Meta第三季资本支出将超运营费用拒售算力令千亿AI投资承压

浑水摸鱼清源

Meta因拒绝通过对外出租算力对冲基建成本，且Muse等AI产品推行激进免费补贴，导致每年逾千亿美元的AI支出缺乏云业务安全垫，自由现金流恐陷入多年低谷。

AI-assisted summary of 【The Motley Fool】Meta真的能从AI中赚到钱吗？ — original source. Check the original for full context.

Source published: 2026-10-22

Key ideas

Meta资本开支面临历史性拐点：预计在2026年第三季度，其资本支出（CapEx）将正式超过日常运营支出（OpEx），年度基础设施承诺规模高达1300亿至1400亿美元，短期自由现金流将显著承压。 (5:00)
消费级AI应用Muse采用极具侵略性的补贴策略：累计下载量达1100万次且预计Q3达1500万次，但每周向每位用户免费赠送100万tokens，且80美元/月的最高档提供120亿tokens，短期内缺乏产生正向现金流的能力。 (6:14)
核心广告印钞机依然稳固但面临自相蚕食风险：过去12个月广告收入达2220亿美元，AI算法优化拉动最新季度广告单价提升12%，但独立AI工具使用时长与未来自主Agent可能分流传统信息流展示与点击广告。 (1:36)
缺乏云基础设施安全网：与微软和Alphabet可通过向企业租赁GPU算力分摊AI基建成本不同，扎克伯格明确拒绝成为“新云服务商”，完全押注于通过自主应用与超级智能直接变现，执行风险显著提高。 (0:00)

Detailed notes

## Background Published on Bilibili by the channel 浑水摸鱼清源 on October 22, 2026, this video presents an analysis from The Motley Fool titled "Meta真的能从AI中赚到钱吗？" (Can Meta Actually Profit From AI?). According to the page description, the program examines whether Meta's massive AI capital expenditures can generate real returns, contrasting its cash-generative core advertising business with past metaverse losses, aggressive consumer subsidies for its Muse application, and tightening free cash flow.

## TL;DR The Motley Fool hosts evaluate Meta's massive pivot toward proprietary AI infrastructure, where computing power is hoarded for direct intelligence monetization rather than leased out as cloud services. While Meta's core advertising empire remains extraordinarily profitable and is enjoying short-term AI pricing gains, infrastructure CapEx is set to overtake operational expenditures by Q3 2026, threatening a multi-year trough in free cash flow. Simultaneously, Meta's consumer app Muse is driving rapid user adoption through heavily subsidized token pricing, creating a significant lag between user growth and actual revenue generation. Lacking the cloud rental safety net enjoyed by peers like Alphabet and Microsoft, Meta carries elevated execution risk under Mark Zuckerberg's control, though the hosts ultimately debate whether its unparalleled distribution and talent justify a long-term buy.

## Atmosphere & style The discussion has a sharp, analytical, and grounded tone typical of seasoned financial analysts, blending respect for Meta's advertising juggernaut with clear-eyed scrutiny of its ballooning balance sheet commitments. The dynamic between the hosts is candid and conversational, enriched by relatable personal anecdotes—such as an unsettling experience with Muse's agentic boundary creep—alongside rigorous financial breakdowns. The mood traces a path from admiration of Meta's historical cash engine to sober apprehension over an impending cash flow squeeze, finishing on a spirited, constructive divergence between cautious restraint and long-term bullish conviction.

