go, which is half the battle when a machine is trying to figure out whether that shape ahead is a pallet or a person.
The software layer includes Halos Core, Halos OS, and the Halos Outside-In Safety Blueprint. That last one is open source, so developers can inspect and build on it rather than treating it as a sealed box.
Nvidia is also leaning on third-party validation. The Halos AI Systems Inspection Lab is the first ANAB-accredited program focused on where AI safety and functional safety meet in physical applications.
The lab works with established certification names, including TĆV Rheinland and UL Solutions.
Agility Robotics gets first dibs
Agility Robotics is the first partner putting Halos to work. The company is integrating Halos OS into Digit, its humanoid robot, for customer deployments.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The customer list is where things get interesting. Deployments are already in the works for Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada.
From cars to humanoids
Halos did not start with robots. Nvidia first launched the system on March 18, 2025, aimed at autonomous vehicles, with hardware and software tools for building guardrails into self-driving cars.
The robotics version extends that architecture to autonomous mobile robots in warehouses and humanoid robots. Nvidia says it draws on more than a decade of work in autonomous vehicle safety.
The bigger shift is philosophical. Traditional industrial safety tends to be rigid and rule-based: fence off the robot, stop everything if someone crosses a line on the floor. Halos is pitched as a move toward adaptive, AI-driven safety, designed to keep machines productive in messy, changing environments while still meeting safety standards.
The money behind the machines
Nvidia’s multi-trillion-dollar market capitalization is mostly a bet on AI data centers. Physical AI is the company’s attempt to make sure the next wave of demand also runs on its chips.
Barclays estimates that revenue from humanoid robots could reach $200 billion by 2035.
What this means for Nvidia and the robotics market
For Nvidia, Halos is a classic platform move. If robot makers build on Nvidia’s safety stack, they are also likely to build on Nvidia’s compute, such as IGX Thor.
Investors should watch a few signals. The first is how quickly the Agility Robotics deployments with Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada move from planned to operating at scale.
The second is whether other humanoid and mobile robot makers sign on after Agility. One partner is a launch. Several partners would look more like a standard.
The third is regulatory. As humanoid robots enter shared workspaces, certification bodies and safety rules will shape who wins. Nvidia getting its inspection lab accredited early suggests it wants a seat at that table before the rules harden.