The global artificial intelligence AI continues to be hot, everyone seems to have heard of it, many technology giants are developing artificial intelligence models, and Apple has even given up the electric vehicle project that has been developed for 10 years in order to concentrate on the development of AI.
The most beneficiary is NVIDIA, which alone occupies 90% of the global AI chip market, so it makes a lot of money. However, the situation faced by Nvidia is also very grim, not only is it restricted from shipping, but it is also "besieged"!
Previously, in front of the global CPU chip giants Intel and AMD, Nvidia could only be regarded as a little brother for many years. Because compared with the core chip CPU of the computer, the GPU chip that NVIDIA focuses on has always been a supporting role.
But what they didn't expect was that such a little brother would soar into the sky, with the continuous boost of the global AI wave, Nvidia's market value has exceeded 2 trillion US dollars, becoming the third 2 trillion giant after Microsoft and Apple.
If this trend continues, Nvidia is likely to surpass Apple in terms of market capitalization and become the world's second-largest company by market capitalization.
Seeing Nvidia's achievements today, the most regrettable one is probably Intel. You must know that in 2000, Intel's market value reached $277 billion, ranking sixth in the world at that time, and as much as 60 times the market value of Nvidia.
But today, Intel's market value is $154.35 billion, and Nvidia's market value is $2.19 trillion, which is more than ten times that of Intel.
The key problem is that Intel not only canceled the GPU project before, but also stopped the acquisition because there was no agreement on what position to give Huang when it had the opportunity to buy it, which has made Nvidia what it is today.
However, Nvidia's growth is so fast, and it is not always smooth sailing. Some time ago, there were many twists and turns, and even the market value evaporated by 920 billion yuan overnight. This has sparked a lot of discussion about whether Nvidia can continue to move forward when it reaches the bottom.
For NVIDIA, it is currently facing a difficult problem, that is, the chip restrictions of the United States are directly not allowed to ship high-end AI chips to Chinese enterprises, and after A100 and H100 are restricted, even the modified A800 and H800 cannot be used.
Although Nvidia has made revisions again, the H20 chip launched is no longer popular with Chinese companies because of its low performance.
You know, Nvidia's revenue from the mainland market has been accounting for about 20%, CEO Jensen Huang publicly said that the mainland market is very important to Nvidia, because it is irreplaceable, but Nvidia can only bear the loss after the restriction.
Importantly, Huang also understands that many companies in mainland China are developing GPUs and will continue to replace them.
In particular, Huawei's AI chip, the Ascend 910B, has begun to be snapped up by Chinese companies to replace NVIDIA's AI chips. iFLYTEK said that the computing power based on Huawei's AI chip can already be compared with the NVIDIA A100.
Not only are domestic manufacturers accelerating the research and development of AI chips, but even American chip giants are constantly launching AI chips. We are mainly to deal with chip restrictions and solve the problem of "stuck neck", and American companies just want a piece of the pie.
Recently, Google, Meta, Intel, AMD and other actions have taken action. Google released TPU v5p for cloud AI acceleration, and meta launched the second-generation self-developed AI training and inference chip MTIA v2, with computing power aimed at H100.
Recently, Intel has released a new generation of cloud AI chip Gaudi 3, which claims to surpass Nvidia H100 in performance.
According to the relevant data released by Intel, Gaudi 3 is 40% faster than the NVIDIA H100 in terms of AI large language model training speed, and 50% ahead of the NVIDIA H100 in the inference energy efficiency of popular large models.
Obviously, Intel's AI chip is benchmarking against Nvidia, and it also shows that its chip performance has been surpassed.
There is also AMD, which has also recently released AI corresponding chips. Among these manufacturers, AMD developed AI chips relatively early, and has launched a variety of AI chips early, which can be said to be one of NVIDIA's strongest opponents in the field of AI chips.
It can be seen that in addition to the impact of chip restrictions in the United States, Nvidia also has to face the continuous substitution of mainland manufacturers and the fierce competition from American chip giants, because they all want to seize Nvidia's market share.
In this regard, some foreign media directly commented, foreign media: Nvidia's situation has reversed, not only chips and other hardware, but also Nvidia's CUDA ecosystem, Intel, Google, ARM, Qualcomm, etc. have established the UXL alliance to compete with it.
As for whether the siege of American companies can succeed, we don't care too much, the important thing is that we completely realize the replacement of domestic chips.