## Bulletin - [05:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=300s) Meta资本开支面临历史性拐点：预计在2026年第三季度，其资本支出（CapEx）将正式超过日常运营支出（OpEx），年度基础设施承诺规模高达1300亿至1400亿美元，短期自由现金流将显著承压。 - [06:14](https://www.bilibili.com/video/BV14tnv6YEY9?t=374s) 消费级AI应用Muse采用极具侵略性的补贴策略：累计下载量达1100万次且预计Q3达1500万次，但每周向每位用户免费赠送100万tokens，且80美元/月的最高档提供120亿tokens，短期内缺乏产生正向现金流的能力。 - [01:36](https://www.bilibili.com/video/BV14tnv6YEY9?t=96s) 核心广告印钞机依然稳固但面临自相蚕食风险：过去12个月广告收入达2220亿美元，AI算法优化拉动最新季度广告单价提升12%，但独立AI工具使用时长与未来自主Agent可能分流传统信息流展示与点击广告。 - [00:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=0s) 缺乏云基础设施安全网：与微软和Alphabet可通过向企业租赁GPU算力分摊AI基建成本不同，扎克伯格明确拒绝成为“新云服务商”，完全押注于通过自主应用与超级智能直接变现，执行风险显著提高。

## Key takeaways - [00:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=0s) 扎克伯格坚决拒绝成为算力出租的“新云厂商”，坚信Meta的AI战略应聚焦于吸纳顶尖人才并构建自主算力，通过消费级智能应用直接变现。 - [01:36](https://www.bilibili.com/video/BV14tnv6YEY9?t=96s) 核心广告业务是Meta不可动摇的基本盘，过去12个月创造了2220亿美元广告收入和1060亿美元营业利润，相比2019年的280亿美元营业利润增长约4倍。 - [01:36](https://www.bilibili.com/video/BV14tnv6YEY9?t=96s) Reality Labs硬件业务过去一年录得190亿美元营业亏损，缺乏第二增长曲线的历史包袱加剧了市场对其数千亿美元AI投入回报率的担忧。 - [03:45](https://www.bilibili.com/video/BV14tnv6YEY9?t=225s) AI短期内对广告业务赋能显著，凭借算法改进和库存优化推动最新季度全球平均广告价格增长12%，美国本土市场增幅更为强劲。 - [03:45](https://www.bilibili.com/video/BV14tnv6YEY9?t=225s) 长期看AI对广告模式存在结构性威胁：用户在独立AI工具上的停留会分流社交媒体信息流时间，而自主Agent的发展更可能颠覆传统的展示与点击付费生态。 - [05:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=300s) 资本支出激增正在改变Meta的财务结构：尽管过去12个月产生1300亿美元经营现金流，但Q3资本支出将超过运营支出，使自由现金流承压。 - [06:14](https://www.bilibili.com/video/BV14tnv6YEY9?t=374s) 旗舰应用Muse虽收获1100万次下载，但每周赠送100万免费tokens，80美元月费档提供120亿tokens，定价远低于竞品，属于完全亏本获取用户阶段。 - [07:54](https://www.bilibili.com/video/BV14tnv6YEY9?t=474s) 当前AI转型酷似2012年Meta向移动端的跨越，但当前年化1300亿至1400亿美元的基建支出规模极为庞大，变现时间窗口远比过去更紧迫。 - [11:30](https://www.bilibili.com/video/BV14tnv6YEY9?t=690s) 市场容易将用户增长假象等同于财务成功，Meta在成熟市场社交用户ARPU超过100美元，AI工具想要达到同等变现水平可能需要5到15年。 - [13:09](https://www.bilibili.com/video/BV14tnv6YEY9?t=789s) 扎克伯格持有超级投票权使外部股东无法干预战略，历史上唯有股价暴跌才能迫使其纠偏，普通投资者承担了更高的执行风险。 - [15:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=900s) 微软和Alphabet拥有庞大的云服务作为财务安全网来吸收未充分利用的算力，而缺乏B端云租赁业务的Meta完全暴露于AI基建开销之下。 - [15:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=900s) 尽管即将经历自由现金流挤压的财务低谷，但基于Meta在AI人才与庞大分发网络上的绝对优势，分析师仍建议在多元化投资组合中逢低买入。

## Detailed walkthrough ### [00:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=0s) Questioning Meta's AI Pivot and the Neo-Cloud Dilemma

The discussion opens with the staggering financial commitment Mark Zuckerberg and Meta Platforms are directing toward artificial intelligence: hundreds of billions of dollars. The fundamental question hanging over this massive spending spree is whether it will ever genuinely pay off for shareholders. The hosts focus first on the underlying business model shift. Historically and currently, Meta generates the overwhelming lion's share of its revenue from digital advertising, yet that traditional mechanism does not currently appear to be the primary monetization thesis for their emerging AI portfolio, including their standalone application Muse.

