Robotics enters post-Jensen Huang era as NVIDIA leadership changes
Days before Christmas in 2023, Serve Robotics was close to bankruptcy. A major investor withdrew from a committed funding round, leaving the sidewalk delivery robot developer unable to make payroll and days from running out of cash.
With few options, co-founder and CEO Ali Kashani contacted NVIDIA co-founder Jensen Huang. Kashani recalled that they spoke at 6 a.m. the next day, and that he had dinner with Huang’s executive team that evening to find a bridge. Within days, Huang led a financing round that revived the company; four months later, Serve rang the Nasdaq bell for its public listing. Kashani said he had heard similar stories from other founders and current and former NVIDIA employees whose lives Huang’s actions changed.
For years, few outsiders did more than Huang to raise robotics’ profile. In his signature black leather jacket, he routinely devoted portions of marathon keynotes to robots, showcasing humanoids, introducing development tools, highlighting startups, championing open models and research, and arguing that robots would transform the world. He also put NVIDIA’s money behind that vision.
Now that advocacy has ended inside NVIDIA. The board accepted Huang’s resignation on Nov. 12, and CFO Colette Kress is interim CEO while the company searches for a permanent successor. NVIDIA said in its third-quarter earnings call that physical AI revenue nearly tripled year over year and is on track to exceed $25 billion in fiscal 2027. That is sizable and fast-growing, but it is still only about 6% of projected annual revenue above $400 billion, and physical AI includes more than robotics.
The attention Huang gave robotics was disproportionate to its near-term commercial contribution. Deepak Jayaraj of Four Growers said NVIDIA introduced Isaac in 2017 and Huang kept robotics in keynotes and the long-term story year after year, making it easier for founders to pitch investors. Unlike Tesla CEO Elon Musk, whose advocacy centers on his own products, Huang promoted an ecosystem of developers, many competing directly. NVIDIA benefited whichever companies succeeded, but the model also let Huang advocate for the broader industry.
NVIDIA’s lasting contribution may be development infrastructure built under Huang: GPUs and CUDA for AI training; Omniverse and Isaac Sim for simulation; Isaac Lab for robot learning; GR00T and Cosmos foundation models; Isaac ROS; and Jetson embedded computing. Jan Zizka of Brightpick and Photoneo said Photoneo joined an early-access program for Jetson TX1 and that Huang was almost pushing the market to make it happen, investing before demand fully materialized. Zizka said open research and software lowered barriers, while embedded computing efforts encouraged AMD and Qualcomm to pursue robotics. NVIDIA also became a prominent strategic investor in Figure AI, NEURA Robotics, Skild AI, Wayve, Waabi and Serve Robotics; beyond capital, its backing was an endorsement.
Inside NVIDIA, Huang’s enthusiasm gave robotics teams an advocate at the highest level. His credibility from previous long-term bets gave him latitude a successor may lack. The next CEO inherits growing data-center competition and potentially greater political and regulatory scrutiny after the Democrats’ strong midterm performance. Robert Little, former CEO of ATI Industrial Automation, worries that opposition to data-center construction could affect NVIDIA’s ability to fund longer-term investments. He said robotics is propped up by data-center money; if growth slows, NVIDIA could cut anything not making money. Conversely, pressure could make diversification more important. A pending $12.9 billion acquisition of Hugging Face, whose LeRobot initiative is important to robotics developers, could signal priorities. NVIDIA did not respond to requests for comment.
Little nevertheless believes robotics has enough momentum to withstand any individual’s departure, saying the flywheel has been created and is spinning. Google DeepMind, Physical Intelligence, Generalist, Skild AI and others are developing robot intelligence; semiconductor companies are challenging NVIDIA in embedded computing. A less aggressive NVIDIA could benefit competitors, and departing researchers would likely spread expertise. Tenzer said the foundations are there and things are accelerating. Zizka called the change a negative and said the best case is continuity.
The bigger test may come in 12 months, Tenzer said. Many humanoid developers raised substantial funding on expectations that AI advances will unlock multi-purpose humanoids for industrial use. He said the next year may be fine, but many companies will then struggle to show actual use cases and real business. If commercial progress falls short, enthusiasm could dissipate. Tenzer said somebody needs to get on stage, show cool things and have the charisma to do it—where Jensen will be missed. Jayaraj said funding is where he would watch first, especially for humanoid companies whose business case depends on breakthroughs years out. Huang’s advocacy gave those bets credibility, he said; founders should ask how much of their plans rest on informal support or tools that have not shipped, because such support can walk out the door with one person.