Rebirth 08: Rise from copycat phones

Chapter 434: GOOG Seeks to Purchase Graphics Cards
Semiconductors and smart terminals are the two major pillars of Zhiyun Group and the two core businesses in Xu Shenxue's career empire.

Most of his other business units are built around these two major business segments. Even the artificial intelligence business is, to some extent, a support for smart terminals.

Pure artificial intelligence is nothing more than a pile of data. Without a physical carrier, it will be difficult to function and even more difficult to bring huge revenue to Zhiyun Group.

This can be seen from the operating performance of Zhiyun Group after the release of Yun AI.

Yun AI from Zhiyun Group is technically the current top generative AI. It has no rivals in terms of both algorithms and reference data. However, the vast majority of users are free users, and only a small number of users are willing to spend money to subscribe to paid features. In addition, some companies will pay to subscribe to the interface.

However, there are not many paid subscribers, only more than 6 million worldwide.

The revenue brought by enterprise users, that is, API users, is not as much as imagined. API revenue accounts for only 15% of Yun AI's revenue, which is far from the project team's previous estimate.

How much revenue have these more than six million paying users and some corporate customers brought to Yun AI?
It is expected that the revenue this year will not exceed 3 billion US dollars. Even if there is an increase next year, it will most likely not exceed 5 billion US dollars.

But what about the spending on the Yun AI project?

It is estimated that the expenditure of the entire project this year will be about 5 billion US dollars, of which the training cost will exceed 2 billion US dollars. Of the other 3 billion US dollars, about 2 billion US dollars are the cost of running the model... The computing power required to run artificial intelligence is very high.

In addition, there are hundreds of millions of dollars in salary costs, administrative and marketing costs, hosting and data costs.

The current cost expenditure of the Yun AI project is actually based on the entire Zhiyun Group system, and many intangible costs are not actually taken into account.

For example, relying on the human resources, hardware resources, brand marketing capital, and public relations resources of the entire Zhiyun Group and its sister companies, these intangible costs are actually very valuable.

Without the massive resources of Zhiyun Group, Yun AI would not have been able to gain so many users and achieve the estimated $3 billion in revenue in its first year.

If other companies were to work on the Yun AI project, even if they had the technology and such an awesome underlying algorithm, the project expenses would be estimated to reach $700 to $800 million over the course of the year.

However, this large-scale generative AI, which has caused a sensation worldwide and is recognized as the first commercialized AI, is now facing losses.

When R&D and training costs are amortized into expenses, they account for the bulk of the entire project's expenses, which in turn leads to huge losses for the project.

This is the situation faced by large-scale generative AI in the conventional sense today... It sounds like a good thing, but in fact it is a serious loss.

Therefore, pure generative AI, especially when providing services to individual users in the form of APP or web pages, is difficult to generate commercial value, at least for now.

Of course, this is also related to the fact that in order to set off a wave of generative AI, Zhiyun Group is currently providing most of Yun AI's basic functions to users for free.

Zhiyun Group is not doing the Yun AI project to make money from the Yun AI project itself... at least not in the early stages.

The Yun AI project acts as a multiplier for Zhiyun Group and even for other companies under Xu Shenxue.

For example, many of Yihai Technology's network services, especially its core product Jump Video, have been connected to the Yun AI service. One of them has a very powerful translation function that can automatically identify users in various regions, and then provide users in various regions with convenient and accurate translation functions. Users can translate with one click, and can translate more than 100 languages ​​and even other unconventional languages, such as emoticons, Morse code, etc.

In addition, Zhiyun Group’s own smartphones and other smart terminals, as well as new products released after April this year, all have built-in Yun AI interfaces in their system updates, which can directly call Yun AI services.

This will undoubtedly bring considerable competitiveness to Zhiyun Group's mobile phones, tablets, smart watches and other smart terminals, thereby bringing better market performance.

In addition, there is another robot, Zhiyun Group's robot is also equipped with Yun AI's interface, making its human-computer interaction function better.

