Those Years When I Was Forced to Top the Forbes List

Chapter 342 The Year of the GPU
After the meeting.

"Mr. Chen, here's our next round of investment funds."

"Mr. Chen, about the sixth-generation production line."

"Mr. Chen, about the joint venture with Huawei."

Xu Ran from JD.com, Fan Renyong from Spreadtrum, and Liu Xiaodong from BOE followed Chen Xuebing into the office.

Lu Weibing was quite sensible; he didn't mention it at all, but he chatted with Wu Jiesi and arrived later.

After the office door closed, Chen Xuebing looked at the group, his face filled with thoughts of money, money, and more money.

Mr. Chen seemed rather relaxed, dismissing them one by one:

"The second tranche of JD.com's Series C investment of $600 million will be released to you on time at the end of this month. You need to pay the money for the Ziyang warehouse on time and do not default. This will be the core transit warehouse for Chengdu and Chongqing in the future, and it will be a sure thing."

"Don't worry, Mr. Chen, we've discussed this in a meeting. The money will be transferred soon. Once Huaxi Group finishes the infrastructure, we'll develop the primary warehouses and expand our operations to the Sichuan and Chongqing markets! Mr. Liu even told me that he'd like to invite you to visit our warehouses in Shanghai and Guangdong! Mr. Liu is in Guangzhou right now!"

"I'll visit your Shanghai warehouse sometime."

Chen Xuebing wasn't being modest; he really wanted to talk to Dongzi about the subsequent network upgrades of logistics, but some technologies weren't available yet, so it wasn't the right time.

After saying that, he looked at Liu Xiaodong, his expression becoming more serious: "I will discuss the investment in the sixth-generation line with the State-owned Assets Supervision and Administration Commissions of Hefei and Beijing, and we will sign the agreement by the end of the month at the latest! Also, go back and tell President Wang that the production line construction should not be overly ambitious. It is best to have two parallel lines for small screens and large screens, advancing in parallel! With BOE's sixth-generation line in place, the next step should be to expand the small screen production line and increase the production of mobile phone screens! I have already chosen a location for you, in Chengdu! Please don't forget, President Wang, I will be responsible for BOE's market operations. You can make the line construction requirements, but the completion of the production line is mostly an economic issue, which I will consider for you. President Wang only needs to focus on R&D. Even if we need to build a line in Shenzhen in the future, I will help you negotiate. You don't need to worry too much about it."

Liu Xiaodong was given a severe scolding and was initially a little confused.

When the topic turned to Shenzhen, he finally understood.

He was Wang Dongsheng's strategist. When Shenzhen contacted them, he was the one who set the tone for their response. Hearing this, he understood that the matter of shifting the blame to General Manager Chen had been discovered.

"Uh," he chuckled awkwardly, "Mr. Chen, did Shenzhen contact you? There aren't many places that truly have the resources to support our line construction."

Chen Xuebing interrupted coldly: "Don't think about these things. Don't you have to pay back the money the local government gave you? If you do your business well and have money, you won't need local support. You can build your own production line! You guys get OLEDs up and running, and I'll guarantee your sales!"

After hearing this, Liu Xiaodong dared not say anything more.

Dozens of manufacturers are in their hands, so they can say that.

Moreover, when this Mr. Chen went to Beijing to talk to the State-owned Assets Supervision and Administration Commission, he not only discussed the issue of exiting the company, but also the next step of reinvesting.

The market-oriented exit of Changzheng Capital does not mean its withdrawal from the shareholder list, but rather that it plans to use its own funds to take over BOE and play a relay game with the Beijing State-owned Assets Supervision and Administration Commission.

Changzheng is gone, but Gu'an will come in later.

This matter was discussed privately between Chen Xuebing and the leaders of the Beijing State-owned Assets Supervision and Administration Commission. The leaders did not reveal any details, and even Wang Dongsheng was not clear about the specifics.

All he knew was that Mr. Chen would remain a major shareholder of BOE.

They are just managers and have no authority to make decisions on this matter.

Moreover, given the current situation, it would be highly beneficial for BOE to have Mr. Chen as a shareholder. The board of directors is unlikely to raise objections with the State-owned Assets Supervision and Administration Commission, and even if they did, the shareholders in the stock market would probably not agree.

The value of the shares held by the State-owned Assets Supervision and Administration Commission has now doubled, so they probably won't agree to it.

Chen Xuebing had the support of all major shareholders and possessed strong capital within BOE, but he didn't want the management to feel that he was too difficult to deal with. He only vented briefly before softening his tone and saying:
"I support you in opening a factory and conducting research and development, but you also need to be responsible to the market shareholders. Having the support of the entire market is your strongest advantage."

