US Stock Ads: After TikTok, is ChatGPT going to start a new "revolution"?

Hello everyone, I am Dolphin Analyst!

The lifeblood of advertising is macroeconomics, and the advertising market in 2022 is expected to fall into dust as the global economy slows down. In the past year, management of advertising giants mentioned economic expectations are confused and advertising expectations are repeatedly conservative. However, after surviving the downward trend of 2022, the landing of the economy seems not to be as bad as imagined. Although 2023 is likely to still be a sluggish environment, at least the expectation of a strong recession is untenable. Perhaps advertisers will remain cautious in the first half of the year, but the growth rate of the advertising market in the second half of the year is likely to rebound before the economy.

Under such a large cycle, Internet advertising giants are not only facing pressure on income, but also eating bitter fruit for their blind expansion in the past three years. As expected, waves of layoffs are coming, and the profits of giants will see turning points before income under a wave of cost reduction and efficiency increase. Dolphin Analyst expects to see significant signs of recovery as early as the second quarter.
In terms of Google-C.US and Meta.US that Dolphin Analyst pays attention to, Meta, which compromised first, regained its investors, but the market still expects to see greater cost reduction and efficiency increase. As for Google, it was hit hard by ChatGPT and fell into confusion, but its profound AI technology and long-term accumulation of search data made Dolphin Analyst feel that there is no need to panic for the time being. Although a big ship is difficult to turn around, Google is not unprepared. The worrying thing is the cost increase brought by AI to the entire industry.

In the following, this article will start from the industry, focusing on the performance of advertising giants in the fourth quarter and discussing the logic changes and valuation judgments of Google and Meta.

I. The expectation of "recession" switching to "weak growth"; advertising leading the bottoming out?

The financial reports of advertising companies such as Snap, Meta, and Applovin in the fourth quarter had poor performance but their guidance was not bad. At the performance conference call, although the management of various companies did not express particularly firm trends for the confusing macroeconomics, there was no worry about the serious deterioration of the macroeconomy in their words. The management of mobile advertising company Applovin even believes that the bottom of mobile advertising will come in the short term.

So why will there be signs of bottoming out first in advertising under the generally expected economic slowdown in 2023?

The advertising market is closely related to macroeconomics, but the expectations and planning of advertisers are difficult to be completely consistent with the actual economic growth rate. With stable economic development, the proportion of advertising in GDP also remains stable. The US advertising/GDP has been around 1% for many years. In the past two years, due to the dividend of onlineization, the proportion of digital advertising has significantly increased the overall proportion of advertising in GDP. However, at turning points in the economic cycle, marketing expenses usually become the first choice for the elastic expansion and contraction of corporate operations. Therefore, the change in the advertising market is often significantly higher than the actual economic changes.

Over the past decade, although the US economy has gone through different cycles, such as the significant slowdown in economic growth in 2015 and 2016, the continuous increase in digital advertising penetration has basically smoothed out the fluctuations in advertising budgets caused by economic cycles.

However, in the past three years, due to the outbreak and withdrawal of the epidemic, as well as policy disturbances imposed by the government, the short-term economic changes have been significant, resulting in more severe fluctuations in the advertising market.

If we further analyze the situation of each quarter, especially for Internet giants that have benefited from the online dividend, the comparison between their advertising revenue and GDP changes shows a steeper slope. Of course, this also includes the impact of the passive pressure on revenue caused by the high exchange rate of the US dollar for non-US region revenue of the giants. However, it can still reflect that the elasticity of advertising fluctuations will be much higher than actual economic changes when the economy enters the turning point stage of different cycles.

For macroeconomics, the possible maximum expected change at present is the difference between the economy transitioning from "strong recession" to "weak growth/recession", which will significantly affect the marketing investment rhythm of advertisers this year.

According to the research conducted by some advertising agencies at the beginning of the year, the period when advertisers were most cautious about economic expectations was basically in the second half of last year, especially when the marketing budget was rapidly adjusted in the third quarter. Although the expectations were slightly more cautious in the fourth quarter, compared with the third quarter, they did not further deteriorate.

Corresponding to the advertising revenue of the giants, it reflects as follows:

① In the third quarter, non-e-commerce advertising collapsed rapidly. Besides the shrinkage of the budget size, many advertisers also shifted to e-commerce advertising closer to the transaction link (higher conversion rate) in the short term (Amazon's growth rate increased in the third quarter).