When breaking down how Meta intends to convert raw computing investments into revenue, the hosts note that concrete public details remain sparse. Rather than renting out compute capacity like a traditional infrastructure provider or a "neo-cloud"—a pathway recently mirrored by companies like SpaceX with cloud infrastructure optimized for AI inference and training—Zuckerberg has made it clear that he has no interest in being a wholesale infrastructure landlord. Instead, Meta is channeling its resources into hoarding top AI talent and building captive in-house supercomputing infrastructure, operating under the philosophy that owning the compute directly is the prerequisite for generating, refining, and directly monetizing frontier machine intelligence.

As John explains regarding Zuckerberg's core philosophy: > “Zuckerberg does not believe that the long-term business model plan should focus on that rather it should focus on monetizing intelligence.”

```mermaid flowchart TD accTitle: Meta AI Strategy vs Traditional Cloud Model accDescr: Comparison between the Neo-Cloud rental business model and Meta's proprietary AI monetization thesis.

subgraph NeoCloud["Neo-Cloud Infrastructure Model"] NC1["Build AI Compute Clusters"] --> NC2["Rent GPU Capacity to Third Parties"] NC2 --> NC3["Collect Cloud Service Fees"] end

subgraph MetaStrategy["Meta's Internal Model"] M1["Build In-House Supercomputing"] --> M2["Attract Top AI Researchers"] M2 --> M3["Train Proprietary Frontier Models"] M3 --> M4["Directly Monetize Intelligence via Consumer Apps"] end ```

---

[01:36](https://www.bilibili.com/video/BV14tnv6YEY9?t=96s) The $222 Billion Ad Machine and the Shadow of Reality Labs

Pulling up historical financial charts dating back to 2012, the hosts underscore just how peerless Meta’s existing core business remains. In 2019, Meta generated $28 billion in operating income; over the past twelve months, that figure expanded roughly fourfold to $106 billion, supported by an astounding $222 billion in total advertising revenue. This engine works by mastering the monetization of human attention across social media at global scale, allowing Meta to run circles around competitors like Snap. However, this unmatched success highlights a persistent vulnerability: Meta has never successfully built a "second act" outside the core social graph Zuckerberg launched in 2004.

The clearest proof of this struggle is Reality Labs, the very division whose hardware and virtual reality ambitions prompted the company to rebrand from Facebook to Meta Platforms. Over the past year alone, Reality Labs burned through $19 billion in operating losses, with material revenues failing to materialize despite the release of smart glasses that the hosts consider mildly compelling. For a business operating at Meta’s scale, sinking billions into speculative hardware initiatives has delivered a thoroughly underwhelming return on investment, casting a long shadow of skepticism over their current, even larger AI spending cycle.

Critiquing the persistent lack of secondary growth drivers, one host observes: > “This business has just been a complete disaster, $19 billion dollars worth of operating losses over the past year Revenue has never really taken off despite the fact that they now have some glasses that are, you know, mildly compelling.”

