The U.S. chip giant Advanced Micro Devices (AMD) held the 'ADVANCING AI' event on Wednesday in San Jose, California, announcing the launch of the highly anticipated new AI chip series, 'Instinct MI300.' T

his series includes the MI300X, focused on generative AI applications, and the MI300A, targeted at supercomputing. Challenging competitor Nvidia's dominant position in the AI chip market, AMD boasts that its new chips outperform Nvidia’s products in running AI software.

Bloomberg reported that Lisa Su, CEO of Advanced Micro Devices, mentioned in an interview that it takes time to assess the impact of AI on productivity and other aspects of the economy.

She said,

"The truth is we’re so early. This is not a fad. I believe it."

Capturing the Attention of Tech Titans

AMD is very confident in the MI300 series, optimistic about winning the favour of some tech giants, and expecting to generate revenue starting at several billion dollars. AMD stated that customers using the MI300 series will include Microsoft, Oracle, and Meta.

The new ultra-fine microchip possesses over 15 billion transistors. Its memory is 2.4 times that of the current market dominator, the Nvidia H100 chip. According to UltraMicro, the memory bandwidth of the MI300 series products is 1.6 times that of Nvidia products, further enhancing performance.

Surpassing Nvidia's H100

Lisa Su stated that the new ultra-microchip is on par with Nvidia's H100 in terms of training AI software capabilities, but it outperforms Nvidia significantly in terms of inference performance.

In fact, NVIDIA has already developed the next-generation chip, H200, which surpasses H100. Scheduled for release in the first half of next year, the H200 will replace the H100. The H200 features new high-speed memory. Additionally, NVIDIA is expected to launch a completely new processor architecture later next year.