Meta: AI investment in 'reckless driving', is the faith in capital going to be exhausted?

Hello everyone, I am Dolphin Research!

$Meta Platforms(META.US) released its third-quarter 2025 financial report after the market closed on October 30th, Eastern Time. Although the "thrilling" net profit of 2.7 billion is "noise," the performance, especially the underlying trends implied by the revenue and expenditure guidance combination, is definitely not something worth celebrating.

Specifically:

1. Strong acceleration in advertising: Advertising revenue grew by 26%, further accelerating growth. According to advertiser research, this was mainly driven by increased Reels duration and AI tools Advantage+ boosting ROI. In terms of actual volume and price, the third quarter continued the trend of the second quarter, with ad impressions accelerating to 14% growth and unit price increasing by 10%.

From the regional trend perspective, the growth in volume should mainly be due to further penetration of Reels on the user side (mainly reflected in Facebook); as for ad unit price, the US market did not see erosion from low-priced ads like Reels due to economic growth, while Europe saw further price growth driven by the strengthening of the euro.

Overall, the advertising landscape is flourishing.

2. "Flat" fourth-quarter guidance: Compared to the continuous acceleration trend of advertising growth in the second and third quarters, the company provided a total revenue guidance of 560-590 billion for the fourth quarter (vs. sell-side market expectations of 574 billion, buy-side expectations between 580-600 billion), corresponding to a growth range of 16% to 22%.

Since advertising revenue accounts for 98% of Meta's weight, this revenue growth guidance is equivalent to advertising growth, in other words, fourth-quarter growth is expected to slow down. Considering Meta always exceeds the highest end of guidance, even if fourth-quarter Meta revenue ultimately meets the highest buy-side expectations, reaching 60 billion, the year-on-year growth would still be 24%, indicating a slowing growth trend.

Under this revenue growth trend, if operating expenses and capital expenditures both rise, it will result in slowing revenue growth, accelerated investment, and squeezed cash flow leading to reduced shareholder returns, while the current valuation remains high at around 28X PE.

3. Opex rising: Progress in large models is not smooth, AI labs are hiring at high salaries, followed by large layoffs. Coupled with high growth in depreciation expenses, this raises concerns in the market about uncontrolled spending.

This concern is being confirmed in the third quarter: R&D increased by 35% year-on-year, and management expenses, due to legal expenses and high employee salaries, increased by 88% year-on-year, completely returning to a high growth range.

Fourth-quarter expenditure guidance: According to the company's updated full-year guidance of 1160-1180 billion dollars (bottom line raised by 20 billion dollars from last quarter), based on the midpoint guidance, the implied fourth-quarter cost + operating expenses will further increase by 38% year-on-year, not considering last year's 1.55 billion legal expenses write-back, a 30% year-on-year increase, significantly higher than the fourth-quarter revenue guidance range of 16%-22%.

Calculating from start to finish, the corresponding fourth-quarter operating profit margin may be squeezed to around 42%, compared to the peak period profit margin of 55%, a significant decline. In other words, starting from the fourth quarter, the market will truly feel the pressure Meta's "reckless" investment brings to the report.

Ultimately, with both revenue and expenditure increasing significantly, third-quarter operating profit was 20.5 billion dollars, an 18% year-on-year increase, although it exceeded sell-side expectations, buy-side had already raised revenue and profit expectations under high growth, the actual beat should not be significant.

By business segment, core business, App services operating profit margin was 49%, below sell-side expectations. On the contrary, in the less concerned Reality Labs, losses decreased, narrowing to 4.4 billion dollars.

Dolphin Research reminds here that due to the continued AI FOMO mentality next year, excessive AI investment. In this situation, Reality Labs business may become a "disaster area" for cost reduction, layoffs, and expenditure cuts, shifting focus to the AI field. This process involves the continuous decline of the main APP profit margin while Reality Labs losses continue to narrow.

4. Reckless AI investment, is the market panicking? The third-quarter performance itself, even the fourth-quarter guidance, the market's biggest concern in the third-quarter financial report is still the qualitative guidance for 2026. This bomb, Meta still mercilessly threw out:

In the second-quarter financial report, capital expenditure guidance had already been raised from 64-72 billion to 66-72 billion, the lower limit increased by 2 billion; this time, after the third-quarter actual capital expenditure reached 19.4 billion, exceeding market expectations by nearly 1 billion, directly raising last year's lower limit of 66 billion to 70 billion, an increase of 4 billion, equivalent to further raising fourth-quarter capital expenditure to around 21 billion.