- **Operating income surge:** Grew from $28 billion in 2019 to $106 billion over the trailing twelve months (~4x growth). - **Advertising revenue:** Reached $222 billion over the trailing twelve months. - **Reality Labs drag:** Accumulated $19 billion in operating losses over the past year without producing meaningful revenue scale.

---

[03:45](https://www.bilibili.com/video/BV14tnv6YEY9?t=225s) AI's Double-Edged Sword: Near-Term Pricing Boost vs. Long-Term Cannibalization

Examining artificial intelligence's direct impact on Meta's current operations reveals a distinct divergence between near-term efficiency and long-term structural risk. In the short run, AI integration has been an unambiguous net positive for the ad business. Sophisticated algorithmic targeting and inventory optimization enabled Meta to deliver a 12% increase in average ad pricing during the most recent quarter, with domestic pricing across the United States performing even higher than that benchmark.

In the longer run, however, AI poses a subtle threat of self-cannibalization. If user engagement shifts toward dedicated AI tools like Muse, that time is diverted directly away from Instagram and the core Facebook feed, where Meta actually monetizes impressions. Furthermore, should consumer behavior transition into an "agentic world"—where autonomous AI agents handle search, scheduling, e-commerce, and discovery behind the scenes—the foundational pay-per-click and feed-impression advertising paradigm could rapidly deteriorate.

Detailing the near-term gains before pointing out the structural threat, one host states: > “One of the things to consider here is that artificial intelligence so far has has actually been a net positive for that advertising business.”

- **Ad pricing momentum:** Rose 12% year-over-year in the latest quarter, driven by internal AI targeting optimizations. - **Geographic outperformance:** The United States market experienced ad pricing gains well ahead of the 12% global average. - **Cannibalization dynamic:** Unmonetized time spent in chat and standalone AI interfaces directly siphons user attention away from monetizable feed impressions.

---

[05:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=300s) The CapEx Inflection Point: When Capital Spending Overtakes OpEx

Shifting attention to corporate cash generation, the hosts point out that Meta has operated as a mature cash cow, producing an enormous $130 billion in operating cash flow over the trailing twelve months. The central dilemma of this mature phase is that Meta is no longer merely racing to onboard the entire planet's population; rather, it is focused on extracting maximum user engagement and higher monetization rates per unit of attention.

However, Meta’s capital allocation has reached an unprecedented tipping point. Driven by data center expansion, specialized silicon, and power infrastructure, capital expenditures are expanding at a breathtaking pace. In the third quarter of 2026, Meta’s CapEx is projected to surpass its total operating expenditures—a milestone that does not even factor in substantial off-balance-sheet commitments and lease obligations. This dramatic escalation raises serious questions about whether pouring hundreds of billions of dollars into data centers simply to make an already-optimized advertising engine 5% better can justify exhausting virtually all of the company’s incoming cash flow.

Emphasizing the historical significance of this spending pivot, one host notes: > “This quarter, the third quarter of 2026, they're going to pass their operating expenditures With? CapEx...”