Liu Xiaodong quickly seized the opportunity: "Yes, with President Chen's help, we now have more confidence, and we will definitely cooperate with President Chen's decisions."

"Next year, BOE must turn a profit and distribute dividends to shareholders so that everyone can see the company's transformation."

Chen Xuebing dismissed both JD.com and BOE, leaving only Fan Renyong, Lu Weibing, and Wu Jiesi, two of his own people, in the office.

Actually, Spreadtrum is considered one of our own.

Chen Xuebing's expression softened: "Old Fan, SHLTE is the capital I used to invest in Spreadtrum. Don't worry about it. I'll arrange the subsequent funding. How's the cooperation going so far?"

This joint venture takes one initial from Spreadtrum and one from Hua Wei, plus the name of its core research area, LTE.

This company is located in Zhangjiang, Shanghai. SH is also the abbreviation for Shanghai, which implies that the Shanghai government is very satisfied and has specially allocated a good plot of land for it.

"There's progress!" Seeing Liu Xiaodong being criticized, Fan Renyong quickly searched his mind for information about the joint venture, and then smiled and said, "They have made a TD-LTE frame structure proposal, which has solved the problem of unifying the dual-mode frame structure. They are currently talking to China Mobile and want to adopt the dual-mode BBU architecture in all newly built base stations."

"Oh? Solved?" Chen Xuebing raised his eyebrows.

This issue was raised a long time ago, and it was he who brought it up.

The base station processes two signals simultaneously, much like dual SIM dual standby. Previously, the problem stemmed from the different signal timings of TD3G and LTE, which would cause signal interference when placed together. Now that dual-mode can coexist in the same device, it means that the signal timing distribution problem has been solved.

If TD3G base stations can smoothly transition to LTE, and the core can be replaced without changing the casing, it will definitely save more than half of the cost compared to repeated construction in the future.

"How much will the cost increase by using this solution?"

"23% to 28%."

"Hmm," Chen Xuebing pondered for a moment, then shook his head: "That would add at least 50 to 60 billion in construction costs. I think we should wait until there is a breakthrough in the core MIMO technology before we talk about it, so that people don't think we're just looking for an excuse to snatch orders. The first batch of construction is in big cities, which will take at least half a year. If we can make a breakthrough in these six months, it will be easier to negotiate."

At this juncture, China Mobile is placing base station orders. In addition to Huawei and Spreadtrum, Datang, ZTE, and Fiberhome Technologies of Wuhan Research Institute of Posts and Telecommunications have all participated in base station construction. Proposing a new solution at this time seems like a rogue tactic to prevent others from participating.

China Mobile probably doesn't want to take responsibility either.

What if it doesn't meet international standards in the future?
What if the construction doesn't achieve the desired effect?

Ideally, you should have a breakthrough in your own technology, or a clear message from an international association regarding the 4G roadmap.

“Let’s work and discuss as we go!” Fan Renyong hadn’t thought that far ahead, but then, remembering something, he said excitedly, “There’s more! This November, an academic team from the University of Toronto proposed a DBN concept in Science, called Deep Belief Network! This concept describes the construction method of deep neural networks, which allows computers to understand complex information layer by layer, just like humans, through hierarchical self-learning. Huawei and we both believe that this idea can be applied to signal processing, using analysis of massive amounts of interference samples for self-training to filter noise in real time. If implemented, it could increase the pass rate of effective data by more than 30%! Moreover, this technology definitely needs to be implemented using chips. After some discussion, we decided to collaborate on the research at SHLTE, with our technical experts taking the lead in the project!”

Chen Xuebing's eyes lit up.

General communication technologies can be summarized into three aspects: signal amplification, signal filtering, and signal conversion.

A 30% pass rate for valid data is a major breakthrough in the international communications industry.

However, the phrases "deep neural networks" and "self-training with massive amounts of samples" inevitably brought one thing to his mind: AI.

Chen Xuebing nodded calmly, seemingly casually reminding him, "This model sounds good. Wouldn't it be more effective to train using a GPU?"

Lu Weibing was taken aback: "Isn't a GPU a graphics card? It's dedicated hardware with fixed-function texture units and rasterization modules. It can't execute general-purpose computing instructions, right?"

Although he wasn't an IC designer, he's now in charge of the company, and he's learned a lot about the theoretical aspects.

Fan Renyong clapped his hands and stood up excitedly:

"Mr. Chen, you're amazing! They're discussing this issue right now! This year, NVIDIA in the US proposed a CUDA technology that allows GPUs to perform parallel computing tasks! Huawei also researched a lot of information before coming up with this idea. How did you find out about this?"

The office fell silent.

"Is it necessary to have a technical foundation to even think about this?" Chen Xuebing asked calmly, leaning down slightly.

