my fintech empire.

Chapter 647 [Chen Yu]

After entering April, the volatility of the concept of galaxies has gradually stabilized from a violent explosion and plunge, and it is not always a daily limit or a daily limit.

However, the subject of quantitative capital is excluded.

The listed subsidiaries of other star systems all disclosed their annual reports at the end of March, but Quantitative Capital has not yet released its annual report.

Although the annual report was disclosed at the end of April at the latest, other subsidiaries of the galaxy have disclosed it. As for your quantitative capital, there is no movement. Everyone thinks that the company's annual report may be a big hit.

In addition to this worry for investors, the stock of Quantitative Capital has risen the most before, from 6.46 yuan all the way to 22.54 yuan, with a cumulative increase of 248.92%. There is a need to make up for the decline.

……

Wednesday, June 4.

At around 10:20 in the morning, Fang Hong came to the headquarters of Qunxing Capital, but his destination was not Qunxing Capital, but a building next door.

The 23rd floor of that building is the headquarters of Quantitative Capital.

"I didn't expect General Manager Chen to come out to greet him in person. What is the origin of that handsome guy who calls himself Fang Hong?" At this moment, the two girls at the front desk of Quantitative Capital Company saw Chen Yu leading Fang Hong into the company, and they couldn't help but Discuss with curiosity.

"It seems that they are about the same age, maybe they are classmates or something..." The girl next to her guessed, and her colleague couldn't help but nympho said: "They are both so handsome..."

At the same time, Chen Yu who came out to receive Fang Hong brought Fang Hong into his office.

The two came to the sofa in the rest area and sat down. Fang Hong looked at Chen Yu sitting opposite him and said with a smile: "Recently I heard that Qin Feng wants to recruit you into his SOCL computing language department, and Lao Huang also wants to invite you to join. Nvidia."

Hearing this, Chen Yu said: "I have always wanted to build a more accurate artificial intelligence quantitative trading model. The complexity of the model is getting higher and higher, and the corresponding requirements for data parameters and computing power resources are getting higher and higher. , computing power, and data are all indispensable, especially the shortage of computing power resources. The existing hardware must meet the computing power I need, the cost is too high, and the efficiency is still too slow..."

The implication is that the current hardware level cannot keep up with his requirements.

Chen Yu said: "When we usually run the model, whether it is deep learning training or reasoning, the first question is how much video memory is needed. Why is the graphics processor of Xingyu Technology so fast? One of the reasons is that the general memory gets rid of the PCIe The limit allows the CPU and GPU to exchange information more quickly."

"I think Xingyu Technology's SOCL has great potential and has its own ecosystem foundation. It is most likely to challenge the status of Nvidia's CUDA, although it seems that there is no competition between the two parties."

Hearing this, Fang Hong looked sideways at him.

Chen Yu looked at him and said in a deep voice: "But Qin Feng is obviously not aware of the relationship between GPU, CPU, SOCL and AI and its significance in the field of artificial intelligence. No, it should be said that he is conscious, at least he understands better than Wall Street, otherwise There will be no SOCL, but Qin Feng's emphasis is far from rising to the level of the STAR series of smart phones."

Fang Hong was quite happy in his heart, this Chen Yu was definitely a talent.

The Quantitative Capital he founded now has a total of more than 300 employees. Fang Hong has known the general situation of this company long ago, but more than [-]% of the employees have academic backgrounds with high-achieving students in the fields of computer science, physics or mathematics, including Chen Yu himself has such an educational background.

Now it is researching the capital market and doing investment transactions, but this team has transformed into a powerful technology development team.

After a while, Chen Yu turned on the computer on the table, and said to Fang Hong: "This is a self-learning neural network AI model we ran, and it has watched tens of millions of videos on Yixing Video. The goal is image recognition, but the problem is that the computing power is not enough. If you want to achieve this goal, you need the support of thousands of CPUs, but if you switch to GPUs, you can do it with only seven."