And these are what large-scale generative AI like Yun AI can bring to Zhiyun Group and its sister companies... The benefits far exceed the revenue of Yun AI itself.

Some outside analysts believe that Yun AI, released this year, has brought at least an additional $15 billion in revenue to Zhiyun Group, which is reflected in the enhanced market competitiveness of various smart terminal products and robotics businesses, increased sales, and increased average selling prices, ultimately bringing in an additional $15 billion in revenue.

Therefore, for Zhiyun Group, it is not important whether the Yun AI project itself makes money or not. What is important is that it can make its smart terminal products and robotics business, as well as its cooperative enterprises, such as HaiLian Automobile's automotive products and Yihai Technology's Internet services, more profitable.

This is the significance of large-scale generative AI such as Yun AI to Zhiyun Group, and why this project is still released even though it is loss-making, and the main functions are provided to users for free.

Because the Yun AI project itself is a supporting facility for Zhiyun Group's various products and services.

And this is not a secret. After a little analysis, people in the industry can basically come to this conclusion... Similarly, this conclusion is also the reason why major high-tech companies are striving to develop. Even if the price of APO graphics cards is expensive, there are orders of tens of thousands or even hundreds of thousands of them.

Whether artificial intelligence itself makes money or not is not important. What is important is to provide value-added services and make the core business more profitable.

This is the core reason why a large number of high-tech companies have joined in.

Taking the search field as an example, the search service of Company A is still traditional search, but the search service of other companies uses artificial intelligence search. The quality of search results is very different... If the difference in the quality of search results is too large, it will affect the number of users and further affect the search advertising revenue.

Similar examples also exist in many other Internet service areas.

For example, the core business of the Internet industry is "advertising".

By using artificial intelligence technology, advertisements can be delivered to specific user groups more accurately, catering to their preferences and further improving advertising efficiency. The result is that users' aversion to advertisements is reduced, the click-through rate of advertisements is increased, advertisers are more satisfied, and they pay more for advertising.

Artificial intelligence technology is not limited to anthropomorphism and content generation...its practical applications are very extensive.

It is not just software and Internet companies that are engaged in artificial intelligence. Major hardware equipment companies are also engaged in artificial intelligence, such as Tesla, Apple... and of course Zhiyun Group itself.

If these companies want to develop artificial intelligence, even though Zhiyun Group has made GTAI2 open source for them to play with, there is still one thing they cannot avoid, which is the need for a large number of APO graphics cards.

--------

In mid-November, a business jet arrived at the Shenzhen Airport.

Business jets are nothing to this international airport. Shenzhen is one of the top cities in the country with a large number of large enterprises. Executives of many large enterprises also travel by business jets, especially executives of several companies under Xu Shenxue. They basically take business jets for long-distance travel.

Zhiyun Airlines' main operating base is Shenzhen Airport, where it operates its own business jets all year round.

However, the business jet that Shenzhen welcomed that day was not from Zhiyun Airlines, nor was it from some regular business jet leasing companies.

Rather, it was a large intercontinental business jet that came from the far opposite side of the Pacific Ocean and took off from an airport near Silicon Valley!
As for the plane itself, it is not important. What is important is the people on the business jet!

The passenger on board is not a wealthy person in the conventional sense, but Sergei, one of the owners of GOOG.

After Sergei got off the airport, he looked at the hustle and bustle of traffic outside the airport. At first he didn't think anything was a big deal, but as the convoy gradually entered the city, he looked out of the car window more and more frequently.

This was not his first time to Shenzhen. In fact, he had been here many times. He had attended the flagship phone launch conferences of Zhiyun Group three times. He also represented Google in the "International Artificial Intelligence Summit" held in Shenzhen last year. After the summit, he was also invited to a small cocktail party hosted by Xu Shenxue!

At that small cocktail party, there were all the top business tycoons from all over the world...either the actual bosses of large high-tech companies like Sergey, or the bosses of large physical companies, and some people from various consortiums who were very low-key in front of the media but basically knew everything they should know.

Sergei, who has been to Shenzhen many times, thinks he has some understanding of the city, but he is still shocked by its rapid development in the past seven or eight years.