② In the fourth quarter, the advertising revenue growth rate of giants continued to decline, but the decline was already slowing down compared with the third quarter.

Therefore, from the perspective of the revised expectation of "weak economic growth", the overly cautious investment rhythm in the third and fourth quarters should not be maintained throughout the year. Perhaps there will still be a conservative mentality in the first half of the year, but as the expectation of economic soft landing becomes stronger, advertisers' marketing activities will quickly return to normal. In addition to the low base last year, it is expected that the advertising market in the second half of this year will show a relatively significant growth recovery.

The following is the analysis of Google and Meta, the key advertising companies covered, respectively, by Dolphin Analyst. ** 二、 Google: ChatGPT's "disruption" crisis?**

In the past, when discussing Google's logic, the discussion of macro impact would be much greater than Google's own operations. This is due to the fact that the search business contributes more than half of revenue and may contribute more than 90% of profit, which has been a deep moat for a long time.

In today's competition of mobile applications, Google still occupies 92% of the market share of overall search traffic. It can be said that if it is a search engine with the same features, there is basically no ability to shake Google's market position.

The appearance of ChatGPT will have a short-term impact on user traffic due to its technological first-mover advantage. However, the actions of backend advertisers may not be so fast. Perhaps some advertisers will adjust the proportion of their budgets due to Bing's increased traffic in the short term, but in the medium and long term, they still need to consider user retention, activity, product response and overall Google's investment plan.

At the same time, how to balance the commercialization of ChatGPT ads with the original commercialization logic of search redirection in the short term is also a problem that Bing needs to consider. However, this gives Google a window of time to improve Bard in a timely manner. Although Google relies on search revenue, its C-end products constitute a traffic ecosystem of nearly 3 billion globally, which cannot be overturned in a short time.

Therefore, we believe that the most likely stable state of the search market is to maintain the current competitive structure unchanged (or Google slightly lower), but the profit margin of the entire search market will be permanently reduced due to the increased cost investment of AI functions.

1. Clarifying ChatGPT's "disruptive" logic

First of all, Dolphin Analyst wants to explain that we do not agree that the appearance of ChatGPT will necessarily cause a substantial disruption (i.e. reshuffle) of the existing search platform competition pattern, but rather a new way of interaction between users and search platforms, which can be viewed as a "technological revolution" that promotes progress of the entire search industry to a certain extent.

The main reasons are:

1) Short-term user experience differences are necessary but not sufficient factors.

The competition pattern of the existing search industry has multiple factors. For search applications, Bing, which has integrated ChatGPT, affects the user experience on the front end, but at the same time, the functional and commercialized scenarios behind the search platform are closely related to the platform's original product ecology and commercial cooperation ecology.

2) The accumulated data and algorithm models in history are the competitive advantages of search giants

Given that the advantage of search platforms in one of the three production elements of AI-Q&A training, data, is significant, the possibility of industry outsiders entering the search field through this opportunity is very low. Existing search giants such as Google, Microsoft, and Baidu have not only accumulated decades worth of historical data, but have also continuously invested in AI algorithms. Although the differences between these search giants require time to catch up, it does not necessarily mean that the gap will continue to widen.

Therefore, from a higher and more long-term perspective, ChatGPT (an AI-based Q&A system) represents a technological innovation in the current search engine landscape that brings new ways of interacting with search platforms and users. It will become an essential technology for future search platforms.

Now, let's discuss the "disruptive" nature of ChatGPT from a business perspective.

The emergence of ChatGPT has provided another way to query and obtain information by changing the way users interact with the platform.

Dolphin Analyst briefly outlines the technical evolution process, but doesn't elaborate. Those interested can consult academic papers.

The large-scale GPT-3.5 model behind ChatGPT belongs to the third generation of the GPT model series, which is an improvement of the Transformer model proposed in a 2017 Google paper. In addition to the GPT model, Google mainly improved the Transformer model into the BERT model. The main difference between BERT and GPT is that the underlying module architecture is different. In terms of front-end applications, Google's BERT model's core advantage lies in "natural language understanding," while OpenAI's GPT model's core advantage lies in "natural language generation."

