Agility Robotics has made a big, simple bet on the future of humanoid robots: one brain, one safety system, and a clear path to the factory floor. On October 22, 2026, the company announced that its next humanoid, Digit 5, will standardize on NVIDIA IGX Thor for its onboard compute. For the general-availability version of the robot, NVIDIA becomes the exclusive provider of that onboard computer.
That may sound like a parts-supplier announcement. In humanoid robotics, it is much more than that. The computer decides how fast a robot can see, understand, plan and move — and whether it can do all of that safely next to people.
One computer for the whole robot
Digit 5's new architecture unifies its onboard compute on IGX Thor modules. Instead of mixing different computers for different jobs, Agility gets a common foundation across the robot. NVIDIA describes IGX Thor as a module built on its Blackwell chip design and aimed at industrial machines, running the AI that interprets camera images, plans the next step and moves limbs, all on board rather than relying on a distant data center.
That on-board point matters in a warehouse. A robot that must wait for a cloud server to tell it to stop is a robot with a dangerous delay. Local compute keeps perception and action tight together, which is exactly what a machine working around people and moving goods needs.
The move also echoes a wider shift in physical AI. Rivian's recent push into 3D world models for self-driving, covered in our story on Rivian Acquires Third Dimension AI to Build 3D Worlds for Self-Driving Cars, shows how much of today's robotics progress is about giving machines a reliable model of the world around them. Digit 5's bet is the hardware side of the same idea: give the world model a fast, consistent brain to run on.
Safety is the real selling point
Agility is clear that this deal is about safety as much as speed. The company is integrating NVIDIA IGX Thor and additional NVIDIA Halos models into its proprietary safe human detection system for Digit. In plain terms, that system is meant to notice a nearby person and slow or stop the robot.
Agility was the first company to announce it would incorporate elements of NVIDIA Halos into its own safety system. It will also take part in the NVIDIA Halos AI Systems Inspection Lab, where Digit's safety-related software, AI components and cybersecurity protections will be evaluated ahead of final third-party certification.
That certification step is the bottleneck for humanoids. Factories will not put a large, strong machine next to workers on the strength of a demo video. They need tested safety, documented behaviour and an independent check. By tying compute and safety together under one vendor's platform, Agility is trying to make that certification path shorter and more repeatable.
It is a useful contrast with the tactile side of robotics. Enact Intelligence Raises $30M to Give Robots a Sense of Touch shows how much work is still going into letting robots feel what they handle. Digit 5's story is the companion piece: before a robot can handle things delicately, it must first be trusted not to hurt anyone while it moves.
Why one supplier, not many?
Standardizing on a single compute supplier is a trade-off, and Agility has chosen simplicity. A single, well-tested platform should be easier to build, certify and support at volume than a mix of parts, especially as the company prepares to sell Digit 5 in larger numbers. NVIDIA's Deepu Talla, vice president of robotics and edge AI, put it this way in the announcement: "Bringing humanoid robots to commercial scale requires AI computing and safety to be engineered together."
There is a commercial logic too. NVIDIA already supplies the chips and software stack behind much of today's AI training. If the same company also supplies the robot's brain, Agility's engineers work inside one ecosystem, from simulation and training to the machine on the floor. That is the same scale logic behind NVIDIA and Firebird Bring 45,000 GPUs to Kazakhstan: Inside the Country's AI Factory Bet, where huge GPU deployments are treated as national infrastructure rather than one-off purchases.
For buyers, the question will be support and longevity. NVIDIA says IGX Thor comes with up to ten years of enterprise software support, which matters to factories that plan equipment purchases over years, not months. A robot is not a phone; nobody wants its brain orphaned after two product cycles.
When can you actually get Digit 5?
Agility says early access to Digit 5 is expected to begin in the first half of 2027, with availability increasing over the following six to nine months for manufacturing, warehouse and distribution operators. A separate Digit 5 announcement notes that, for the first time, Agility plans commercial deployment outside North America, starting with the European Union and United Kingdom, with a CE mark and additional regulatory marks to enable market access there.
That timeline is worth reading carefully. Early access is not general availability, and a humanoid that works in one pilot warehouse still has to prove itself across different sites, shifts and tasks. Training grounds matter here too — Industrial Next Built a 10,000-Square-Foot Arena Where Robots Train for the Real World is a reminder that robots, like people, need realistic practice before they are trusted on the job.
What this means for workers
The honest version of this story is not "robots replace everyone." Digit is aimed at physically demanding and repetitive tasks in warehouses, manufacturing facilities and distribution centers, while workers focus on higher-value work, according to Agility. Whether that promise holds depends on how deployments are actually run: who sets the pace, who supervises the robot, and what happens to the jobs around it.
What is clearer is the direction of travel. Humanoids are moving from research labs and trade-show stages into paid, certified, commercial work. The winners in that transition may not be the flashiest robots, but the ones factories can certify, support and trust next to their people. Digit 5's NVIDIA deal is Agility's answer to that test.
Would you work alongside a humanoid like Digit 5? Let us know in the comments.
Comments
Post a Comment