As one of the American CEOs who is most familiar with Chinese culture, Zuckerberg won a big victory before the Lunar New Year in the Year of the Dragon. On February 2, Meta released its unaudited fourth quarter and full-year financial report for 2023. The report showed that Meta’s 2023Q4 revenue was US$40.111 billion, a year-on-year increase of 25%; net profit was 140. 1.7 billion US dollars, a year-on-year increase of 201%; Meta's fiscal year 2023 revenue is 134.902 billion US dollars, a year-on-year increase of 16%; as for profit, it is 47.651 billion US dollars, a year-on-year increase of 62%.
Since then, there has been a series of explosive news, which has been widely circulated in recent days: Meta paid dividends to shareholders for the first time in history; Meta's market value increased by US$206.14 billion in a single day, becoming the company with the largest single-day increase in market value in the history of the US stock market; Meta's full-year US stock income reached the highest record in the company's history.
All changes occurred within one year. Meta's market value soared from US$315.5 billion to US$1.2 trillion, an increase of almost US$880 billion. Based on the time node at the end of 2023, it is approximately equivalent to 19 JD.com.
This has elevated Zuckerberg to one of the most legendary CEOs in the United States, and is bound to become the most controversial topic figure. At the same time, it leaves everyone with a question: What has this guy done?
One year, from quagmire to heaven
If you stand at the beginning of 2023 and look back at Meta, you will feel that this company is going to be finished.
At that time, in February 2023, Meta released its 2022 full-year and Q4 financial reports. Quarterly revenue fell by 4%, and profits dropped by 55%. This announced a complete failure in the Metaverse that has lasted for one year since Facebook changed its name to Meta. RealityLabs, the much-watched Metaverse business unit, lost $13.7 billion a year, exceeding Microsoft’s total investment in OpenAI.
More hidden dangers lie in changes in the external environment.
In April 2021, Apple adjusted the privacy policy of the iOS system - APP developers need to obtain user consent before they can track their operating information on other APPs. The proprietary term is called advertising identification (IDFA). Previously, social media companies like Facebook, whose main source of profit was online advertising, relied heavily on IDFA for customized advertising push.
This also directly led Zuckerberg to turn to the Metaverse. He urgently needed a new entrance that could bypass the operating system in order to reconstruct the basic logic of user information collection.
In October 2022, Musk took over Twitter and immediately began to significantly lay off employees, optimize Twitter's cost structure, and increase profitability. On the one hand, people believe in Musk and believe that this will bring huge pressure to Meta; on the other hand, layoffs are always a rough but effective management method. This further weakened the capital market's confidence in Meta.
But what most people didn't expect was that this was actually the last time Meta showed its decline so far. In Q1 of 2023, Zuckerberg finalized the second round of layoffs of 10,000 people, and the company's stock price rose in response. Everyone agrees with Zuckerberg’s statement that layoffs will help improve work efficiency. Facts have proved that Meta's financial reports have indeed begun to rise steadily since then.
In addition, starting from Q1 of 2023, the development trend of the global online advertising business has fully complied with Snap CEO Evan Spiegel’s expectations: the decline in advertising seems to be bottoming out.
The advertising business revenue of Meta, Google, Snap and other companies will gradually begin to recover in 2023. In the third quarter, Meta’s financial report showed that advertising business revenue was US$33.643 billion, a year-on-year increase of 24%, accounting for more than 98.5% of its revenue in the quarter.
If the above are all external factors, then Meta’s own investment and success at the AI level have become the internal factors that turned defeat into victory.
Zuckerberg is like a poker player with strong capital and excellent patience. In the most difficult times, he does everything possible to stay at the poker table. When the east wind blows, he starts a rapid attack.
Apple’s calculations are no match for the “new industrial revolution”
Some reports about Apple's privacy policy changes will follow this main line: Apple discovered that many companies are reselling users' private information, Cook publicly criticized the data collection behavior, and the Apple privacy team decided to take action to put the devil back in the box.
But placing this kind of "privacy man"-like persona on Apple's head feels a bit inconsistent.
After all, not only are Apple's product profits ridiculously high, but the scandal of iOS 15 secretly scanning users' photo albums has long been made public. In addition, although the modification of the privacy policy will harm companies that focus on online advertising in the iOS ecosystem, it will be beneficial to the development of Apple's own search advertising business.
After the implementation of this privacy policy in 2021, Apple's market share in search advertising has more than tripled in six months - as an advertiser, although you can no longer easily obtain IDFA (advertising identification) to achieve accurate advertising recommendations, you can purchase Apple's search ads so that your APP will always appear next to the APPStore search box.
Location of Apple search ads, picture from Apple’s official website
This is a privilege to control user entrance from upstream. Apple can do it, Google can do it, but Meta cannot.
The relationship between Apple and Meta is actually quite tense. There is direct competition between the two parties at the level of instant messaging tools, direct competition at the level of XR, and competition at the level of advertising business. In 2018, Facebook was caught in the Cambridge Analytica scandal. A reporter asked Cook, what would you do if you were Zuckerberg? Cook replied, I wouldn't be put in his situation. In 2020, when Apple began to push for changes to its privacy policy, Meta published a report in the New York Times, accusing Apple of not caring about the lives of small and medium-sized enterprises. It's just that this kind of resistance, which is limited to the public relations level, is quite weak.