OpenAI's ChatGPT can become popular not only because it is used in conversational chatbots and widespread among the general public but also because of the model's advantages (including data cleaning, labeling, model structure design, and training inference technology accumulation). If we further trace back, the model advantages of ChatGPT can be attributed to highlights:

① Large pre-training model (corpus) data

② Incorporating human feedback for supervised learning during training

① After three generations of evolution of GPT 3.5, it has significantly surpassed GPT1,2, and the standard version of the BERT model in terms of parameter quantity (Google's new version BERT model parameter quantity has reached 481 billion and the PaLM model parameter quantity has reached 540 billion) as well as pre-training data volume, which is now one order of magnitude higher.

② The most critical aspect of "silky" conversations with ChatGPT is the reinforcement learning by human feedback (RLHF) added to the model training compared to previous models. Currently, Google, Baidu, and other major manufacturers with large models rely more on unsupervised learning for training. Therefore, compared to its predecessors, ChatGPT can converse in a way that is more in line with human language habits, greatly reducing the "machine feeling."

The revelation of the secret of ChatGPT's training also represents the possibility for future generations to learn and catch up, especially since Google's underlying framework and data volume are not inferior. However, this change in interaction methods poses two thorny challenges to the commercial logic of traditional searches (or now Google's own absolute monopolist):

(1) The original commercial logic needs to be transformed

In simple terms, the biggest change brought about by ChatGPT compared to traditional search engines is its ability to organize and summarize information. This means that users no longer need to collect the desired information through jumps between different website addresses.

However, these jumps between websites are actually a necessary commercial function for traditional search businesses. Therefore, after changing this business logic, ChatGPT will naturally also change the way search businesses commercialize and monetize.

Therefore, for traditional search engines, finding a more fitting way to commercialize after the change in the way they interact with users is crucial. Otherwise, due to the disconnection of the webpage jump function, they will lose part of the existing advertisers' budgets.

(2) After AI conversation becomes a necessary function for search, the economic calculation needs to be recalculated

Cost is the most critical issue to be faced in the short and medium-term. After introducing AI to search, if the monetization logic at the front-end does not bring about a qualitative improvement in the efficiency of advertisers' ROI, the extra operating costs caused by users using AI functions will cause the bid price to not increase in sync with the cost. This will be reflected as a direct drop in the profit margin of search ads.

This is also a point that Microsoft CEO Nadella mentioned when talking about the application of ChatGPT to Bing: "…dedicated to reducing the profit margin of search…"

Of course, this means not only Google but also Bing's profit margin will decrease. However, for Bing, introducing ChatGPT ahead of time will quickly increase its user traffic in the short term (with daily downloads reaching 100,000, ten times the usual amount). Therefore, even if the original advertising conversion rate has not improved, due to the influx of fresh traffic, Bing will still partially gain the favor of advertisers in the short term. This is like last year when TikTok just started commercializing. Although its conversion rate was not high due to the age structure of users and the accuracy of recommendations, many advertisers still chose to allocate part of their budget to it because of the new traffic it brought.

On the other hand, for Google, it may still face a loss of some search share in the short term. Google's response can be to further reduce ad prices to reduce losses and affect ad profit margins.

But Dolphin Analyst believes that this short-term loss can be recovered after Bard is integrated into Google search engine. The real reason for the long-term decline in search profitability still comes from the high AI training costs and operational costs of using AI for dialogue with users.

(1) Training costs

According to research and public disclosure, OpenAI's ChatGPT training cost is about 20 million US dollars, of which offline training cost is 12 million US dollars, and the cost of retraining a large model with a 45TB corpus is about 4.6 million US dollars. The training cost theoretically increases with the increase of data scale.

However, if Google's current training plan for PaLM model is used and GPT3 is trained with Google's GCP TPU v4 chip, the training cost is only 1.4 million US dollars.

At the same time, because of the increase in model utilization and energy efficiency, the cost can also be greatly reduced by using the original training plan. According to estimates, if the latest H100 of NVIDIA is used, the internal cost will be further reduced.

Of course, for search giants that have already invested historically, the cost pressure in this area may be much less than that for new platforms without historical accumulation (such as the construction of basic training servers and the establishment of annotation personnel teams). However, to achieve the current level of ChatGPT intelligence, other giants still need to add training and debugging to the existing model.

As users' requirements for user experience become higher, especially the demand for content update frequency (currently, the data used for ChatGPT training is mainly before 2021), the expansion of the backend corpus and new training must keep up with the constant emergence of new content data. 2) Operational Costs

Although the training costs have exceeded millions of dollars, the main pressure on the ChatGPT product costs lies in the operational phase. In theory, it is difficult to determine the upper limit of the number of times ChatGPT runs the model during the interaction process due to user volume and interaction with ChatGPT that cannot be controlled.

