NVDA

<p>Jensen Huang stated that there are no physical laws limiting AI data centers from scaling to millions of chips, and we can now scale AI software to run across multiple data centers. We are prepared to scale computing at an unprecedented level, and we are just getting started. In the next decade, computing performance will double or triple each year, while energy demand will decrease by 2-3 times annually, which I call the super Moore's Law curve</p>

- NVIDIA CEO Jensen Huang discussed the company's advancements in AI during an interview, emphasizing the potential to scale AI data centers to one million chips.  
- He highlighted the shift from traditional computing models, noting that NVIDIA's innovations, such as NVLink and heterogeneous computing, aim to overcome the limitations of Moore's Law, projecting significant increases in computing performance and efficiency over the next decade.  
- Huang also mentioned the importance of AI in chip design and the transformative impact of generative AI across various scientific fields, suggesting a future where AI becomes integral to both software development and operational efficiency.

Wallstreetcn

NVIDIA

Jensen Huang: AI data centers can scale to millions of chips, performance doubles annually, and energy consumption decreases by 2-3 times per year (full text attached)