The real turning point will come at the end of 2022, when the arrival of generative AI completely muddies the waters.
After the release of LLama2, Meta became de facto the most successful company in developing open source large models. This success is directly reflected in its two online advertising products, one called MetaAdvantage and the other called AISandbox.
MetaAdvantage was launched in March 2022. Its core is to introduce machine learning capabilities into the production and delivery of advertisements. This product has become Meta's main focus in fighting the trough period. The core of MetaAdvantage is to gradually transfer human decision-making to AI in the advertising and marketing process.
For example, Meta's advertising allows advertisers to configure target groups, but AdvantageDetailedTargeting allows Meta to extend ad targeting to groups other than the preset target audience. Meta's official announcement is that after enabling AdvantageDetailedTargeting, conversion costs are reduced by 37%.
Another example is Advantage+Creative, which can automatically optimize advertising effects, including screen filters, text positions, etc.; Advantage+Placements, which allows the platform to help advertisers optimize placements; Advantage+shoppingcampaigns function is to optimize promotional activities.
A set of MetaAdvantage tool applications, according to Meta's official promotional figures, can reduce CPA (cost per action, such as billing every time a user clicks on an ad) by 22%.
AISandbox is a more radical AI experiment method after the initial success of Advantage - it is a sandbox, which was opened to a small number of users in the early stage to test the most advanced AI for Advertising functions, mainly generative artificial intelligence, such as automatically generating multiple versions of advertising copy, generating images based on text, and automatically adjusting image sizes according to the resource requirements of each Meta platform.
This series of achievements in the field of AI initially relied entirely on Yann LeCun, Meta's most famous AI scientist, and the FAIR (Fundamental Artificial Intelligence Laboratory) he established. However, OpenAI's achievements suddenly exposed the possible conflicts between Zuckerberg and Yang LeCun to the public - on the one hand, everyone accused Yann LeCun of being too academic in his management and research style. Technology has led to Meta’s AI level being excluded from the first echelon in business (in May 2023, US President Biden and Vice President Harris held a White House AI Summit and invited Microsoft, Google, OpenAI, and Anthroic, but it was not called Meta); on the other hand, judging from the content of the tweet, Yang Likun does not seem to be very optimistic about the development of large models.
In February 2023, Zuckerberg made a decisive adjustment and announced that Meta would form a top product development group to focus exclusively on content-generated AI technology. Some media revealed that the team’s formation information had just appeared on Meta’s intranet, and more than 2,000 job transfer applications were immediately received, because no matter how Meta lays off employees, the artificial intelligence department is absolutely safe.
Interestingly, the head of the new department is Vice President Ahmed Adar, who has worked at Apple for 16 years and now reports to Meta Chief Product Officer Chris Cox.
The person on the left is Ahmed Aidar
Meta's AI focus has since gradually shifted to Adar and his new department, and the industry regards him as a key figure in driving Meta's generative AI forward. As for Yang Likun's FAIR, it became turbulent. In September 2023, the media broke the news that the FAIR team had internal strife due to uneven distribution of computing power, and half of the core authors of several key papers resigned.
New products and new leaders have greatly strengthened advertisers' confidence in Meta. Although data shows that after Apple's privacy policy came into effect, advertisers' costs increased by 37%, the success of LLama2 makes everyone believe that Meta's progress in generative AI can gradually regain the initiative.
Open source has become Meta’s biggest reliance
No matter what kind of controversy Yang Likun is burdened with, using an open source strategy to fight OpenAI and occupy a place in the era of generative AI is definitely an extremely successful plan. LLama2's performance is equivalent to GPT-3.5 in several mainstream test benchmarks. Once the performance of the open source model is close to that of the closed source model, no one will pay for closed source anymore.
LLama2 received more than 150,000 download requests within a week of its release. Currently, LLama2 has gained 8.5kforks and 41 Contributors on Github, and 27 Teammembers on Huggingface.
A large amount of usage feedback and code contributions from all over the world have made LLama2 evolve rapidly. Zuckerberg himself also talked about LLama and open source strategy during the earnings call in February this year.
He explained: Open source AI technology is to promote technological innovation, improve model quality, establish industry standards, attract talents, increase transparency and support long-term strategic considerations. Meta's strategy has long been to build and open source common infrastructure while retaining specific products and implementations as proprietary technologies.
The key to this is that open source has more advantages in AI reliability and review because the code is open and transparent, and because it attracts a large number of external contributors, LLama itself is closer to industry standards, allowing Meta to quickly become the first echelon in the generative AI competition.
In addition, the open source LLama2 has also become the best publicity for MetaAI's capabilities. LLama's pedestal status in the large model industry is an important reason why advertisers are willing to try it and the secondary market is optimistic about Meta.
The open source LLama helps Meta complete the iteration of its technical infrastructure, and Adar from Apple helps it achieve productization and commercialization. It has become the core layout of Zuckerberg's leadership of Meta into 2024. This makes public moral condemnation very weak. Zuckerberg can apologize and count the money at the same time.
2024 is likely to become a watershed. With the blessing of AI, social media is breaking away from the shackles of public opinion and morality. With its powerful content generation capabilities and recommendation mechanisms, it can more effectively increase monthly activity, improve profitability, and accelerate its growth, thus transforming into a brand new species different from the mobile Internet era.