Last year, Professor Tom Goldstein of the AI department at Marland University estimated the operating costs of OpenAI: The computing cost behind a 30-word response is $0.01.

Assuming an average of one query per day per user among the current 100 million registered users, this means that ChatGPT's daily operating cost (excluding labor costs, property and equipment costs, basic operating expenses, etc.) is close to $1.2 million.

If AI Q&A becomes an essential tool for search engines in the future:

According to Microsoft's recent disclosure, global users search 10 billion times a day. Assuming that one search represents one conversation (a 30-word response), if AI Q&A models like ChatGPT and Bard are used for all searches, it will cost nearly $100 million to global search engines each day.

For Google, which currently holds 92% of the search share, in extreme cases (100% use of AI Q&A), the expected cost increase for one year will be up to 33.6 billion yuan. If half of the search volume uses AI, the cost will also increase by 16.8 billion yuan, equivalent to 13% of Google's total cost in 2022.

However, the cost calculation for a single conversation above is based on a charge of $3 per hour for one A100 on Azure servers. For server platforms like Google's self-built underlying platform, internal costs can be halved, so the corresponding cost increment is around 8 billion yuan per year, which is equivalent to 6% of the total cost in 2022 and affects the operating profit margin by 3%.

Of course, if the word count of a single Q&A increases (such as from 30 words to 50 words), and more users use AI Q&A to search for content (search penetration rate exceeds 50%), the impact on Google's profit will be even greater.

2. Google's valuation: Derivation of three search competition scenarios in addition to invisible cost increases

Although Dolphin Analyst believes that within the six-month window, Google can launch a new search engine integrated with Bard and other work software, which can rely on the nearly 3 billion traffic ecology of its product matrix and cooperation with advertisers to maintain its leading position in the search field in the long term. But due to considerations of risk, we estimate that by 2025, Bing may have gained some market share from Google during the current window of opportunity for catching up, as well as through the use of office suite products and eroding the market share of other small and medium-sized search platforms.

① Optimistic expectations: The competitive pattern between Google and Microsoft remains unchanged, but the two giants will steadily increase their market share (+5pct) by eroding other small and medium search platforms. In addition, Bard can also provide external authorization to drive cloud business growth.

② Pessimistic scenario: If Google fails and gives up 10% market share (including the potential share of erosion of small and medium-sized platforms, the actual share given up is 15%), Bing's daily download peak after the launch of ChatGPT will be 100,000, and usually 10,000. Compared with Google's previous 300,000 daily stable download volume, the market share would be about 15%.

③ Neutral scenario: There is some impact, but Google's market share remains unchanged while Microsoft's market share increases.

Overall, under the three different competitive patterns, the CAGR of Google Search for the next five years is 10%, 8%, and 5%, respectively.

In terms of costs, as AI question-and-answer functions are becoming indispensable cost inputs, whether they can bring higher advertising efficiency to search engines and attract additional budgets from advertisers (or subscribe to individual users for a fee) remains to be seen. As a result, compared to previous expectations (Google's 4Q22 call: cost structure will be significantly improved in 2024), Dolphin Analyst has lowered its expectations for Google's cost optimization speed in 2023-2024, but it is expected to accelerate repair after 2025 with the improvement of cloud business profitability.

The final valuation result, assuming WACC = 9.59%, g = 2.5%:

① Under neutral expectations, Google DCF is valued at $15.2 trillion, with a single share at $117. Due to the short-term high yield of US bonds, if the risk-free rate assumption is raised from 3% to 3.5%, the corresponding values would be $14 trillion and $109 per share.

Under a pessimistic outlook, in order to compete defensively, Google's operating cost investment in AI Q&A remained unchanged, but advertising monetization did not expand synchronously, resulting in a decline in market share. DCF valuation is 11.6 trillion, or 90 USD/share (risk-free return rate is 3%).

Under an optimistic outlook, Google leverages AI capabilities to not only resist Microsoft's attack, but also erode the advertising market share of more small and medium-sized platforms, accelerating the increase in operating profit margin. The final DCF valuation is 17.5 trillion, or 136 USD/share (risk-free return rate is 3%).

