NVIDIA (NVDA-US) CEO Jensen Huang recently predicted at an event that humanoid robots will undoubtedly become mainstream in the future, and their manufacturing costs may be much lower than anticipated.
Huang participated in CadenceLIVE Silicon Valley 2024 and engaged in a discussion with Cadence President and CEO Anirudh Devgan. During the dialogue, Huang noted that the era of accelerated computing has arrived. The fundamental changes and transformations in underlying computing platforms are the foundation of Cadence’s development, upon which everyone relies on Cadence technology. NVIDIA specializes in accelerated computing.
He emphasized that achieving generative AI would be challenging without transitioning to accelerated computing.
Energy Efficiency with AI Huang believes that running tens of thousands of general servers would incur costs 10 to 20 times higher and consume 20 to 30 times more energy. Accelerated computing is indispensable in this scenario. Concurrently, AI technology can also reduce energy costs.
He stated, “AI actually helps save energy. Without the AI models you create, how could we save costs by 6 to 10 times? Once a model is trained, millions of engineers benefit, and over decades, billions of people enjoy reduced costs.”
Huang pointed out that a small part of a program’s code accounts for a significant portion of tool runtime. For instance, Computational Fluid Dynamics (CFD) might use only 3% of the code but represent 99.9% of the runtime, while the remaining 97% could be rewritten using AI and accelerated computing to accelerate applications by 100,000 times.
The Significance of Accelerated Computing Huang also highlighted the importance of accelerated computing in areas such as drug discovery, data centers, smart vehicles, and humanoid robots.
Firstly, in drug discovery, Huang mentioned that over the past 30 years in biomedicine and computational fluid dynamics, AI accelerated computing has been achieved through CUDA and various Domain-Specific Languages (DSL) and specific libraries.
Secondly, for data center computing, Huang believed that by investing in AI and data centers, NVIDIA could design better, more energy-efficient products.
Next, in smart vehicles, Huang expressed a preference for abstracting cars into “humanoid robots” with autonomous systems.
Huang noted that robots have many joints and sensors, making functional safety crucial, requiring computer design and verification methods, with broad applications of AI technology in this field.
Prediction of Affordable Humanoid Robots Huang predicted that in the near future, humanoid robots will become the standard equipment for everyone, with manufacturing costs potentially much lower than expected.
“Why can’t we purchase a humanoid robot in the price range of $1-2,000, similar to affordable cars?” Huang questioned.
He highlighted that most environments are designed for humans, from production lines to warehouses, implying that in such settings, humanoid robots could be more versatile, productive, and adaptable.