```mermaid flowchart LR accTitle: Meta Capital Allocation Shift accDescr: Diagram showing how Meta's $130 billion operating cash flow is being absorbed by AI CapEx eclipsing operating expenditures.

CF["$130B Operating Cash Flow"] --> Div{"Capital Deployment"} Div --> OpEx["Operating Expenditures (OpEx)"] Div --> CapEx["Infrastructure CapEx (Surpassing OpEx in Q3 2026)"] CapEx --> Infra["Custom AI Silicon & Data Centers"] CapEx --> OffBS["Off-Balance-Sheet AI Commitments"] ```

---

[06:14](https://www.bilibili.com/video/BV14tnv6YEY9?t=374s) Muse's Explosive Adoption and the Subsidized Token Trap

Meta’s flagship consumer AI play, Muse, has achieved explosive top-line adoption immediately following its launch. The application recorded roughly 11 million cumulative downloads by the preceding week, with projections indicating it could comfortably eclipse 15 million downloads by the time third-quarter earnings are formally reported. While the hosts acknowledge using the tool productively in the background for routine workflows, they emphasize that Meta is essentially subsidizing its users through aggressive token giveaways.

Currently, Muse grants users 1 million free tokens per week, completely unmonetized. Even for power users willing to pay, Meta’s top-tier plan is priced at just $80 per month for an enormous 12 billion monthly tokens. In contrast, competing AI labs like OpenAI and Anthropic price high-tier compute plans at $100 or more while delivering considerably smaller token allocations. Meta is deliberately undercutting the market and hemorrhaging compute capacity to buy user adoption, leaving investors entirely in the dark regarding when—or how—these subsidized interactions will generate positive unit economics.

Highlighting the financial reality of this aggressive discounting, John remarks: > “To me, even that most expensive tier, it's not going to be Cash generative for the company.”

- **Cumulative user downloads:** 11 million downloads recorded, with projections targeting 15 million by the Q3 earnings release. - **Free tier allowance:** 1 million free weekly tokens distributed to each user without advertising or subscription paywalls. - **Premium pricing tier:** $80 per month for 12 billion monthly tokens, far exceeding compute allowances offered by competing $100+ third-party tiers.

---

[07:54](https://www.bilibili.com/video/BV14tnv6YEY9?t=474s) Parallels to Facebook 2012, Boundary Creep, and Free Cash Flow Pressures

The current inflection point closely mirrors Facebook's post-IPO era in 2012 and 2013, when user numbers exploded while the company burned capital trying to pivot to mobile, causing the stock to languish until mobile ad monetization was finally cracked. The difference today is the sheer scale of the cash drain: Meta is committing an estimated $130 billion to $140 billion annually to infrastructure, meaning monetization cannot afford a multi-year delay while free cash flow trends downward.

Retention and utility also remain murky beyond headline download figures. One host shares a personal anecdote about uninstalling Muse after it overstepped its intended role: after granting it access to manage calendars, the application autonomously monitored private transactions and issued unsolicited alerts regarding an online dinner order. Despite these early product friction points and declining free cash flow, the hosts conclude that Meta’s sheer distribution muscle makes it dangerous to bet against their long-term user aggregation.

Expressing confidence in Meta's ability to lock in an unmatched consumer footprint, John concludes: > “On the path that they're on, they are going to have the user base in AI that is going to make a lot of companies jealous.”

- **Historical parallel:** Reminiscent of 2012–2013, when Facebook delayed near-term profits to capture global mobile market share before building mobile ad infrastructure. - **Annualized spending run-rate:** Projected infrastructure commitments of $130 billion to $140 billion per year make fast monetization significantly more urgent than in previous cycles. - **User experience friction:** Real-world reports of agentic boundary creep, such as AI services over-indexing on calendar and transaction tracking without explicit user requests.

[10:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=600s) The Promise of Muse and Meta's AI Talent

The rapid pace of artificial intelligence development makes it incredibly difficult to predict how these technologies will ultimately mature and what kinds of business models will emerge around them. Because the landscape is shifting so quickly, there is still a vast amount of unknown territory. However, a massive, highly engaged user base remains a critical competitive advantage, which makes it unwise to count Meta out of the AI race just yet.

A major reason for optimism is the exceptional AI talent Meta has managed to attract. The company has built a product that is far more compelling than many skeptics anticipated, largely by rethinking their approach from the ground up. This success spans both the underlying foundational models and the consumer-facing product side with their Muse application. Interestingly, Meta is designing these systems in ways that do not merely collect data, but actively help monetize their core business more effectively and lay a solid foundation for future advertising integration.

> “So I'm unwilling to count meta out of this game right now, especially considering some of the AI talent that it's attracted.”

---

[11:30](https://www.bilibili.com/video/BV14tnv6YEY9?t=690s) The Illusion of User Growth vs. Financial Reality

From an investment perspective, there is a dangerous tendency among market participants to equate rapid user adoption with immediate financial success. While Muse is experiencing impressive user growth, a massive time lag exists between getting people to use a free or heavily subsidized tool and actually generating meaningful revenue from them. To put this in perspective, Meta currently generates over $100 in revenue per user in highly developed markets like the United States through its core social media platforms. Reaching that level of monetization with an AI tool like Muse—whether through subscriptions, advertising, or behind-the-scenes transaction fees—could easily take anywhere from five to fifteen years.

