Reborn in 2010, the AI ​​in my mind is too powerful.

Harry nodded, then said somewhat seriously, "Indeed, our OpenCL support is very poor, and the GPU driver team's budget is also insufficient, leading to a high turnover rate of OpenCL engineers. So we'd like to ask, if we agree to give you some shares, could you contribute your technical expertise as equity and help us resolve this issue?"

Chen Yiran was stunned. He didn't expect AMD to be so blatant in showing weakness and so open in trying to recruit him.

[Agree to his request; this is like Qin Shi Huang returning home—we've won big! Nvidia, that bad boy, is stingy, so AMD is easy to bully. To do our work better, we need to get our hands on the internal design documents for the next-generation graphics core architecture, early drafts of instruction set extensions, and microarchitectural details of the stream processors. That's not too much to ask, is it?]

With these tools, I can directly optimize for AMD's future hardware within DeepFlow, ensuring the entire framework runs faster than Nvidia cards on AMD cards. We essentially have a detailed roadmap for all of AMD's hardware for the next two to three years, which is very beneficial for the future, allowing us to gain a lot of valuable experience and data.

Chen Yiran ignored Jasmine. He was actually inclined to agree, as it seemed to perfectly align with his goal of slowing down Nvidia. However, he had to maintain his composure. He nodded and said, "I can consider it. Tell me your detailed terms first. You know I'm not short of money; shares alone won't sway me."

Harry was clearly prepared; he pulled a folder from his briefcase, a copy of an internal memo.

"2%, unlocked over four years, one-quarter each year, with each batch tied to a delivery milestone."

"What node?"

"In the first batch, the new driver achieved more than 70% of the general computing performance on AMD GPUs, comparable to NVIDIA's equivalent hardware."

In the second batch, DeepFlow's OpenCL backend went live, and the first 5,000 developers successfully installed and used it.

In the third batch, AMD GPUs increased their market share in general computing to over 10%.

The fourth batch will see the entire framework synchronously supporting our next-generation graphics core architecture.

This shareholding is already quite high; Chen Yiran is no longer the same person he was when he was first reborn. He's now considered a rising star in Silicon Valley, and he knows a lot of relevant information.

This kind of equity investment is common in Silicon Valley. IBM is very fond of giving equity to maintainers of open source projects, but 2% is an extreme amount.

He will basically take all of the annual equity incentive pool that the board of directors can authorize.

What made Chen Yiran even happier was the issue of next-generation graphics.

He asked calmly, "I don't even know what your next-generation architecture will look like, so how can you promise synchronization? Will you provide me with architecture-level design manuals?"

Harry solemnly stated, "The architecture-level design manual will be open to you, including the stream processor scheduling logic, microarchitectural details, and data sharing protocols for heterogeneous system architectures. Your access will be limited to the top two levels within AMD, allowing you to see the same things as our own chip architects."

[Oh ho ho, I'll take advantage of him completely. But that won't do either; we need to push our luck and be able to make demands regarding the instruction set of the next generation of AMD GPUs.]

Chen Yiran deliberately paused for a few seconds to pretend to think, and then said, "I have three conditions. The first is that I need to be able to make requests regarding the instruction set of the next generation of AMD GPUs."

Secondly, I need a seat at the OpenCL organization, specifically at Khronos.

Thirdly, purely because of a lack of trust in your management, I need you to promise that you will not abandon investment in optimization, and that each year's investment will be a certain percentage of GPU revenue; we can calculate the specific figure later.

Harry felt much more at ease. Chen Yiran's first two conditions seemed excessive, but he would soon become a shareholder of AMD. If he acquired all 2% of the shares, he would be the number one individual shareholder, and even if institutional shareholders were included, he would rank among the top fifteen.

Moreover, the third condition further proves that Chen Yiran intends to work with AMD to achieve long-term mutual benefits.

He nodded and said, "I can agree to these conditions today. The OpenCL2.0 working group was launched last year, and several companies have sent people. We will let you join as a technical consultant for our company."

Chen Yiran nodded: "If these conditions can be met, I think we can reach an agreement. Let's call it a day. I will contact the lawyer and we can schedule a time to discuss the specific details."

"Yes, but hopefully as soon as possible. We have a lot of work to do, and approving such a large-scale restricted stock award is quite troublesome."

The two stood up and shook hands, and Chen Yiran escorted Harry out of the company.

Chen Yiran didn't stay at the company for long; she went straight home and started discussing things with Jasmine.

"They seem to get along very well, but are you sure there won't be any problems?"

AMD is really good, much better than Nvidia, and easier to bully. Letting it develop shouldn't be a problem.

"My concern is whether this will create a new Nvidia. I hope they can compete with each other, but neither of them can grow big enough, and that they can all develop in tandem."

No, no, Nvidia is hogging the benefits. They use CUDA, which can only run on their own hardware. OpenCL is much better. It's just an open standard, not owned by AMD. Even if China blocks it in the future, it can still be used on their own hardware.

Another point is that if we eventually manufacture our own chips, we can simply develop our own backend and interface directly with DeepFlow without any performance loss.

"Okay, the goal is to dismantle the game rule that GPU computing can only be monopolized by one company. If one company will eventually monopolize the future of AI computing, it has to be us."

Chen Yiran took out her phone and called Ted.

"Ted, come to my house immediately, I have something important to discuss."

"Ah, what's wrong that your company's legal department can't handle? As far as I know, Wang Zhiwu is quite a good guy."

Chen Yiran was speechless. After a three-second silence, she began to complain: "Is this how you praise your own son? Before, you would ask if I had any business as soon as you answered my call. Now that your own son is here, why do you change your tune and stop helping your law firm get business?"

Ted said matter-of-factly, "I'm just a partner in the law firm. I'm not the only one making money. My son's future is more important. What kind of job is it? Are you sure Wang Zhiwu can't handle it?"

"This is a serious matter. He has no experience, and I'm worried about him. Why don't you come and take a look?"

Just as Chen Yiran was about to play some games and take a break after hanging up the phone, it rang again.

Upon picking it up, Chen Yiran saw it was Qiu Ciyun's. He knew there must be some new information regarding the lithography machine.