Clearly, with Blackwell's stocking, Meta's capital expenditure gate has completely opened. Looking ahead to 2026, although Meta did not provide specific figures for total expenditure, the qualitative guidance is already very "aggressive," starting to lay the groundwork for investors for next quarter's quantitative guidance:

a. 2026 capital expenditure: The growth opportunities in 2026 are too tempting, although specific data for 2026 capital expenditure has not been determined, it is expected to "invest aggressively" through self-built and third-party co-built data centers, the absolute value of capital expenditure will be "notably larger."

b. 2026 operating expenses: The growth rate in 2026 will be "significantly faster" than 2025, mainly due to higher cloud expenses and amortization of infrastructure costs, secondly due to the substantial recruitment of AI and technical talent in 2025 leading to a significant increase in employee expenses.

5. Cash usage and shareholder returns: At the end of the third quarter, Meta had cash + short-term investments totaling 44 billion dollars, down 3 billion from the previous quarter, mainly due to increased PPE investment consuming cash flow.

This quarter's free cash flow was 10.6 billion, dividends were 1.2 billion, and buybacks were 3.1 billion. Among them, buybacks significantly decreased compared to last quarter's 10 billion. This year, shareholder returns totaled just over 30 billion by the third quarter, compared to 34 billion in the same period last year. This quarter, on average, shareholder return yield has dropped to 2%.

6. Performance indicators overview

Dolphin Research's view

Overall, it can be seen that the third-quarter performance itself is good, the main issue is the revenue and expenditure trends implied by the quantitative and qualitative guidance:

a. First, from the fourth-quarter revenue growth guidance, Meta is entering a high revenue growth period, the subsequent growth trend will reverse the previous two quarters' accelerated revenue growth trend, and will slow down, or optimistically estimate for 2026, it may at most stabilize in the 20-25% growth range.

b. But under this background, capital expenditure and operating expenses start qualitative guidance is basically "full throttle": not to mention, look at the qualitative wording - "aggressively, notably, significantly." These three words placed in front of investment and expenditure, any one thrown out, would make investors "frown," all three thrown out, the market will inevitably panic.

c. Increased investment, reduced cash flow: the main shareholder return - stock buyback suddenly said a lot, while the stock price is still rising, directly dragging shareholder return yield down to 2%.

d. According to the closing market value of 1.89 trillion, 2025 post-tax operating profit of less than 70 billion, Meta's current PE valuation is around 28X.

Putting these four pieces of information together, the contradiction between Meta's performance and valuation in 2026 is evident:

2026 will be a strictly high investment period, this high investment covers the AI FOMO mentality, further marginally expanded capital expenditure high investment in 2026 itself, high salary talent competition, and the capital high growth already formed in 2025, starting from the end of 2025, continuously producing significant performance erosion on the report.

In this situation, Meta being able to provide a "WOW" market-wide high revenue growth guidance is crucial. And precisely on this issue, Meta did not provide guidance for further marginal acceleration of revenue in the fourth quarter, instead, the growth rate marginally slowed to (16%-22%). In this situation, the market may optimistically only be able to expect 2026 growth rate around 20-25%.

The result of revenue, valuation resonance becomes: slowing revenue growth, high investment starting to erode operating expenses on the report, while capital expenditure will "move forward passionately," directly squeezing the ability space for shareholder returns.

Thus, 2026 will be a strictly mismatched revenue and expenditure period for Meta, profit growth is likely to be low double digits, or high single digits growth. This profit growth expectation matching the current high PE of 25-30X is clearly untimely. Therefore, the next valuation cut is likely to be a high probability event.

But here Dolphin Research wants to remind, don't be scared blindly. Dolphin Research has been emphasizing valuation cuts, not logic cuts. Although the third-quarter performance guidance indeed confirms market concerns: Meta is building extensively, recruiting talent, performance will be uncomfortable, but this is just the company's own investment-output cycle mismatch.

First, the industry is still stable: although there are tax disturbances, the fiscal and monetary easing in the US in 2026 gives Meta, which is heavily bound to macroeconomic Beta through advertising monetization, an invisible performance safety bottom line.

And most importantly, in terms of competitive landscape, Dolphin Research believes the market's concerns about TikTok's return and Sora's sudden emergence are somewhat overestimated, Meta's social entertainment empire remains stable:

a. Sora's sudden emergence, the narrative seems scary. But currently, user retention rate performance is quite average, we believe most users should treat Sora as a tool, generating videos on Sora and then posting them on original social platforms. Therefore, the initial stock price performance clearly overreacted, Dolphin Research's commentary reference "Who will lose a layer of skin when OpenAI wants to scrape oil?".

b. Old rival TikTok, although the transaction is about to be completed (there are no substantial obstacles), TikTok's data center migration to Oracle in the US may lead to some front-end functionality limitations, whether this will lead to user loss is uncertain, therefore before the migration work is truly settled, Dolphin Research expects the overall impact on Meta should still be controllable.

