June 11 News.NvidiaChief Executive OfficerJen-Hsun HuangWe met with Prime Minister Keir Starmer this week to discuss the opening of London Technology Week. AI future.
According to foreign media Tom's Hardware, the two sides jointly announced a series of initiatives to incorporate AI into national economic planning, supporting large-scale infrastructure construction, talent training, and in-depth government-industry cooperation. Jen-Hsun Huang also emphasized that AI hardware performance has increased by 1 million times in the past decade -- but there are also hidden concerns behind this.
Huang Renxun said."AI has progressed a million times in the last 10 years.. The speed of change is mind-blowing." He didn't specify whether he was referring to a leap at the software level or a breakthrough in hardware.
NVIDIA claimed last year that the performance of its Blackwell B200 processor on inference tasks20,000 times more than the 2016 Pascal P100According to 1AI, the B200 provides about 20,000 FP4 TFLOPS while the P100 has only 19 FP16 TFLOPS. According to 1AI, the B200 provides about 20,000 FP4 TFLOPS of arithmetic power, while the P100 has only 19 FP16 TFLOPS. although the two are not completely analogous, they are still representative of real-world applications.
In addition, Blackwell is 42,500 times more efficient at generating tokens per unit of energy consumption.
In addition to stronger GPUs, NVIDIA, xAI, Microsoft, and others are also building AI systems that are far superior to last generation, according to the report. In some ways, it's true that current AI hardware is a million times more powerful than it was a decade ago. For example, xAI's current supercomputers run onWith 200,000 Hopper GPUs..
Musk has also revealed that in the future, he will also build a super cluster containing 1 million Blackwell GPUs, which will not only set a new performance record in the field of AI, but also become one of the most powerful supercomputers in the industry.
U.S. companies aren't the only ones investing heavily in AI infrastructure. The UK has committed to investingApproximately $1 billion（NVIDIA will set up an AI research center in the U.K. to focus on robotics and environmental modeling and materials science, and the first batch of funds has already been received. NVIDIA will set up an AI research center in the U.K., focusing on robotics, environmental modeling and materials science. It is also launching a national developer education program to enhance the technical talent pool; and will work with the UK financial regulator to create a secure testing environment for AI, while participating in the promotion of 6G research.
The UK's current technological prowess still pales in comparison to US supercomputers. The country's most powerful AI system is Isambard-AI, which will soon go live with 5,500 Grace Hopper 200 processors.