Visual China.
Text |Prophet's Hall.Jia returned to China, "Let's suffocate our dreams together", shareholders and creditors were completely suffocated.
Zuckerberg, burning 400 billion in ten years is suffocating 10, in the past two or three years, more than 150 billion yuan has been burned to buy GPUs to suffocate the dream of large models 20。Today, the market is no longer hesitant, after the release of the 23Q4 earnings report, Meta has the highest increase among the seven major U.S. stock technology companies (Apple, Google, Amazon, Microsoft, Meta, Nvidia, Tesla) - about 20% - as evidence.
Compared with the attitude of other technology companies at home and abroad and Meta chasing emerging technology tracks, you may think that the latter is really "stupid", stupid to the point of bubbling!
As everyone knows, it is the above-mentioned seemingly stupid behavior (or persistence, doing something, not doing something) that makes the capital market willing to believe the story told by Zuckerberg, not worrying about major investments becoming sunk costs, and believing that Meta will inevitably attend (at least late) in the next era, and that Meta's success is the success of all shareholders.
It is necessary to observe Meta from the details and observe this company with a good layout for the future. After all, in a world where the technology cycle is rapidly transitioning from depression to boom, we can't afford to fail overnight.
An AI navigation seen online**.
Tao Te Ching - will take what you want, you must stick to it.
Since Meta (then Facebook) acquired Oculus in April 2014, the loss of the VR business has always been like a sword of Damocles, hanging over the heads of Meta shareholders.
From 2019 to 2023, Meta disclosed the financial figures of Reality Labs, with an operating loss of up to $51.2 billion. 2014.4-2018 did not disclose relevant figures, but the operating loss of at least $5 billion in five years is probably inescapable.
Reality Labs has a total loss of $56.2 billion, equivalent to 400 billion yuan, and this is just the beginning.
Because it's hard to say how long the huge losses of the Reality Labs division will continue, anyway, the burning of money will not stagnate in 2024, with an operating loss of $20 billion! Meta is well prepared, with this paragraph in its earnings report: "We expect Reality Labs operating losses to increase significantly in 2024." ”
In the past five years, the proportion of Reality Labs in total revenue has never exceeded 2%, and the related revenue will decline by 12% year-on-year in 2023, so we are ready to invest heavily in 2024.
Look at the smart byte, in early 2022, it acquired PICO, a leader in China's VR equipment, and in early November 2023, nearly half of PICO's employees were transferred or laid off, and Byte decisively withdrew. In contrast, Meta, which acquired Oculus, has a huge loss in sight for ten years, and if it continues to burn tens of billions of dollars a year for three or five years, it is not impossible for trillions of yuan to smash into it. The so-called head iron, but so, even the head iron has been unable to express, poverty limits my imagination.
Of course, the investment in the metaverse is still only the beginning, and this stubble has not yet blossomed. Zuckerberg is on the big model again.
Zuckerberg revealed that by the end of this year, Meta will have 350,000 Nvidia H100 (150,000 in 2023), and if other GPUs are included, it will have about 600,000 H100 equivalent computing power, what is this concept?
The ** of a single sheet h100 is 2550,000 yuan, 600,000 is 153 billion yuan. Yes, Meta will spend 153 billion yuan just to buy cards and run large models in the past few years. Last year, we listened to experts in the first meeting to say that the threshold fee for large models is 10 billion yuan, which is still too young.
The capital expenditure of the GPU in the optical infrastructure is already astronomical, but Meta's large model capability is still "relatively unworkable", just to name two examples:
What else can I say, praise it for iFLYTEK, quack and chirp. In contrast, the capabilities of the meta large model are simply stretched.
Zuckerberg said that LLAMA 3 will have stronger ** generation capabilities and advanced reasoning and planning capabilities similar to Google's GEMINI model, and the performance of this model is comparable to GPT-4, and is even expected to reach a higher level - well, LLAMA 3 has not yet been released, not even PPT, that is, the gap with iFLYTEK is not a little bit
One of the most commendable things is thatThe large-scale model research and development of iFLYTEK is simply turning decay into magic.