Actually, the above predictions are based on two relatively conservative assumptions:

(1) The market size of the search industry has not attracted more advertising budgets due to the introduction of AI functions such as ChatGPT;

(2) Google Cloud has not been estimated to use Bard's AI capabilities to gain more market share with Microsoft.

Meta: Investment also needs to go "bubble"

Having gone through the tragic 2022 and moved towards the new year of Meta, the problem has decreased instead. As Dolphin Analyst detailed in the Meta Q4 earnings report review "Multiple Buffs Favorable, Meta's Gorgeous Turn?", the three issues surrounding Meta last year: Apple IDFA, TikTok, and unrestrained investment in VR all showed a trend of easing at the end of the year.

Of course, among the three issues, the impact of proactively reducing investment in VR on valuation and investor sentiment was the greatest. In the medium and long term, the competitive threat of TikTok that Meta faces will not stop here. Dolphin Analyst is more inclined to believe that political factors will not exist for a long time and thereby crush TikTok. Similarly, the impact of Apple ATT is more about marginal slowdown than objective disappearance, which will also cause some advertising budgets to shift to e-commerce platforms (Amazon).

Comparing Meta's operating profit margin before and after the epidemic (expanding investment in the Metaverse), even if based on the latest adjusted operating expense guidelines for 2023, the company's profitability level is still far off. Currently, the support for VR is still being emphasized in Meta's expenditure guidance of 89-95 billion in 2023.

However, it is difficult to significantly increase the monetization of VR hardware in the short term, but expectations for software revenue require imagination, resulting in the entire company still showing a business that burns billions of dollars a year and has a long way to go before reaching a profitable inflection point. But this does not conform to the operating decision-making of a rational management team, which will definitely endure pressure from investors.

Therefore, we may consider this irrational continuous investment as an operating assumption under extremely pessimistic expectations for Meta. Under a neutral outlook, the company is expected to continue to significantly reduce investment and significantly pull back the level of profit margin. And on the revenue side, although TikTok encountered obstacles in the short term and competitive pressures have eased somewhat, Dolphin Analyst believes that if viewed from a relatively neutral perspective, political factors will not have a significant impact on TikTok's development in the medium to long term, and advertisers' short-term caution will not become the norm.

Even eMarketer believes that TikTok's current low eCPM will allow it to quickly capture more market share. With its global traffic of 1.5 billion, including 900 million TikTok users outside China, it can achieve $18.5 billion in revenue (excluding Douyin) by 2024, accounting for 2.4% of the overall global digital advertising market ($780 billion).

Therefore, Dolphin Analyst does not currently consider the short-term easing of competition, which may lead to an increase in Meta's market share in the medium to long term social advertising market, and will not make special adjustments to the trend of advertising revenue growth. As for hardware revenue on the VR side, its contribution to overall revenue is relatively low, and rough estimates are only based on IDC's projected shipment volume and the current average price of Oculus.

Based on different expectations for the company's operations:

① For a neutral expectation, Dolphin Analyst believes that, under pressure from shareholders and the development of the VR industry, Meta will continue to cut costs, and we expect that operating expenses will return to the level before the epidemic/VR massive investment within three years, and the gross profit margin will be slightly lower than in previous years due to the relatively high cost of Reels.

Therefore, compared with market expectations, Dolphin Analyst has higher expectations for Meta's profitability recovery speed (especially in 2024), and we believe that, under shareholder constraints, Meta has the motivation to further reduce expenses, in addition to cutting 11,000 employees.

In the case of WACC=10.5% and g=2.5%, the Meta DCF valuation is $557 billion, or $215 per share. If the short-term risk-free rate is assumed to be raised to 3.5%, the corresponding market value is $520 billion, or $200 per share.

② Under pessimistic assumptions, the Meta management team will continue to make high investments in VR and maintain the human efficiency level after downsizing, accepting a long-term operating profit rate of 27% in 2023. Although this situation is extremely unfavorable for shareholders, the extreme valuation in this scenario also represents the potential bottom for Meta in the future.

We assume that due to continuous "unreasonable" investments, the long-term operating profit rate will still hover around 29%, far below the normal level of 40% before 2019. In this case, Meta's DCF valuation is less than 400 billion U.S. dollars, with a single share of 153 U.S. dollars.

Combined with the current market value (440 billion), this indicates that the market expects some further cost reduction and efficiency improvement actions, but not many. Therefore, this means that the expected difference may come from Meta's subsequent more aggressive operational adjustments.

