NVIDIA paid $12.9 billion to acquire the parent company of Pollen Robotics, the Bordeaux-based startup known for its Reachy telepresence humanoid robot, according to details surfaced in IEEE Spectrum's weekly robotics video roundup. The acquisition, announced this week, represents NVIDIA's largest robotics investment to date and signals an aggressive vertical integration strategy in humanoid development. While NVIDIA has supplied compute platforms to humanoid manufacturers including Figure AI, Agility Robotics, and Apptronik, owning hardware design assets allows the company to define reference architectures rather than simply enabling them. The parent company name was not disclosed in the initial reporting, but the transaction closed in August 2026.
Pollen Robotics shipped its first Reachy units in 2021 as an open-source telepresence platform targeting research institutions. The robot features expressive antenna-style appendages above its head and human-scale arms capable of bimanual manipulation tasks. Unlike Boston Dynamics' Atlas or Figure's 01, Reachy was never intended for autonomous warehouse work or outdoor mobility. Instead, Pollen positioned it as a teleoperated avatar for remote collaboration, with universities purchasing units for human-robot interaction studies. The company raised approximately €3.5 million in seed funding from European investors in 2020, though it never achieved the scale of U.S.-based humanoid ventures. What made Reachy valuable to NVIDIA was not its commercial traction but its hardware design philosophy: modular, open-architecture, and built around standardized compute modules that could easily integrate NVIDIA's Jetson platform.
The timing aligns with NVIDIA's GR00T announcement earlier in 2026, a foundation model designed to give humanoid robots generalized physical intelligence. GR00T requires massive compute both for training and inference, and NVIDIA has repeatedly stated that humanoid robotics represents a trillion-dollar opportunity over the next decade. But foundation models alone do not create a robotics business. Tesla demonstrated this with Optimus, where controlling both the AI stack and the physical platform allowed faster iteration than any partner-dependent approach could achieve. NVIDIA lacks Tesla's vertical manufacturing capacity, but acquiring Pollen provides a design studio and engineering team with four years of experience building articulated humanoids. The $12.9 billion price tag suggests NVIDIA sees the acquisition as more than an acquihire. For context, Amazon paid $1.7 billion for iRobot in 2022, and that deal included a consumer product shipping at scale. Pollen's revenue was negligible by comparison, implying NVIDIA valued the team, IP portfolio, and strategic positioning rather than existing cash flow.
Industry observers note that NVIDIA's move mirrors strategies from the autonomous vehicle sector, where Waymo and Cruise both moved toward tighter hardware-software integration after years of partnership models failed to deliver commercial outcomes at speed. Humanoid robotics faces similar challenges: mechanical design, actuator selection, power distribution, thermal management, and sensor fusion all constrain what any AI model can achieve. Companies that own the full stack can optimize across all layers simultaneously. Those that do not must negotiate with hardware partners over every design decision, slowing deployment cycles and fragmenting the ecosystem. NVIDIA's acquisition also complicates relationships with humanoid startups that rely on its GPUs. Figure AI, which raised $675 million in a Series B earlier in 2026, uses NVIDIA hardware extensively but competes directly with any NVIDIA-owned humanoid platform. Whether NVIDIA operates Pollen as an independent reference design lab or pivots it toward proprietary commercial products will determine how the rest of the industry responds. Apptronik, Agility, Sanctuary AI, and others now face a supplier that also competes in their market.
What to Watch: NVIDIA's GTC conference in March 2027 will likely feature updated Pollen hardware running GR00T models, providing the first public demonstration of the combined platform. Separately, track whether Figure AI, Agility Robotics, or Apptronik announce alternative chip partnerships with AMD, Qualcomm, or custom ASIC developers in response to NVIDIA's competitive positioning. The Humanoids Summit in Seoul on September 22-23, 2026, may offer early signals as executives navigate the new supplier-competitor dynamic. Finally, monitor Pollen's hiring activity in Bordeaux and any NVIDIA job postings for robotics mechanical engineers, which would indicate whether the company plans to expand beyond reference designs into volume manufacturing.