Therefore, without the reappearance of the pressure from 2022 on the most critical issue, Meta will not have a major crisis. Currently, due to profit margin pressure, 2026 performance growth is definitely going to slow down significantly (unless Meta AI performs heroically within a year), but this still makes the valuation appear to have little buffer.

But if we move past the short-term capacity mismatch in 2026, if OpenAI to C products still require a relatively long "breeding period." Then, after 2027, moving past the short-term investment-output mismatch, from a long-term perspective, Meta's strong social moat under AI support still has the ability to lead the industry.

Therefore, from a long-term value perspective, it might be worth estimating the present value corresponding to the 27-year expectation of 25x P/E (valuation center) or 15x EV/EBITDA, discounted back at an annual required return rate of 10%, estimated to correspond to a present value of 2 trillion.

Of course, this value target corresponds to the current fundamental market value level, which is completely chicken ribs. And the hitting zone for Meta should correspond more to the opportunity pit after this pullback, for example, due to OpenAI entering social apps, TikTok restarting, coupled with a comprehensive stimulus from a certain earnings season report bomb in 2026, falling out a relatively good long-term "opportunity pit" (such as 20X PE or below vs. today's post-drop about 26X), then looking for a suitable entry timing.

Below is a detailed interpretation

I. Advertising is still "accelerating"

Third-quarter Meta revenue was 51.2 billion, accelerating year-on-year growth to 26%, exceeding the upper limit of guidance, but since the market had already anticipated the third-quarter revenue growth (third-party tracking platform data shows core advertising indicators accelerated growth in ad impressions to 15% in the third quarter), it is not considered very unexpected.

And the high growth is mainly because Meta, under the support of products like Reels, continues to consolidate user quality dimension indicators such as duration and DAU stickiness, driving the advertising business, which accounts for 98% of revenue, to continue joining growth. And VR new products (Quest 3S and the new generation of Ray-Ban AR smart glasses) were launched at the end of September, so they did not drive the current period.

Q4 revenue guidance: Meta management expects 4Q25 total revenue in the range of 56-59 billion, corresponding to a year-on-year growth of 16%-22%, with exchange rates still driving 1 point of growth.

This guidance is basically consistent with the sell-side consensus expectation of 57.3 billion, but lower than the buy-side expectation of 58-60 billion. Moreover, from the guidance trend, last quarter's guidance was 17-24% growth, vs. this quarter's 16-22%, there is a risk of actual fourth-quarter revenue growth slowing down.

Specific business breakdown:

1. Advertising business: acceleration in impression growth, possibly due to further penetration of Reels

For the advertising business, Dolphin Research consistently prefers to break down the current period's volume and price growth trend changes to better understand the current macro environment, competition, and other issues.

1) Ad impressions: The success of Reels is still releasing dividends

Third-quarter ad impressions growth further accelerated to 14%.

On one hand, it comes from the user base still expanding, DAP (Meta Group APP family bucket DAU) year-on-year further accelerated growth to 7.6%; on the other hand, from the trend of recent quarters, it should still be in the accelerated penetration of Reel ad monetization. Dolphin Research estimates that the average ad impressions per user in the third quarter grew by 5% year-on-year, especially this growth was achieved under the continuous rise of ad unit price.

And from historical experience, during stable APP type structure periods, when Meta is in an ad unit price upward period (macro good, advertiser demand good), it will not deliberately raise ad fill rate to excessively monetize, reducing user experience. Only during ad unit price downward periods, with revenue pressure, will it raise ad fill rate to boost revenue.

But currently, it is a relatively rare period of both ad and single-user ad impressions rising. Therefore, Dolphin Research judges that it should mainly be due to Meta Reels short video consumption duration increasing, after which the naturally high fill rate of short video ads leads to a structural increase in the group average.

However, the ad loading rate of old platforms is already not low, so Meta will continue to promote the commercialization of WhatsApp and Threads to create new ad inventory.

DAP growth + single DAP ad impressions growth drove total impressions growth to 14% in the third quarter, performing quite well. And by region, the total impressions still clearly show Meta's control over the ad release gate as an ad platform:

In Asia, because ad unit price basically did not grow, ad impressions directly increased by 23% year-on-year; while in regions like Europe and the US where ad unit price grew by 10%+, ad impressions were in the high single-digit range (also due to the increase in Reels ad proportion).

