Unveiling Nvidia’s Strategic Investments in Robotics: Jensen Huang Promises Major Breakthroughs in Three Years

NVIDIA’s Investment Landscape in Robotics: Jensen Huang Predicts Major Breakthroughs Within Three Years

At the recent GTC 2026, NVIDIA’s presence in the robotics sector remained highly prominent. Jensen Huang boldly stated that the company aims to generate $1 trillion in revenue by next year, while also noting that robotics represents a $50 trillion manufacturing market. NVIDIA has been deeply involved in this field for a decade, showcasing 110 robots at the conference, with nearly every robot company globally collaborating with NVIDIA. In Huang’s latest group photo of robots, over 30 machines were featured, many of which hail from China, including companies like BYD, KUKA, ZhiYuan, Xiaopeng, and Geely.

Looking back at the development of embodied intelligence in recent years, NVIDIA has strategically positioned itself across various areas such as foundational technology platforms, investments, and deep collaborations, ensuring that any company venturing into humanoid robotics cannot overlook its influence. According to a partial count by the Robotics Forward-looking, NVIDIA has invested in at least 13 robotics-related companies in recent years. Among these, NVIDIA’s early investments included 9 companies focusing on specific applications like weeding, surgery, and manufacturing, while more recent investments have targeted 7 companies in the embodied intelligence domain.

Huang has never shied away from expressing his enthusiasm for robotics. He has explicitly stated that robotics will become NVIDIA’s “second growth curve” following AI, mentioning three types of robots expected to achieve large-scale production: automobiles, drones, and humanoid robots, with humanoid robots projected to have the highest output. This comprehensive approach, spanning technology research, industry collaboration, and ecological investment, not only demonstrates NVIDIA’s commitment to the robotics arena but also reveals its ambition to set the rules for the Physical AI era.

1. Early Investments in Niche Areas to Identify Potential Robotics Players

NVIDIA has strategically locked in several potential players in niche robotics markets through early investments, particularly focusing on domains with clear applications such as weeding, surgery, mobile delivery, and industrial manufacturing. The financing has mainly targeted B-D rounds, indicating that NVIDIA prefers to accelerate investment when companies are relatively stable and have clear commercialization paths, acting more as an “amplifier” rather than taking on the uncertainties of early-stage technology.

Some of the robotics companies NVIDIA has invested in include:

• Serve Robotics: Focused on developing AI-driven autonomous sidewalk delivery robots, resembling small yellow figures. As of last December, over 2,000 units of their third-generation sidewalk delivery robot have been deployed.
• Outrider: Specializes in automated delivery, enhancing operational efficiency across the supply chain.
• Machina Labs: Integrates AI and robotics to quickly produce advanced composite materials and metal products, featuring a manufacturing platform called Robotic Craftsman.
• Moon Surgical: This surgical robotics company primarily focuses on laparoscopic procedures.
• Neocis: Concentrates on robots for dental implant surgeries, with its core product Yomi assisting in pre-surgical planning and navigation.
• Carbon Robotics: A unicorn in laser weeding robots, with plans to expand beyond agriculture in the future.
• Bedrock Robotics: Developed the Bedrock Operator system, turning existing engineering equipment into autonomous machines with precision and safety.
• Bright Machines: Provides software-defined smart manufacturing solutions, offering flexible automation support across product lifecycles in various sectors.
• Oxa: A UK company specializing in autonomous driving software for industrial scenarios in logistics, airports, and ports.

2. Betting on Leading US Embodied Companies, Rarely Taking the Lead

In recent years, NVIDIA has invested in several leading US embodied intelligence companies, particularly favoring those related to embodied models. However, NVIDIA’s role in these investments has mostly been as a non-leading participant. The rationale behind this strategy is clear: NVIDIA emphasizes its position as a foundational ecosystem provider, primarily focused on delivering technological infrastructure rather than deeply involving itself in the governance and operations of the invested companies. Additionally, similar to its investment logic in the AI field, NVIDIA often forms technological and commercial partnerships with the startups it invests in.

Some of the robotics companies in this category include:

• Figure: Launched Figure03 last October, with its first “dual-system” VLA model Helix upgraded to Helix02 this January, expanding from upper body control to full-body autonomy.
• Wayve: An autonomous driving company taking an unconventional approach, working on embodied intelligence and end-to-end deep learning models.
• Generalist AI: Focuses on developing embodied foundational models targeting dexterous operations for robots.
• Agility Robotics: Introduced the Digit humanoid robot designed for logistics and manufacturing, which has been implemented in various factories like Amazon and Schaeffler.
• Skild AI: Developed the Skild Brain, adaptable to different robot forms, capable of quickly reconstructing movement strategies even when robots are damaged.
• Field AI: Aims to develop an embodied intelligence brain for robots, with its Field Foundation Models (FFMs) designed for risk perception.
• Dyna Robotics: Released the DYNA-1, the first all-weather, efficient, and stable autonomous dexterous operation model.