Dolphin Analyst "Google" historical articles

Financial Report Season (only showcasing 2022)

February 3, 2023 conference call " More concerned about revenue growth than simply cutting costs (Google 4Q22 conference call summary)"

February 3, 2023 financial report review " Short-term pressure is not small, Google needs to learn from Meta"

October 26, 2022 conference call " Short-term optimization of resources, opportunities still lie in search and YouTube (Google 3Q22 conference call summary)"

October 26, 2022 financial report review " Google: Recession approaching, the king of advertising is down"

July 27, 2022 conference call " Google: High economic "uncertainty" in the second half of the year, focusing on investing in better long-term prospects (conference call summary)"

July 27, 2022 financial report review " Google: "Hard-handed rolls" under the expectation of a sudden drop"

April 27, 2022 conference call " Management avoids talking about TikTok, but the competition behind Shorts is still intensifying (Google conference call summary)" 2022-Apr-27 Financial Report Review "Google: Facing Headwinds, Big Brother Is Struggling Too"

2022-Feb-02 Teleconference Meeting "Increasing Investment, Accelerating Recruitment, Google Actively Seeking Expansion (Teleconference Summary)"

2022-Feb-02 Financial Report Review "Dazzling Performance, Rare Stock Split, Google Is Soaring Again"

In-Depth

2022-Jul-01 "TikTok To Teach "Big Brothers" How To Work, Google And Meta Are Going To Change"

2022-Feb-17 "Internet Advertising Overview-Google: Watching The Wind And Clouds Rise"

2021-Feb-22 "Dolphin Research | Detailed Analysis of Google: Is the Recovery Trend of Advertising Leader Over?"

2021-Nov-23 "Google: Performance and Stock Price Soaring, Strong Recovery is the Theme of This Year"

Dolphin Research "Meta" Section Historical Articles

Quarterly Reports (Showing Only 2022)

2023-Feb-02 Teleconference Meeting "Zuckerberg Is Talking About "Efficiency" All The Time, Has Learned To Behave (Meta 4Q22 Results Teleconference Transcript)"

2023-Feb-02 Financial Report Review "Good News Framed, Meta Turnaround?"

2022-Oct-27 Financial Report Review "Under Questioning 'Encirclement', Despite Severe Losses, Zuckerberg Still Bets On The Metaverse (Meta 3Q22 Teleconference Summary)"

2022-Oct-27 Financial Report Review "[Headstrong Meta, Still Betting On The Metaverse Under Bleeding](https://longbridgeapp.com/topics/3579820? invite-code=032064)》

2022年7月28日电话会《宏观、苹果 ATT、竞争多个逆风，管理层的短期展望很保守（Meta 电话会）》(https://longbridgeapp.com/topics/3210898)

2022年7月28日财报点评《没有 "Google 式" 的预期反转，Meta 颓态难掩》(https://longbridgeapp.com/topics/3209031)

2022年4月28日电话会《为应对竞争，不着急推进 Reels 商业化（Meta 电话会纪要）》(https://longbridgeapp.com/topics/2445846)

2022年4月28日财报点评《暴涨改信仰？Meta 拐点还未到》(https://longbridgeapp.com/topics/2442693)

2022年2月3日电话会《能否期待 Reels 像 3 年前的 Stories 再次激活 Meta 用户增长？（电话会纪要）》(https://longbridgeapp.com/topics/1891697)

2022年2月3日财报点评《雷上加雷，改名 Meta 后 Facebook 变身 "衰神"》(https://longbridgeapp.com/news/55050905)

** 深度 **

2022年7月1日《TikTok 要教 "大哥们" 做事，Google、Meta 要变天》(https://longbridgeapp.com/topics/3016272?invite-code=032064)

2022年2月17日《互联网广告综述——Meta：战斗力低下是原罪》(https://longbridgeapp.com/topics/1924950)

2021年9月24日《苹果拔刀，第一个 "见血" 的巨头是 Facebook？》(https://longbridgeapp.com/topics/1165524?invite-code=032064)

2021年8月6日《Facebook：深挖全球头号网民收割机的 "生意含金量"》(https://longbridgeapp.com/topics/1012966?invite-code=032064) 2021年11月23日《Facebook: Heavy Money Turns to "Meta", Turning Point is Not Far Away After Double Pressure》

Risk Disclosure and Statement of this Article: Disclaimer and General Disclosure of Dolphin Investment Research