Newshot After AMD's Acquisition of World Labs, 李飞飞 and 苏姿丰 Share a Stage for the First Time: Robots Will Become a "New Engine of Compute," with a Market Potentially Worth Hundreds of Billions of Dollars in the Coming Years
IT Home, November 12: On November 12 local time, at the CoRL 2026 Conference on Robot Learning, World Labs founder and CEO 李飞飞 and AMD Chair and CEO 苏姿丰 held a conversation, exchanging views on topics including physical AI, robotics development, compute demand and future model roadmaps.
李飞飞 believes robots will drive computing power into its next stage of development; 苏姿丰 said physical AI, especially the robotics field, could become a market worth hundreds of billions of dollars in the coming years.
The conversation was moderated by CoRL conference organizer Peter Stone, and was the first public joint appearance by 李飞飞 and 苏姿丰 after AMD announced its acquisition of World Labs.
According to the official CoRL 2026 agenda, 李飞飞 also delivered a keynote titled "Atlas: A Unified World Model for Generation and Reconstruction" that day, presenting World Labs' latest research results in the world model field.
Speaking about the robotics market outlook, 苏姿丰 said artificial intelligence is still in its early stages, and physical AI, especially robotics, holds enormous opportunity, with the market potentially reaching hundreds of billions of dollars in the coming years. She believes robots need to truly understand physical laws, models and their surrounding environment, and that coordinated optimization across hardware, software and systems will play an important role.
苏姿丰 also said AMD already has a sizable physical AI and robotics research team, though the related efforts are still at an early stage. AMD already has technologies such as embedded FPGAs, processors and AI engines that can support equipment such as industrial robots. In the future, the company hopes to further strengthen cooperation with universities, startups and research institutions to advance the robotics technology ecosystem.
On the direction of AI model development, 李飞飞 put forward the view that "AI is compute, and compute is AI." She believes a convergence trend is emerging among world models, language models and multimodal models, and that hardware and software also need to evolve together.
Discussing the application of large language models and vision-language models in robotics, 李飞飞 said these models will play an increasingly important role, but the systemic problems in robotics cannot yet be considered solved. Robots need to interact directly with the physical world, and there are still many technical issues to explore.
苏姿丰 agreed, saying different application scenarios require different types of models, and no single model or single technical approach can solve all problems. In some scenarios, frontier large language models may perform well, while in others, specialized models may have more advantages.
Compute demand was also an important topic of the conversation. 苏姿丰 believes robotics development will further drive demand for different forms of computing power, including cloud compute for model training, enterprise internal computing resources, edge computing and computing devices deployed directly in physical environments.
She noted that the future will require computing power of different scales and forms, with overall efficiency improved through coordinated optimization of hardware, software and systems. As robots increasingly enter real-world environments, metrics such as low power consumption, high efficiency and low latency will also become more important.
李飞飞 emphasized that robots will drive computing power into its next stage of development, and that this demand exists not only on the model training side but will also extend to edge computing. She believes robotics development has faced many difficulties in the past because devices must understand and interact with the physical world, and current AI technology advances are bringing new opportunities to this field.
The two also discussed the issue of academic research institutions obtaining compute resources. 李飞飞 believes that at this critical period of rapid AI development, it is very important to provide sufficient computing resources to the public sector and academic researchers. She mentioned the previously promoted U.S. National AI Research Resource (NAIRR) Act, hoping to support researchers in AI evaluation, benchmarking and other work through public resources.
苏姿丰 agreed, and said AMD is building computing capacity on the scale of thousands of megawatts. She believes industry companies should devote some computing resources to long-term research and open them to universities and the next generation of innovators, helping more researchers participate in technology development.
In addition, 苏姿丰 revealed that AMD already has a physical AI and robotics research team, and hopes to further strengthen related research with the addition of 李飞飞 and World Labs, and to establish connections with the entire robotics ecosystem.
The conversation also touched on AI energy consumption and model efficiency. 苏姿丰 believes reducing power consumption, improving efficiency, extending service life and lowering latency are all directions that future computing technology needs to continuously optimize. 李飞飞 said the human brain's high energy efficiency is the result of billions of years of evolution and cannot simply be compared with the training energy consumption of current AI models, while AI is also not necessarily just about replicating the human brain.
Related Articles
- Report says AMD gaming GPU supply will drop by at least 20%-50% in the fourth quarter, with shortages lasting until January next year
- AMD "lightens the load" on server memory, with DRAM demand potentially reduced by up to 75%
- Xbox version of "GTA 6" preload files show clues of AMD FSR 4.1, which may not be supported at launch
- AMD announces the launch and shipping of EPYC 9006 SP7 processors, supporting 16-channel DDR5 memory
- Samsung and AMD upgrade semiconductor test systems through cooperation, introducing Zen6-architecture EPYC Embedded 9006 Series SP7 processors
- Microsoft Project Zenith officially launches, bringing an out-of-the-box programming experience to AMD Ryzen AI Halo