3. Focusing on Simulation to Propel AI into the Physical World

Since 2018, NVIDIA has entered the field of embodied intelligence with a clear goal: to create a foundational ecosystem covering the training, testing, and deployment of robots. On the hardware front, NVIDIA has introduced edge chip modules like Jetson Orin and Jetson Thor, catering to various computational needs for service robots and humanoid robots, establishing itself as the “physical brain” of robotics. Building on this computational foundation, NVIDIA has constructed a complete link for robot development and training through two core technology platforms: the Isaac platform, a comprehensive AI robot development platform, and Omniverse, a high-fidelity virtual physical world providing scalable training environments.

The breakthroughs in foundational models represent NVIDIA’s core competitiveness. The Cosmos foundational model platform allows AI to learn the physical causal laws of the world, predicting future scenes from current frames. The Project GR00T initiative focuses on developing foundational AI models for humanoid robots and has already reached GR00T N1.7. At GTC 2026, Huang previewed the upcoming GR00T N2, a next-generation foundational model based on DreamZero research, expected for release by the end of this year. This model, built on a new world action model architecture, significantly enhances the frequency at which robots successfully complete new tasks in unfamiliar environments, outperforming the VLA model by over two times.

Supporting this trajectory is NVIDIA’s GEAR research lab, led by Jim Fan and Yuke Zhu, dedicated to building foundational models for embodied agents in both virtual and physical worlds. The lab focuses on four research areas: multimodal foundational models, general-purpose robot models, virtual world agents, and simulation synthetic data. Since February of this year, GEAR has introduced several groundbreaking achievements:

• EgoScale framework: By utilizing over 20,000 hours of self-centered human data, it breaks the scale bottleneck in human-robot dexterous operation transfer.
• DreamDojo: Trains a universal robot world model using a massive database of 44,000 hours of first-person human videos, enabling robots to possess controllable “imagination.”
• DreamZero: Based on a 14-billion-video generation model and 500 hours of teleoperated data, it guides robot action generation by predicting future video frames, achieving zero-shot completion of unseen tasks.

NVIDIA’s strategy in robotics follows a familiar path from its AI endeavors: leveraging larger, longer, and more diverse data sets as fuel, supported by stronger computational power and engineering capabilities, gradually integrating previously scattered key technological modules into a more universal model system.

4. Deeply Engaging with Industry Leaders to Accelerate Application Scenarios

Beyond the technological closed loop, NVIDIA has already accelerated the application of its robotics technology across various scenarios through deep integration with industry leaders. During his GTC 2026 speech, Huang mentioned that NVIDIA provides three types of computing platforms for robot manufacturers, along with open models, libraries, and frameworks for flexible use. Numerous top robotics companies are building on these platforms: ABB, FANUC, and KUKA collectively account for nearly half of the global industrial robot installations and have integrated the Omniverse library into their robot simulation tools. Companies like Figure, ZhiYuan Robotics, and 1X are employing Isaac Lab, Newton, and Cosmos simulation libraries, utilizing Jetson and Thor for edge inference. AI-native companies such as Skilled AI and Field AI are also building their general-purpose robot brains on NVIDIA’s Isaac and Cosmos technology stack.

NVIDIA is currently the only company on which every robotics firm builds its computing platform. In addition to robotics companies, NVIDIA has partnered with industry leaders like Foxconn to explore the application of humanoid robots in manufacturing. At GTC 2026, Foxconn showcased humanoid robots developed in collaboration with NVIDIA for industrial applications, demonstrating their ability to perform high-precision, repetitive industrial tasks such as picking, installation, and material handling. Earlier in 2025, Foxconn’s chairman Liu Yangwei publicly stated that they are working together to develop humanoid robots in Kaohsiung, Taiwan, with plans to apply them in healthcare and service sectors. By late October 2025, Foxconn announced at the NVIDIA GTC conference that it would deploy humanoid robots at its Houston factory for the production of NVIDIA’s AI servers, with the collaboration expected to commence in the first quarter of 2026. The humanoid robots come in both bipedal and wheeled configurations to suit different production environments. Huang previously predicted that widespread adoption of humanoid robots in manufacturing would occur within five years, and over the next decade, more manufacturing companies may follow in Foxconn’s footsteps.

Conclusion: Extensive Investments Combined with Hands-On Involvement, Positioning NVIDIA as a New Infrastructure Provider for the Robotics Industry

While NVIDIA’s investments in AI cover various vertical applications, its strategic ambition in the robotics sector is evidently more grandiose and resolute. NVIDIA not only strategically identifies potential players in the global embodied intelligence landscape through precise investments, building a vast ecological collaborative network, but also chooses to engage directly in the field, comprehensively establishing the core technological infrastructure of the robotics industry. By positioning itself as a new infrastructure provider, NVIDIA aims to lead the rules and developmental direction of the Physical AI era. Although the robotics business currently represents a relatively small scale for NVIDIA, the company is accelerating its progress toward powering billions of robots and hundreds of thousands of robotic factories in the future.