Meta still has to spend more than 150 billion yuan to buy a GPU running model (the company's capital expenditure in 2023 is about 195 billion yuan) before it is expected to catch up with the GPT-4 level in 2024, and the annual capital expenditure is only about 10 billion yuan (10.9 billion and 8.3 billion in 2021 and 2022, respectively), and iFLYTEK's capital expenditure from January to September 2023 is 17500 million yuan, the investment is insignificant compared with meta, but the large model developed is more powerful than meta, I don't know which god can tell me, is the R&D personnel of these two companies too powerful, or there is another winning R&D secret, or others. **Waiting for the answer, it's in a hurry.
To add to this, the talent barriers of large models are very high. Bezos, the founder of Amazon, has a point of view: there may be two or three hundred genius researchers around the world who have actually contributed to the large model, and the agglomeration effect of genius scientists is very strong.
Whether it is the metaverse or a large model, there is absolutely a world of difference between players entering the cutting-edge track and whether they have cash cow business support or not. With that diamond diamond, picking up porcelain is extremely reassuring, and the possibility of playing tickets and cutting leeks is greatly reduced. Of course, if you're a pioneer like OpenAI who has delivered amazing transcripts (ChatGPT active users and considerable income), and there is no shortage of venture capital chasing, then I didn't say it.
The cash cow business is very important, but different companies have different attitudes towards protecting cash cows.
Like Apple, it does everything to protect the iPhone and iOS, and has practiced an amazing Tai Chi Kung Fu for this purpose.
The European Union's Digital Markets Act, which will come into effect in March this year, targets Apple's monopoly on app distribution, requiring the latter to allow users to use apps outside of the App Store, if Apple still wants to sell their phones in the EU. In this regard, Apple directly launched a protocol, which can be used in other channels, but the first installation of the app has to pay 05 euros to Apple,**More than a million times of the app also have to pay 0 for each new ** after that5 euros for Apple. Developers choose. The EU is dumbfounded.
It's easy to confront the EU, and it's even easier to deal with individual companies. The game company Epic (Fortnite developer) sued the "Apple Tax" in-app purchase monopoly, Apple directly updated the developer agreement to allow app developers to provide third-party payments, but sorry: third-party payments also have to charge commissions, from 30% of Apple's tax to 27%, counting third-party payments also have to make money, which means that the total commission has not changed, and even third-party payments will be higher. Anyway, Apple offers other options, so you can't say I'm a monopoly at this point.
Apple's Tai Chi masters are mostly focused on legal affairs. What Yuhang Evergreen, Nanshan Pizza Hut, Haidian Tumbler, and Longgang invincible opponents in China may not be one in front of Apple's legal counsel. With the success of Apple's various devices and systems, it is more and more common for external innovators to use patents to protect their interests. In order to block them and maximize Apple's own interests, Apple's legal Tai Chi has rapidly improved, for example, Apple is now a typical representative of "reverse hijacking" patentees.
The so-called "reverse hijacking" of patents is to use a strong legal department, social resources and financial advantages to invalidate the patent with all its might, and if the patent cannot be invalidated, it will take several years or more to fight a lawsuit, so that the other party will be exhausted, or even die directly (most scientific and technological innovation enterprises will survive).
four or five years), Apple will win.
If you look at it again, Meta is simply a good baby.
On August 1, 2023, Meta announced that it would change the legal basis for processing user data from "legitimate interest" to "consent". This means that Europeans have the option to say about tracking and profiling for advertising purposes"Disagree"。
For a long time, user data has been the anchor of the platform's advertising recommendation accuracy, and more data leads to more accurate advertising. Conversely, if there is less data, advertising revenue will be greatly affected. If there are no other favorable changes, it will be difficult for Meta to make up for the losses even if it launches an ad-free subscription app.
Some people in the legal profession believe that Meta is not so easy to compromise, and I don't think there is such a concern. As long as you understand the dynamics of the industry and Meta, you will find that even if it cannot collect user data to a greater extent, it can find new operational magic weapons at the technical level.
