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

Data centers are never a simple task; there are many things to consider.

Power capacity is the most fundamental factor, determining how many servers can be hosted. Backup redundancy and secure connectivity also require careful consideration.

The most critical metric is Power Usage Effectiveness (PUE). Simply put, it's the total power consumption of the data center divided by the actual power consumption of IT equipment. The closer the number is to 1, the better, as this proves that most of the data center's power is used for servers, and not a large amount of power is wasted on things like heat dissipation.

In areas with dry and cool climates, free external air cooling can be used, and the power efficiency may reach below 1.10.

High-end data centers in Silicon Valley typically use direct expansion cooling coils and air-side energy-saving devices, achieving power efficiency of 1.2 or 1.1.

For older, traditional water-cooled units, this figure is rather average, only reaching 1.3.

Don't underestimate the difference of only 0.1; when the scale is large, the annual electricity bill can easily differ by tens of millions of dollars.

Finally, there are some basic factors, such as the size of the park, the area of ​​the land, and the surrounding infrastructure.

Apple has built four major data centers in the United States. The first, the flagship data center in North Carolina, which began operation in 2010, is a 100% renewable energy data center. The other data centers also meet high standards and are primarily powered by local solar or wind energy.

However, Chen Yiran was not satisfied. His needs were greater; he required an independent, completely autonomous data center where no one could compete for computing power. Aurora Group's data centers were all in Europe and had not yet expanded to the United States.

Fortunately, Silicon Valley, as a technology hub, has a well-developed related industry. And the timing is also excellent, as Facebook has just launched its self-built data center in Oregon.

Their contract with a third-party data center in Santa Clara expires in three years, forcing them to seek low-cost subletting on the market to recoup some losses.

All of them are high-spec server rooms. Although the power efficiency did not reach below 1.1, it is already quite good that it reached 1.1.

Chen Yiran decisively took over and began to mass-produce Oracle's latest all-in-one machines, all of which used the latest Mocha processors.

Ellison felt that Chen Yiran always came up with interesting product names, and this time he also carefully selected one, finally choosing Black Stone as the name of the new generation of all-in-one machines.

The BatuHitam is a Tang Dynasty shipwreck discovered in 1998 by a German salvage company in the waters off Belitung Island, Indonesia. It contained more than 60,000 pieces of Chinese porcelain and was a merchant ship that traded with China from Arab countries.

Ellison plans to name the subsequent series after various famous ancient sailing ships, and has even reserved the name of Zheng He's treasure ship for the most stunning product.

Chen Yiran's business alliance chain is becoming increasingly clear. This time, the chip was designed by Aurora Group. Oracle bought it, put it into their all-in-one PCs, and sold them, making a profit.

The ultimate consumers are Apple and Samsung, who will purchase large quantities of all-in-one PC hardware and software products and leverage their brand value to sell them and make money.

In short, each company has a bright future.

Chen Yiran sat on the sofa in the server room, his fingers tapping incessantly on the screen as he began modifying various settings.

[That's fast! I've done some calculations, and I think at this level, we can at least reach the 2020 standard.]

"It's still not good enough. There are still too few things it can do. Look at this automatic programming, what kind of crap is this? The level is too poor."

"It's good enough as it is. There are still so many empty racks; filling them all should be enough. After the data centers in Europe went live, thanks to their better infrastructure, we can rival the computing power we had in 2020 before our rebirth. ChatGPT 3.5 is nothing compared to what we have here."

"What's the point of ChatGPT? We can't use computing power on such useless things. It's meaningless except for making money. It's more reliable to prioritize serving Aurora Group's internal R&D platform. In just a few months, so many breakthroughs have been achieved thanks to the platform."

DJI has a better control system now, we need to hurry. According to you, we might need to get involved in the war this August.

[No major problem. Our current technology is sufficient to change the overall logic of local wars, especially protracted wars.]

"Those tycoons in the Middle East have tons of money; they make money so fast. I now think it's a waste to use AI on ordinary users. Besides arguing with their babies all day, what else can it do? We still need to prioritize our big clients; ordinary consumers should take a backseat."

【(°ー°〃) Look at your ugly face now! The dragon slayer has become the dragon himself! Who was it that criticized Huang for forgetting his roots and only caring about selling graphics cards to big companies, causing graphics cards to become increasingly scarce and expensive for gamers?】

"Your position determines your perspective. Now that I'm a capitalist, of course I want to maximize my profits. Sigh, the main problem is insufficient production. If production were sufficient, I could take care of everything. Besides, Claude should have enough for now."

I think you can stop this game now. Didn't you see the warning light on your left front?

Chen Yiran glanced at the screen on the left, but ignored it; he couldn't just give up right after the game started. Fifteen minutes later, Chen Yiran's Shadow Fiend in the mid lane had so much damage that the second tower didn't dare to leave, and he decisively typed GG.

He glanced at the notification on the screen and nodded: "It's done. The data has been completely uploaded to Apple's data center. We just need to integrate it and it'll be ready to use. We'll then see how much of it is available to regular users based on Apple's server capacity."

[Okay, that's within the plan, but there's a problem. You and Leather Jacket Huang were supposed to have afternoon tea at 3 pm, and it's already 2:30 pm. You also started a game of Dota. Is there enough time?]

"It's alright, it's his fault for colluding with Google, let's just put him in his place."

Chen Yiran kept her appointment on time. At 3 o'clock, she and Huang, who was wearing a leather jacket, sat down together in a small coffee shop, both with ulterior motives.

Leather-clad Huang looked around and said, "I heard you have some issues with your sense of security, but this is too much. And how did you convince Starbucks to help you close the store? No amount of money could have done that."

"Starbucks is also run by people. The boss here gets paid by me. I've heard that many companies monitor coffee shops and restaurants near other companies."

So, to prevent any potential problems, I took precautions and made sure to take care of all the shops around the company. Strictly speaking, the current shop managers and staff are also employees of our security company; they just happened to be hired here as shop managers.

Huang in the leather jacket seems to be under a lot of pressure; Chen Yiran is very similar to the main villain in many movies.

He took a sip of coffee and said, "We've known each other for a long time. My son often plays games with you. I heard you beat him in StarCraft 2 every day, but you also screw him over in Dota."

I don't want to beat around the bush. I want to know directly: will your Mocha chips be used in Aurora-related companies, or will they be fully open to the market?

"You've already partnered with Google, isn't it a bit pointless to ask me this now?"

The man in the leather jacket, his hand trembling, said, "No, no, I've only had some exchanges with Google; we haven't actually collaborated. My main question today is whether you can guarantee that you won't interfere with the training process, and will only do inference?"