Every time I come here, I can see something new.

This city, like Zhiyun Group, is constantly updating and innovating.

Similarly, Sergei also discovered that this city also bore a lot of traces of Xu Shenxue!
For example, the car he is riding in now is not a Mercedes-Benz or BMW, but a bulletproof version of the Hai Lan SLC produced by the Hai Lan Automobile Company.

This model has been considered by many media personnel and automotive practitioners as one of the world's top four luxury executive sedans, alongside the Mercedes-Benz S-Class, BMW 7 Series and Rolls-Royce!

It is the car of choice for many wealthy people in China, as well as some wealthy people in high-tech industries overseas!
The SLC is not only a large luxury car, but also a high-tech one!
That’s not all. Sergey discovered that there were many vehicles with green license plates on the road, most of which were Hailan cars, but there were also some cars of other brands.

"Their electric vehicles are developing very fast, and people are very receptive to them."

"No wonder GM and Ford are so determined to raise tariffs on cars. If they don't, Detroit's lazy style will sooner or later put them out of business by Chinese electric cars, and even Tesla will find it hard to resist!"

Sergey only needed to take a look at the various new energy vehicles with green license plates on the streets, and think about some information he had seen recently about Hailan cars, and he was able to draw some basic conclusions instantly.

Those people in Detroit are definitely doomed!

When it comes to gasoline cars, we can’t beat the Japanese.

When it comes to emerging electric cars, we can’t compete with the Chinese.

Moreover, there is also Tesla in the United States. Musk dreams of taking over Detroit... How can Tesla develop without destroying General Motors and Ford?

Although many people say that Tesla's competitor is Hailan Automobile, this is a very superficial statement.

The smart electric vehicle market is still a super blue ocean market. The biggest competitor of smart electric vehicles is the traditional fuel vehicles that still occupy the mainstream of the market.

The global annual automobile sales volume is 70 to 80 million, and only a few of them are smart electric vehicles. At most, there are only 2 million.

Everyone in the industry knows that whether it is Hailan Automobile or Tesla, they seem to be competitors, but in fact they are allies... Didn’t you see that they just signed a strategic cooperation agreement in the first half of the year?

Crucial supply contract for solid-state batteries and vehicle-mounted computing chips!

Tesla's China factory and Hailan Auto's Mexican factory hold shares in each other!
Even a fool can see that the world's number one and number two smart electric vehicle companies have reached a strategic cooperation agreement.

And who are their enemies?

All traditional oil vehicle manufacturers in the world!

The automobile market is worth 70 to 80 million vehicles. Even if smart electric vehicles take up half of the market share, there will be 30 to 40 million vehicles.

As the leading and second largest players in the smart electric vehicle industry, they rely on the local market while expanding overseas markets, and finally each take up a market share of several million vehicles. Is this reasonable?
However, those traditional oil vehicle manufacturers are not stupid. Many of them have already begun to transform, and the automobile manufacturers in Europe and China are the most radical.

European automakers have already started the transformation, but the transformation has not been smooth...

China's traditional automobile manufacturers also transformed very early, but their transformation was much smoother.

In addition to BYD, other traditional automobile manufacturers such as GAC, BAIC, and SAIC have begun to transform and launch smart electric vehicles from their main brands one after another.

With the support of various new energy vehicle policies this year, some achievements have been made to a greater or lesser extent.

Even though it cannot be compared to the strong HaiLan Automobile, the future is promising.

Seeing the recent development of smart electric vehicles reminded Sergey of the situation in the mobile phone market many years ago!
Back then, Zhiyun Group and Fruit Company seemed to be competing in this way, but in fact they joined forces to clear the field and finally eliminated a large number of traditional mobile phone manufacturers.

In Sergey's opinion, the current automobile market seems to be experiencing a similar thing. Of course, he also knows that the automobile market is more special and complex. Smart electric vehicles cannot sweep the world as quickly as smart phones and bury a large number of traditional mobile phone manufacturers in just a few years.

