Some people call 2023 the first year of leapfrog development of AI technology.
First, ChatGPT was born, refreshing people's cognition of artificial intelligence, and then various highly "anthropomorphic" mature works swept the Internet.
AI's amazing technological iteration tells people with facts that the era of artificial intelligence has arrived.
As a follower and catch-up, China's computing power speed is also running wildly on the road of AI.
However, roadblocks appeared at this time.
After speculating on AI for more than half a year, the market rumors are also going to be stuck in the neck?
01
Lao Mei is doing things, and in early July to restrict AI chips?
Recently, the scientific and technological and financial circles have been rumoring that the United States will impose unilateral sanctions on China's AI chips.
According to market sources, the United States is considering imposing new restrictions on the export of artificial intelligence chips to China.
One theory is that the U.S. Department of Commerce will act as soon as early July, and without obtaining a license, the Commerce Department will ban manufacturers such as NVIDIA, Micron and AMD from shipping AI chips to customers in China.
Nvidia's A800 chip, which is exclusively for China, will also be banned without a license.
In fact, due to the "core restriction order", NVIDIA's business in the Chinese mainland market has been almost devastating.
In order to circumvent restrictions to meet export control regulations, NVIDIA has specially developed a "downgrade" A800 chip to meet the demand for GPUs in the Chinese market.
I thought that this expedient measure could reduce the loss of business in China, but I did not expect that the US Department of Commerce would expand the scope of AI chip sales.
If this news is true, once the A800 chip is restricted from sale, it will be as unbearable as Mount Tai for both NVIDIA and China's AI industry.
Not only that, it is reported that the Biden administration is also considering limiting the provision of cloud services for Chinese AI companies.
02
AI stuck in the neck, and the stock market was the first to be hit
These two things are indeed so significant that as soon as the rumors came out, they immediately caused a lot of damage to the market.
After all, all domestic computing power software is currently built on GPU chips in the United States.
NVIDIA's GPU chips are the well-deserved kings in the AI field, accounting for more than 70% of the supplier sources.
Take the hottest AIGC, similar to ChatGPT's ultra-large AIGC, which requires at least 145600-233700 NVIDIA A100 GPUs.
If it is really as rumors say, it means that the "hard currency" in AI is directly stuck at the source.
Those large domestic manufacturers, Baidu, Ali, Tencent and other artificial intelligence research and development teams, the upper limit of computing power in the future may be greatly reduced.
It is precisely because of this that when the market worries of the "superimposed version of the core limit" appeared, the A-shares that were in the wind were the first to be impacted, and the entire AI and digital economy industry chain fell in panic.
03
Crack the neck problem, "the iron still needs to be hard by itself"
The upgraded version of the "core limit order" has not yet come, and some stocks that rub AI heat in A-shares have been scared first.
In just a few days, the once-popular AI concept stocks fell back to their original shape.
But then again, there are so many so-called AI concept stocks in A-shares, and there may be few "real materials".
Most of them are masters of capital speculation, but there are few core technologies that can be obtained, which is even more worrying than the "core limit order".
Since 2019, tens of thousands of Chinese companies have participated in the field of chip semiconductors, and many companies have also folded here.
In 2021 alone, more than 3,400 chip companies disappeared from our country.
By 2022, more than 5,700 chip-related enterprises in China have been revoked or cancelled.
Among them, many of them are pseudo-high-tech enterprises that are just to "cheat compensation", and some have to withdraw because the core is too expensive, just like the OPPO Zheku team, which claims to have smashed hundreds of billions of dollars.
For cracking the "stuck neck" problem, Ren Zhengfei once said: the chip semiconductor field cannot be smashed with money, and physicists, chemists and mathematicians need to be invested.
I think so. Large infrastructure may be piled up with money, but the solution of the chip neck problem not only increases the investment in research and development funds, but also needs to absorb a large number of top technical talents, when talents and technology accumulation to a certain extent, may be able to achieve curve overtaking like new energy vehicles.