Third-quarter comprehensive ad unit price accelerated by one percentage point year-on-year, to 10%. By region, it is very clear, Europe due to exchange rate tailwind, year-on-year growth rate quickly further increased to 19%, and US domestic ad unit price also further grew to 13%.

Dolphin Research believes that behind this volume and price growth, besides macro tailwind, essentially it is still related to TikTok being continuously used as a bargaining tool and unable to develop normally, Meta during this period using AI technology to enhance user stickiness and consolidate ecosystem barriers. Only under this basic background can Meta achieve true "volume and price flying together."

Also from this perspective, Dolphin Research believes Meta investing in AI to consolidate ecosystem barriers, enhance user stickiness and duration; using AI investment to improve user content recommendation and ad recommendation accuracy is completely reasonable, Meta has no long-term logic problem, the problem is just Meta pursuing the dream of long-term stars and sea, short-term investment-output mismatch, leading to profit pressure and valuation appearing relatively high.

2. VR: New product contribution in the next quarter

Third-quarter Reality Labs overall revenue was 470 million, this revenue scale, honestly, no matter how much it grows, it is basically a rounding item of advertising revenue, or a Marginal Error. Growth was 74% year-on-year, mainly benefiting from Quest 3 and Ray-Ban Meta smart glasses.

And this time the core information, in Dolphin Research's view, is the operating loss unexpectedly narrowed, with only a 4.4 billion loss. Dolphin Research has already indicated that with increased AI investment, it is necessary to consider the possibility of short-term cost reduction and efficiency improvement, and reduced investment in this long-term investment, slow-effect, large-loss business.

Product rhythm, at the end of the third quarter, Quest 3S was released and the new generation Ray-Ban smart glasses were released, driving fourth-quarter performance.

(1) Quest 3S is a new VR headset continuing the cost-effective style of Quest 3 (released last year), can switch between AR/VR modes with one click, supports interaction between physical and virtual environment scenes, but the starting price is only $299.99. Meanwhile, Quest 3 is reduced by $200 to $499.99, currently Quest 3 is still the best-selling product, with the same basic model, Quest 3's storage space is 4 times that of Quest 3S.

The overall idea of the low-price strategy is correct, this is more conducive to rapid hardware penetration. However, the VR industry currently lacks high-quality content, so sales are basically driven by holiday promotions, expected to help fourth-quarter performance.

(2) The new Ray-Ban Gen3 mainly adds an AR display screen, neural interaction wristband, and the chip is replaced with a custom NPU, which increases AI computing power by 4 times, facilitating more AI function upgrades. For example, the new glasses support real-time subtitle translation, video calls, AI visual recognition, and other AI functions, priced at $399/$799 (price is high, may affect sales). In 2025, mainly launched in the US, expanding to Europe and other countries next year.

II. AI FOMO is booming, investment is soaring

In the third quarter, the depreciation from heavy asset investment has not yet significantly backfired, only growing by 23% year-on-year, total cost growth was 25%, so the gross margin in the third quarter was still further improved, but according to the company's 2026 expenditure guidance and reason explanation, the situation in 2026 is estimated to reverse, the substantial increase in capital expenditure in 2025 will start to reflect on the income statement.

In the third quarter, the number of employees decreased by nearly 1,000 in the previous quarter, and increased by 2,500 in the third quarter, raising the total number of employees to 78,500, partly leading to high growth in R&D and management expenses in the third quarter.

Specifically, R&D expenses grew by 36%, management expenses due to legal expenses, personnel expenses, etc., increased by 88%, ultimately revenue growth was impressive, but profit was 20.5 billion, an 18% year-on-year increase, not dazzling, operating profit margin fell to 40%.

By business segment operating profit, an incremental information that can be seen is that high-margin business - main business APP family bringing advertising + sporadic other income, operating profit margin is actually significantly below expectations, fell by nearly 4 percentage points quarter-on-quarter, to 49%; instead, Reality Lab did not have the expected large loss.

Third-quarter capital expenditure reached 19.4 billion, after Blackwell's stocking, Meta Capex increased by 2.4 billion quarter-on-quarter, and according to guidance, next quarter's expenditure should already be raised to around 21 billion.

Also because the fourth quarter is already 21 billion, the company also said it will significantly increase capital expenditure in 2026, Dolphin Research roughly estimates a 100 billion capital expenditure in 2026. If the company's revenue can maintain at 20% in 2026, it means entering the AI cycle, Meta's capital expenditure intensity (Capex/revenue) from 2024-2026 will be a high slope curve of 24%, 36%, 42%, the tendency of a light asset company to become heavy asset is very obvious.

At this time, if there is no accelerated revenue growth trend to match, the probability of valuation cut will become certain.