Now, 100% automobile tariffs are coming!
It is still unknown how much waves the subsequent disputes surrounding smart electric vehicles will cause.

Of course, this has little to do with Sergey himself. He doesn't play with traditional cars. Even if he wants to play with cars, he would play with driverless cars. Unfortunately, the driverless car project has been around for so many years, but there has been no result.

In the final analysis, it is because our company’s artificial intelligence technology is not good enough!

Why doesn’t artificial intelligence technology work?
On the one hand, there is certainly a shortage of top engineers... A third-party headhunting company released a report that 80% of the top talents in the field of artificial intelligence are in China.

Most of China's top artificial intelligence talents are in Zhiyun Group, and the remaining few are in major Internet companies.

Zhiyun Group alone has most of the world's top talents in the field of artificial intelligence!
This is really enviable and jealous!
In addition to the shortage of talent, another problem is the lack of hardware.

To be precise, it is the server GPU, or to be more precise, it is the shortcoming of the APO graphics card!
The computing power required to train and run artificial intelligence is too huge. Whether it is the underlying artificial intelligence algorithm developed by Google or the open source GTAI2 algorithm of Zhiyun Group, the demand for computing power is huge.

If you want to train a reliable artificial intelligence, you will need a GPU server array consisting of tens of thousands of APO graphics cards.

If you want to do super-large-scale generative AI, you will need more.

In the past few years, Google has been purchasing a large number of APO graphics cards from Zhiyun Group, but due to the limited production capacity of APO graphics cards, the actual number of APO graphics cards obtained is still limited.

At least for Google, it is far from enough!

Today's Google requires more graphics cards, and more advanced graphics cards, not old ones.

Especially now that chip technology is developing rapidly, the iteration speed of APO graphics cards is also very fast. A few years ago, when the APO2000 and APO3000 graphics cards were launched, everyone talked about how powerful their computing power was. After all, these two graphics cards both use chips with double exposure technology, and their computing power was indeed very impressive back then.

However, just one year later, Zhiyun Group launched the APO4000 graphics card, which uses a 200-nanometer process and has much higher computing power than APO3000/.

This is not the end of the story. A year later, at the beginning of this year, the APO4500 using a -nanometer process appeared, further improving performance.

In October this year, Zhiyun launched the APO4600, which also uses the -nanometer process, further improving performance.

Among the above-mentioned graphics cards, each generation has greatly improved performance... Of course, the price is also getting more and more expensive.

In the artificial intelligence industry, when measuring computing power resources, it is often calculated using graphics cards equivalent to APO4500. The APO graphics cards currently owned by Google are APO2000/3000 with 4000-nanometer process, and APO with -nanometer process accounts for the majority.

There is also APO4500 using the nm process, but the proportion is not large. The main reason is that this graphics card is in serious shortage, and even big customers like Google cannot get sufficient supply.

If calculated as the equivalent of APO4500, Google's current computing power resources are only equivalent to about 2... It sounds like a lot, but if it is used to deploy GTAI, it is likely not enough...

Google has too many users. Once GTAI2 is directly deployed and provided to users for free, the huge number of Google users will overwhelm their GPU servers in minutes.

Zhiyun Group uses GPU servers to deploy Yun AI. Because it uses AI graphics cards, it has stronger performance and lower computing costs.

At the same time, the underlying algorithm of Yun AI used by Zhiyun Group is four or five generations more advanced than GTAI2, and the demand for computing power is actually lower.

But even so, the computing power of the AI ​​series graphics cards used to train, deploy and run Yun AI is equivalent to that of 4500 APOs, and this number is still increasing.

From this data comparison, we can understand why Sergey came to Shenzhen specifically!

He came to buy the APO4600 graphics card!
Google's desire for high-performance APO graphics cards led him to come here in person, and when he came, he did not come empty-handed, he also brought a blank check!
As long as Xu Shenxue is willing to sell graphics cards in large quantities, Xu Shenxue can fill in any amount!

If Sergei frowns even once, he will lose!

We, Gu Gou, are not short of money!

(End of this chapter)