Why does a computer need a dedicated graphics card? Can't the CPU perform graphics calculations?

Without waiting for their reply, he continued:

"It's because CPUs have a limited number of cores. No matter how complex the design is, with two or four cores, when faced with a bunch of simple problems like 'one plus one equals', even with the strongest computing power, you still have to calculate them one by one."

"Problems like graphics units are a series of simple problems. Each pixel requires a small amount of computation, but the number of pixels is huge. GPUs integrate thousands of simplified cores and use a single instruction multithreading mode, which can process massive amounts of similar calculations simultaneously."

"Isn't this approach very similar to the signal processing that Mr. Fan mentioned?"

As Chen Xuebing spoke, he smiled faintly and said, "Moreover, the news that NVIDIA released last month has caused quite a stir in the United States. Their CUDA technology has solved the problem of C language extension interfaces, allowing direct writing of GPU computing kernels, and they have already programmed a Tesla architecture that can support dynamic allocation of computing resources."

The invention of the GPU was formally proposed by Nvidia in 1999.

In previous years, GPUs could only perform preset graphics operations, much like a printer that could only draw straight lines, not solve equations.

But just last month, Nvidia's new interface and self-developed architecture made the theory a reality.

He knew about this news, but his focus was not on the broader technology, but rather on Nvidia.

However, this did not prevent him from taking the opportunity to teach by doing.

"I've said before that we should pay attention to technological development. Mr. Lu, as the head of a technology company, you need to set up your antenna and receive more information from the outside world!"

Lu Weibing was having a bit of a headache.

Damn it, I ran right into the line of fire today, the boss was giving orders.

However, this issue is indeed within his responsibilities. The chairman previously mentioned designing GPUs when discussing the future development of mobile phones. He failed to pay attention to such an important matter, so he cannot shirk responsibility.

"It's my problem. I'll take some time to look into it."

After finishing his sentence, Lu Weibing asked with a hint of doubt, "Chairman, do you have some ideas about GPUs?"

The chairman is a person who loves to learn, but he would never learn something irrelevant. If he can talk about something in great detail, it means he is definitely interested in it.

This conjecture has no theoretical basis, because their IC department tried to learn how to design GPUs, but couldn't get the hang of it at all.

Even someone as stubborn as Zhang Hongbin from the IC department, after trying to learn for a while, said he couldn't do it and then quietly gave up the GPU design plan.

His thinking was actually just an intuition.

Chen Xuebing laughed.

Keen, fucking sharp.

Old Lu was also cultivated.

He pondered for a moment and slowly said, "Currently, our design capabilities for GPUs are still in a vacuum. We have no design experience at all in modules such as vertex processing, rasterization, and texture mapping. Moreover, high-end GPUs use the 90nm process, which is quite difficult to design. But difficulty does not mean it cannot be solved."

GPUs have created the world's most valuable company.

People get rich starting with the concept of blockchain, some get moderately rich from mining, and some get rich overnight from AI.

He doesn't know what the future of AI will hold, but providing AI training tools certainly has a future.

With the messages from DBN and CUDA combined, he felt the time had come.

The DBN concept was proposed just last month.

If this concept refers directly to AI, then November 2006 can truly be considered a turning point for AI.

With these theoretical foundations, some actions can now be planned.

“NVIDIA has now given us an answer: the use of GPUs is not limited to display; they can also be applied to multi-threaded computing. We will definitely need them in our big data centers in the future. Therefore, we must put GPU design on the agenda immediately. We need to make good use of the research center we will be opening in the United States, recruit relevant talents from the United States at high prices, and send people there to learn and bring back some design experience to China.”

We have the reasons, and we have the methods.

However, it's unclear how difficult it will be to put into practice.

Another thing has been added to the list.

When Lu Weibing heard that the chairman had even drafted a plan, he knew that he probably couldn't avoid this matter, so he had no choice but to state his position:

“No problem. President Lin has already gone to Silicon Valley for inspection, and Emerson’s people have gone with him. However, establishing a research center still requires approval from CFIUS (Committee on Foreign Investment in the United States). President Lin said that if all goes well, it should be approved by March next year. I will call President Lin in a while to ask him to keep an eye on the recruitment. But as a foreign research company in the United States, we really have to be careful when recruiting. If we are not careful, the whole company will be deported back to China, or we may not even be able to leave.”

Chen Xuebing knew he was under a lot of pressure, so he nodded and said, "I told you to be self-reliant before, but the situation in the United States is an exception. I will try my best to solve the problems of money and compliance for you."

After saying that, he finally looked at Wu Jiesi, who had remained silent all along.

"President Wu, you're aware of our group's current situation, right? Why don't you give me some advice?"

(End of this chapter)