Last year, the rise of ChatGPT in the United States made Nvidia, which developed AI chips, hot, and quickly surpassed Intel, TSMC, etc. to become the king of the global chip industry, but before Nvidia came back from the euphoria, a chip company said that their AI chip performance was ten times that of Nvidia, and this start-up has a lot of roots with Nvidia.
The startup, called Grop, is based in Silicon Valley, and its new AI chip is called Quantum AI, which is ten times more powerful than Nvidia's top-of-the-line chips, and the industry may be worried that the startup is bragging, but if you analyze the company's people** carefully, it is worth trusting.
It is reported that this group of technical engineers believe that they have been treated unfairly at NVIDIA, and that their technical ideas are related to differences with NVIDIA, so they chose to leave NVIDIA and founded GROP.
The unfair treatment of GROP's founding team may be related to the strength of founder Jensen Huang, who founded NVIDIA in 1993, and around 2005, Huang recognized the huge potential of GPU chips and began to bring GPU chips to the server chip market.
Huang's years of correctness in technical direction have given him full confidence, and he has naturally maintained a strong control of many technical directions, which conflicts with the fairness and inclusiveness that American technical engineers yearn for, leading to the departure of these technical engineers to start another AI chip company.
It is not whimsical for Grop to develop AI chips with stronger performance, Google, Amazon and other Internet companies have developed chips for their own needs, among which Google's Alpha Dog is using its own AI chips, and Microsoft has also recently intended to develop targeted custom chips to further avoid the use of NVIDIA's general AI chips and be caught up by other AI companies.
These American companies have developed AI chips based on GPU chips with NVIDIA, which has undoubtedly inspired China's chip industry. Due to well-known reasons, the United States does not allow NVIDIA to sell the most advanced AI chips to Chinese companies, and since the end of last year, NVIDIA has not even been able to customize A800 and H800 for Chinese companies**, and recently launched the H20 chip with castrated performance, for which Chinese chip companies have developed domestic AI chips.
Previously, the industry was worried that most of these AI chip companies in China did not have a GPU chip technology foundation, and whether they could develop AI chips to bypass NVIDIA's technology, and now these companies in the United States have developed special chips for AI, which undoubtedly proves that it is completely possible to bypass NVIDIA's GPU chip technology.
This will have an important impact on China's AI chip market, those who are worried that the performance of domestic AI chips can not be compared with Nvidia's chips will be reduced, and even try to use domestic AI chips, providing opportunities for domestic AI chips to break NVIDIA's monopoly situation, as long as domestic companies adopt domestic AI chips in a big way, then China's AI chip technology is expected to improve rapidly, and at the same time break NVIDIA's CUDA ecological monopoly.
It can be said that the development of AI chips by these companies in the United States has shaken the king of NVIDIA, and the capital market has also reacted to this, resulting in a sharp drop in NVIDIA's stock price, losing more than 500 billion yuan in market value overnight, highlighting investors' worries about NVIDIA's brilliance In an instant, the AI chip market will soon show a situation of a hundred flowers.