Robotics Weekly: Nvidia's Isaac GR00T, Sam Altman's Stealth Bet, and Waymo's New Ojai
This week's digest covers Nvidia and Unitree's new Isaac GR00T reference design, Sam Altman's investment in the stealth startup Alfred, and Waymo's rollout of the purpose-built Ojai robotaxi. We also explore Wayve's new research lab, Serve Robotics' expansion into laundry services, and a study on robotaxi traffic impacts.
Podcast В· 3 min
Nvidia and Unitree Launch Isaac GR00T Reference Design
Nvidia has partnered with Unitree and Singapore's Sharpa to introduce the Isaac GR00T Reference Humanoid. This platform integrates a nearly six-foot Unitree H2 Plus body with Nvidia’s Blackwell-powered Jetson Thor compute module and Sharpa’s tactile hands, all running on an open-source AI stack. The design is intended to provide a standardized hardware and software foundation for researchers. By offering a unified reference design, Nvidia aims to standardize the development process, allowing researchers to compare results on equal footing rather than reinventing hardware stacks. The initiative includes support from major institutions like Stanford, ETH Zurich, and UC San Diego, with units expected to ship in late 2026. This move positions Nvidia as the critical infrastructure layer for physical AI, mirroring its dominant role in cloud AI training. By anchoring the platform to its own silicon and software ecosystem, Nvidia is effectively setting the standard for the next generation of humanoid research and deployment.
Sam Altman Backs Stealth Robotics Startup Alfred
OpenAI CEO Sam Altman is reportedly backing Alfred, a nine-month-old stealth startup based in Hawthorne, California. The company is developing software tools designed to accelerate the R&D cycles for robots and electric vehicles, addressing the complex, manual engineering processes that currently slow down development. The startup is led by former Tesla designer Ankit Ukil and ex-Meta Reality Labs engineer Dömötör Gulyas, with a team comprised of veterans from Tesla, Ford, and Honda. Alfred focuses on automating the 'grunt work' of engineering, aiming to provide tools that help manufacturers close the feature gap with advanced competitors. While the company is still in stealth, it is already in talks with automakers, defense firms, and robotics companies. The investment highlights the growing trend of capital flowing into the software layer of physical AI, as investors look beyond hardware to the tools that will enable the next wave of robotics innovation.
BYD Confirms In-House Humanoid Development
Chinese EV giant BYD has confirmed it is developing humanoid robots in-house, with plans to potentially utilize its global dealer network for future sales and demonstrations. The company's 'Yao‑Shun‑Yu' program has reportedly reached a seventh-generation prototype, with approximately 150 units currently undergoing trials on its production lines. BYD views automotive AI and robotics as sharing the same foundational technology. The company intends to operate an open platform, building its own machines while also hosting partners, similar to its supplier-heavy model in the automotive sector. It has already secured strategic stakes in robotics firms like PaXini and AgiBot. If successful, BYD's manufacturing scale and existing retail footprint could allow it to commercialize humanoids rapidly. While the dealer network strategy remains speculative, the company's ability to integrate robots into its own factories provides a significant testing ground for real-world deployment.
Waymo Introduces Ojai Robotaxi
Waymo has begun giving select riders access to the Ojai, a new purpose-built electric minivan robotaxi developed in partnership with Chinese EV maker Zeekr. The vehicle is designed to improve fleet economics and scalability, featuring a low step-in height, flat floor, and sliding doors for better passenger accessibility. The Ojai runs on Waymo’s sixth-generation self-driving stack, which includes 13 cameras, four lidar units, and six radar sensors. While the vehicles are manufactured in China, Waymo has confirmed that all Chinese-connected software is removed before import. This launch comes as Waymo seeks to overcome operational challenges, including service suspensions in various cities due to weather-related issues. A purpose-built platform that can handle tougher conditions is a strategic necessity for the company to bridge the gap between the promise of robotaxis and reliable, large-scale commercial reality.
Study Finds Robotaxis Do Not Reduce Traffic Congestion
A new two-year study of Waymo’s California operations suggests that robotaxis do not significantly reduce urban traffic congestion. Analyzing roughly 14 million trips and 86 million vehicle miles, researchers found that only 54% of those miles carried a passenger, indicating a persistent 'empty-mileage' problem similar to that of traditional ride-hailing services like Uber and Lyft. The study highlights that because many robotaxi trips are short and often replace walking, biking, or public transit rather than private car ownership, their net effect on traffic load is minimal. The findings mirror earlier data on human-driven ride-hailing, suggesting that autonomy alone is not a panacea for urban gridlock. This data poses a challenge for regulators and operators. Without policy interventions such as caps on deadheading or requirements for higher occupancy, cities risk locking in a system that increases vehicle miles traveled under the guise of high-tech efficiency.
Serve Robotics Expands into Laundry Services
Serve Robotics has launched a pilot program in Los Angeles to use its sidewalk delivery robots for laundry services, partnering with the on-demand service NoScrubs. The pilot aims to increase the utilization of the company's fleet during the 'dead hours' between meal delivery rushes. The service leverages Serve's existing fleet of approximately 500 robots in Los Angeles, allowing the company to generate additional revenue without the need for new hardware. This move marks Serve's first expansion beyond food delivery, serving as a test case for potential future applications in dry cleaning, pharmacy, and grocery logistics. As Serve continues to burn cash, optimizing unit economics through higher robot utilization is critical. The company faces competition from rivals like Starship, Nuro, and Kiwibot, making the success of this diversification strategy important for proving that the sidewalk delivery model can scale beyond food.
Wayve Launches Dedicated Robotics Lab
UK-based autonomous driving startup Wayve has launched Wayve Labs, a new research unit focused on embodied AI beyond automotive applications. Led by former Microsoft researcher and Wayve chief scientist Jamie Shotton, the lab will explore frontier topics such as spatial reasoning, causality, and risk-aware decision-making in physical environments. While there are no immediate plans for commercialization, the lab is actively hiring and building a team to leverage Wayve's existing real-world driving data and compute resources. The initiative follows a $1.5 billion funding round earlier this year backed by Microsoft, Nvidia, Uber, and Mercedes-Benz. Wayve's strategy differs from competitors like Waymo and Tesla, which are building closed systems. By positioning its embodied AI stack as potential infrastructure for other robotics companies, Wayve Labs aims to create a foundational technology layer that could power a wide range of future robotic applications.
Exoskeleton Simulates Weightlessness on Earth
Researchers from DFKI and the University of Duisburg-Essen have developed an AI-driven robotic exoskeleton designed to simulate microgravity conditions on Earth. The suit estimates the wearer's arm weight and applies counterforces to mimic the sensation of weightlessness, allowing astronauts to train for precision tasks required on the Moon and Mars. The system was tested during Airbus Zero-G parabolic flights, where volunteers performed tasks on touchscreens during 22-second microgravity arcs. Half of the participants trained in the exoskeleton for a month, providing researchers with data on how simulated weightlessness compares to real-world conditions. Beyond astronaut training, the technology has potential applications in healthcare as a low-cost neurotech tool. It could assist stroke patients in relearning fine motor skills, demonstrating how specialized space robotics research can yield practical benefits for terrestrial rehabilitation.