Nvidia Breaks Moore’s Law: Huang Announces Record Progress in AI Chips | Hardware and software of computer | Computer hardware parts list | Hardware catalog | Turtles AI

Nvidia CEO Jensen Huang said his company’s AI chips are outpacing the progress set by Moore’s Law, with performance improving faster than in decades. With its new superchips, Nvidia is pushing the boundaries of AI inference technology, promising a future of lower costs and greater efficiency.

Key Points:

• Nvidia AI Chip Performance Is Advancing Faster Than Moore’s Law
• Nvidia’s new superchips are up to 40 times faster at handling AI inference.
• Nvidia’s innovation is focused on the entire technology stack, from chip design to algorithms.
• Huang predicts that inference technologies will become cheaper thanks to higher-performance chips.

Nvidia CEO Jensen Huang recently said that its AI chips are advancing at a rate that is faster than Moore’s Law, a historic law that has guided technological progress in computing for decades. Moore’s Law, formulated in 1965 by Gordon Moore, co-founder of Intel, postulated that the number of transistors on computer chips would double every two years, thus improving the performance of the devices. This concept has dominated the technology scene, driving continued advances in computational capabilities at decreasing costs. However, in recent years, this rate of innovation appears to have slowed, with some struggling to hold the law to the benchmark for technological advances. Huang, however, argues that Nvidia is defying this reality with a pace of innovation that is faster than historical predictions.

In an interview with TechCrunch, the CEO explained how Nvidia is accelerating its technologies thanks to an approach that synergistically integrates architecture, chips, systems, libraries and algorithms, reducing development times and pushing performance limits. Huang also illustrated the tangible progress of Nvidia chips, highlighting that the company’s latest superchip, intended for data centers, is able to run AI inference workloads with up to 40 times greater performance than previous models. This leap in performance, Huang argues, not only increases efficiency but also paves the way for a reduction in computing costs, potentially reducing the expenses for AI inference activities, which are particularly high today. Inference, in fact, is the phase in which AI models generate answers or predictions, and it is precisely in this area that Nvidia chips are making the difference, becoming essential for large technology companies involved in the development and implementation of AI.

In this regard, Huang noted that one of the keys to overcoming the limitations of Moore’s Law is the ability to work on all layers of the technology at the same time. This approach allows for exponential performance increases, making each part of the system evolve at a speed that would not be possible if we were limited to improving only one of the individual elements, as traditionally happens in the chip field. In particular, Nvidia is focusing its efforts on optimizing the inference phase, in which the potential of AI models is exploited to process and respond to questions or requests in real time. Huang predicts that, thanks to the continuous evolution of its chips, the costs for performing these operations will decrease over time, making AI-based technologies more accessible.

To support his statements, Huang recalled that today Nvidia chips are about 1,000 times more performing than those made ten years ago, marking an unimaginable progress compared to the past. This accelerated pace is the result of continuous innovation that characterizes Nvidia, committed to producing increasingly powerful chips, which not only meet the needs of large technology companies but also pose new challenges to the entire industry. Huang’s goal is not only to improve performance, but also to reduce costs, a key aspect for the expansion of the use of AI in everyday applications. Solutions offered by Nvidia, such as the H100 superchip, are already widely adopted for training and running AI models, but now new prospects are opening up in the field of inference, which could have a decisive impact on the future of AI.

The evolutionary path traced by Nvidia seems to promise a future in which AI performance will not only improve, but will also become more affordable, with the aim of breaking down the economic barriers that today limit access to these advanced technologies.

Despite the challenges and doubts of some analysts, Huang is confident that AI innovation, fueled by the power of its chips, will continue to follow an upward trajectory.