Hearing this, Fang Hong stared at the screen and said, "Well, I know what you want to say. Although a GPU's single computing unit is not as versatile as a CPU, it can perform a lot of calculations at the same time."

Fang Hong's identity as the original owner in this life was born in the Department of Computer Science. Although he may not be able to compare with Chen Yu and Xu Jingren in this respect, it is no problem if he applies for a job in a major technology company. This is also an advantage that other investors do not have.

Chen Yu nodded and said: "That's right, just like the printing performance at the opening ceremony of the [-] Olympic Games, it would be quite complicated if one or a few people who mastered the overall changes could control the array in real time, but the actual performance Each member only needs to remember when to stand up and when to squat down, so that the whole can present complex and changeable effects.”

"These members are like small computing units in the GPU. Although they don't have global information, they can show the effect we want when they work together. AI computing is a scene that requires a lot of computing at the same time. , including the AI ​​trading model we run. Now we are using GPUs for deep learning training.”

"If you just say that GPU is more suitable for AI, it is definitely not true, but you have to mention Nvidia's CUDA and Xingyu Technology's SOCL. Lao Huang released CUDA1.0 five years ago, which is a GPU for computing. Parallel computing platform and programming model, although it is mainly used to accelerate image processing, there is no revolutionary thing."

"But I believe that Lao Huang has identified the potential of using GPU for computing. His spare no effort to support it is the best proof. Every chip of Nvidia supports CUDA, and at the same time, it opens CUDA to the public, intending to build The CUDA ecosystem speaks for itself.”

At this moment, Fang Hong had a good idea about Chen Yu, but he didn't say anything, but continued to listen to Chen Yu calmly: "The current stock price of Nvidia is around 12 US dollars, and the total market value is less than 75 billion US dollars. It can be seen that Wall Street can't understand it, let alone what artificial intelligence is. In the eyes of Wall Street, in order to cooperate with the CUDA framework, Nvidia doubled the cost of graphics cards but couldn't sell them at a higher price, and the profit was once too low. look."

Fang Hong is calm at the moment, but he is happy in his heart. Shi Yao has discovered a treasure, and he has a very accurate view of the potential of Nvidia in the field of artificial intelligence.

Fang Hong, who has the memories of his previous life, knows very well that if nothing unexpected happens, Nvidia will skyrocket in the next few years. First, it will take off the huge demand for computing resources in the cryptocurrency mining trend. Artificial intelligence took off at this time when the tide just passed.

Ten years later, when ChatGPT became popular, within a few days, a company came out and announced how powerful it was to release a new model. One of the indicators was how many high-quality GPUs the company claimed to have. Whenever a company announced that it was capable of When participating in this competition in the field of AI and failing to produce products, they will say how many Nvidia A100 graphics cards their company owns, and they will be able to release their own large models soon.

It can be said that the AI ​​track at that time was almost inseparable from the hardware supply of Nvidia. Nvidia was also equivalent to the underlying infrastructure company of AI, and it almost monopolized this industry. no.

But apparently few people are aware of this now. Even Wall Street keeps asking Nvidia why no one is using CUDA. This makes Nvidia abandoned by capital. are still low.

At this moment, Nvidia's market value is not even 75 US dollars, and Xiaomi's valuation is 400 billion US dollars.

Wall Street capital can't understand it, but Fang Hong knows that there are people who understand it. At present, there are at least two people in China who understand it.

One is Chen Yu in front of him. Judging from his discussion, it is obvious that he firmly believes that GPU will become the standard configuration in the field of deep learning in the future.Another person who understands is Qin Feng. Xingyu Technology has launched new businesses this year, including computers, tablets, etc., and SOCL has also emerged as the times require.

At this moment in time, Fang Hong definitely cannot let Lao Huang's Nvidia monopolize the supply of AI's underlying hardware.

……

(End of this chapter)