AMD, Google, China Race Nvidia for Humanoid Robot Brains

A Widening Battle Over the Silicon and Software Inside Humanoid Robots

The race to build humanoid robots is no longer just about mechanical arms and legs — it's increasingly about who supplies the chips and the underlying "brains" that let these machines see, reason, and act. A cluster of recent announcements shows Nvidia's dominance in AI hardware and robot foundation models facing fresh challenges from AMD, Google DeepMind, and ambitious national projects in China and Japan.

AMD Muscles Into Nvidia's Territory

Foundation Future Industries, the robotics venture backed by Eric Trump, announced that its upcoming MK-2 Phantom humanoid robots will run on AMD Ryzen processors rather than Nvidia hardware 1. The company said the partnership with AMD is aimed at co-developing autonomous humanoid robots intended for both military and industrial applications 3. Because Nvidia currently supplies the bulk of AI chips used across the robotics and broader AI industry, Foundation's choice to build around AMD silicon is being read as a direct challenge to that dominance 1. The deal underscores how chipmakers are jockeying for position in a market that is still nascent but widely expected to expand rapidly as humanoid robots move from labs and demos toward real-world deployment in factories, warehouses, and defense settings 3.

Google DeepMind Expands Its Robotics Footprint

While AMD is targeting the hardware layer, Google DeepMind is pushing deeper into the software and foundation-model side of embodied AI. German robotics company Agile Robots has become the latest firm to partner with DeepMind, agreeing to integrate the lab's robotics foundation models into its own robots 2. In exchange, DeepMind gains access to real-world operational data collected by Agile Robots' machines, which can be used to further refine its models 2. This arrangement reflects a broader pattern in the industry: AI labs are trading their foundation models for the physical-world data that robotics companies generate, creating a feedback loop that improves both the software and the fleets that run it. Agile Robots joining DeepMind's growing list of partners signals that the fight for influence over humanoid and industrial robots is playing out not just in chips, but in whose neural networks actually control the robots' behavior 2.

China's Robbyant Targets Spatial Perception

In China, embodied AI company Robbyant has introduced two new models — LingBot-Depth 2.0 and LingBot-Vision — designed specifically to sharpen robots' spatial perception and visual understanding 4. These tools address one of the more stubborn technical hurdles in embodied AI: giving robots a reliable sense of depth, distance, and surroundings so they can navigate and manipulate objects safely in unstructured environments. The launch fits into China's broader push to build homegrown foundation models for physical AI, positioning domestic firms like Robbyant as competitors to Western labs working on similar perception and reasoning systems for robots 4.

Japan's Noetra Frames Itself as a National Imperative

The stakes are being framed in even starker terms in Japan, where government-backed startup Noetra is developing a foundational model for physical AI and robotics 5. Its CEO described the effort as Japan's "last chance" to strengthen its domestic supply chain for critical robotics and AI technology, a comment that reflects growing anxiety in Tokyo about falling behind the United States and China in foundational AI capabilities 5. Noetra's government backing highlights how humanoid robotics and embodied AI are increasingly viewed as matters of national technological competitiveness, not just commercial opportunity, echoing similar strategic pushes seen in Beijing and Washington 5.

Why This Matters

Taken together, these developments show an industry fragmenting on multiple fronts at once. On the hardware side, AMD's entry into Foundation's humanoid robot program challenges the assumption that Nvidia will automatically power every major robotics project 13. On the software side, Google DeepMind is embedding its foundation models into more partners' robots, trading intelligence for data in a way that could entrench its position much as Nvidia has entrenched its own 2. Meanwhile, China and Japan are each pursuing their own foundation models and perception systems, driven partly by commercial ambition and partly by a desire for technological self-sufficiency 45. As humanoid robots inch closer to commercial and military deployment, the question of which chips, which models, and which national ecosystems end up controlling their "brains" is becoming one of the more consequential competitions in the broader AI landscape.