Jensen Huang: the future belongs to reasoning AI, but for this it is necessary to reduce the cost of calculations by several times

Jensen Huang, Nvidia’s longtime CEO, said that the future of AI lies in systems capable of reasoning. However, to realize this concept, the computational cost must be significantly reduced. Huang emphasized that his company strives to annually increase the performance of chips by 2-3 times, while maintaining the current level of their cost and power consumption.

Image source: NVIDIA

In a podcast hosted by Rene Haas, CEO of Arm Holdings, Huang shared his vision for AI development. The next generation of intelligent systems will be able to respond to user queries by going through hundreds or even thousands of steps of analyzing their own findings, he said. This ability for deep reasoning will be a key differentiator from modern AIs such as ChatGPT, which Huang admitted he uses daily.

Nvidia is committed to laying the foundation for such a breakthrough and has an ambitious goal of increasing the performance of its chips by 2-3 times annually while maintaining the same cost and power consumption levels. This approach aims to revolutionize the ability of AI systems to recognize complex patterns and make informed conclusions.

«We are able to provide unprecedented cost reductions for intelligent systems. We all recognize the value of this achievement. With a significant reduction in costs, we will be able to implement complex processes such as reasoning at the inference stage,” Huang emphasized.

Nvidia dominates the AI ​​accelerator market, controlling more than 90% of the market. However, the company is not limited to just producing chips. Its strategy includes the development of computers, software, AI models, network solutions and other services. This diversification is aimed at stimulating wider adoption of AI in the business processes of various companies.

Despite its leading position, Nvidia faces growing competition. Major data center operators such as Amazon Web Services (AWS) and Microsoft are developing their own alternative solutions to reduce their dependence on Nvidia products. In addition, AMD, Nvidia’s longtime rival in the gaming chip market, is actively entering the AI ​​accelerator market, which could increase competitive pressure on the industry leader.