On April 27, 2021, Apple released iOS 145, in fact, it is a preview of the impact of the EU "Data Markets Law" on Meta's advertising business, the iOS version includes the new provisions of the App Tracking Transparency Privacy Function, only when the user clicks "Authorize and Agree", the App can track the user's actions in other apps and ** and obtain data.
Undoubtedly, the vast majority of users choose to "ask apps not to track". Meta also suffered a heavy blow as a result, and this heavy blow was not so obvious in the first year (2021), although Meta lost $10 billion in the first nine months, but the "mask reason", the global Internet application ARPU cooking oil, and the user update system was not timely. After 1 year, the damage was really reflected when the iPhone carried a new system, and the ARPU trend after 21Q1 was obvious.
However, in Q4 2023, Meta's ARPU will hit 13A new high of $12. What's driving up its ARPU, the "advantage tool" is one of the most critical variables. To put it simply, the Advantage tool brings generative AI advertising to advertisers, using algorithms to test the effectiveness of their ads and select the most effective advertising options (automatic changes**).
Even though the pool is smaller (less information gathering makes the acquisition of potential users smaller), the probability of being hooked and conversions for each fish in the pool is also larger, and the advertiser's IRR is improved. Advertisers are more efficient at creating ads and have significantly lower costs.
Some companies are stuck in the cold and cannot grow, some companies rely on slogans to fantasize about growth, some companies rely on unhealthy operating models (losses) for inefficient growth, some companies rely on legal overbearing growth, and companies like Microsoft and Meta have really used AGI, AIGC and other technologies to regain growth and lead the trend. You could say it's inadequateBut what seems to be the hardest path is often the shortest path.
Code here,I found that the number of words is seriously exceeded.,The remaining points of meta are stupid, just say it simply.。
First, the "super data" will no longer be disclosed, and Tencent will be stunned when the bayonet sees the red roll effect.
Whether it's network effects, or value emergence. In the field of social software, the huge user group itself is a symbol of enterprise value. Every time Tencent discloses its quarterly results, it always broadcasts the number of MAUs on WeChat in a conspicuous position, and the 23Q3 report is on the third page, 13400 million users.
The phrase, "I have 1.3 billion users," represents a hand that could completely overshadow Carrys' boast of Daenerys Stormborn of House Targaryen rightful heir to the iron throne. rightful queen of the andals and the first men. protector of the seven kingdoms. the mother of dragons. the khaleesi of the great grass sea. the unburnt,the breaker of chains.。
As for Meta, the number of users of the app family bucket is as high as 39800 million. However, it made it clear in the financial report that starting from 24Q1, it will no longer disclose data such as DAU and MAU. Instead, it was decided to see the bayonet red, and further disclose the number of impressions of advertisements and changes in advertising, in the context of increasingly strengthened user privacy protection, only AI technology can have the confidence to raise the number of advertisements and **.
The second is that Zuckerberg will not fool, the PPT performance is not good, and he is not worthy of suffocating his dreams.
At last year's Connect conference, about 30 minutes, Zuckerberg improvised a demonstration of pulling across. Don't ask you to "suffocate your dreams", make a ** in advance, and pretend that the AI version of Snoop Dogg is very smart. At least you have to set off the atmosphere in place, don't be embarrassed, buddy, there's no one in this speech
The third is that "the cognition is not good", and the maximum value can be created by not knowing how to stay in the company.
Is it not good to spend money in the company, a cumulative repurchase of 220 billion US dollars before 2022 and 200 in 2023$300 million, and then said an additional $50 billion in buyback quota. Not only that, but 0A quarterly cash dividend of $5 is paid quarterly and every quarter thereafter. Are you an internet company, that's stupid. You learn, the $5 billion repurchase quota will be repurchased by Q3 last year$5.1 billion.
Compared with some other companies, Meta's seemingly silly behavior actually expresses a few words:If you are willing to bet heavily on the future, the outside world will believe that you have a dream; The repurchase and dividend payout of real gold and platinum, the outside world is willing to believe that you have the ability to realize your dreams and share success with shareholders; As well as sincerity is powerful**. Otherwise, Meta, whose profits all come from advertising, how can it reap the sky-high wealth of the stock price after the financial report.