Consequently, Meta is highly likely to enter a financial "trough" or valley over the next few years. Even if the core advertising business remains healthy, growing at a respectable 10% to 20% year-over-year with improving operating margins, the company's free cash flow is poised to take a serious hit. Huge capital expenditures and off-balance-sheet obligations will periodically impact the numbers, leading to intense public questioning of Meta's capital allocation, much like the skepticism that surrounded their Reality Labs metaverse investments. While public sentiment around Meta is incredibly strong today due to the cultural success of Muse, that popularity has not yet translated into a viable bottom-line contributor.

> “The concern that I have from an investment standpoint is I think that investors are currently equating user growth and user adoption with a product like Muse and financial success.”

- **Core Market ARPU:** Meta currently generates over $100 per user in developed markets like the U.S. - **Monetization Horizon:** Achieving comparable average revenue per user (ARPU) on AI products like Muse is estimated to be 4, 5, 10, or even 15 years away. - **Core Business Growth:** The core business may continue to grow at 10% to 20% year-over-year even as free cash flow feels the pinch of AI infrastructure spending.

---

[13:09](https://www.bilibili.com/video/BV14tnv6YEY9?t=789s) High Execution Risks and Zuckerberg's Super-Voting Power

While Meta has historically navigated massive transitions successfully—such as the shift to mobile or pulling back on Reality Labs to return capital to shareholders via dividends—the current AI transition carries an order of magnitude more execution risk. Ten years ago, capital expenditures were a fraction of what they are today. The sheer scale of the financial commitment required to build out AI infrastructure means that if this cycle does not pay off, the financial pain for shareholders will be far more severe than any past misstep.

Compounding this risk is Meta's corporate governance structure. Because Mark Zuckerberg holds super-voting shares, public investors have absolutely no power to vote him out or force a change in strategic direction. Historically, the only mechanism that has successfully forced Meta's management to course-correct is a severely declining stock price. If these massive AI investments fail to yield returns, Zuckerberg and the highly compensated AI specialists he recruited will remain incredibly wealthy, while ordinary retail shareholders will bear the brunt of the losses. For this reason, some investors are choosing to avoid buying the stock at its current valuation, anticipating a painful financial valley ahead.

> “At the same time, I will say that the execution risk is much higher here than it's been in the past.”

```mermaid flowchart TD accTitle: AI Monetization Paths: Meta vs. Cloud Giants accDescr: A flowchart comparing Meta's direct consumer AI monetization path with the cloud rental safety net of Microsoft and Alphabet.

subgraph Meta ["Meta's High-Risk Path"] A["Massive AI CapEx"] --> B["Consumer App (Muse)"] B -->|"No Cloud Rental Business"| C["Direct Monetization (Ads/Subs)"] C -->|"Long Lag (4-15 Years)"| D["High Execution Risk / Negative FCF"] end

subgraph CloudGiants ["Microsoft & Alphabet's Path"] E["Massive AI CapEx"] --> F["Cloud Infrastructure"] F -->|"Rent Out Compute to Others"| G["Immediate B2B Revenue"] G -->|"Financial Safety Net"| H["Lower Short-Term Risk"] end ```

---

[15:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=900s) The Lack of a Cloud Safety Net and the Final Verdict

Unlike its mega-cap peers, Meta lacks a diversified business model to cushion its massive infrastructure spend. Alphabet and Microsoft can comfortably spend billions on data centers because they run massive cloud computing divisions. If their consumer-facing AI applications fail to monetize immediately, they can still generate billions of dollars by renting out their raw computing power and enterprise AI tools to other businesses. While focusing heavily on renting out infrastructure may have temporarily slowed Alphabet's consumer product innovation, it provides a powerful financial safety net that Meta simply does not have.

Without a cloud business to fall back on, Meta's financial statements are highly vulnerable to rough quarters ahead, particularly regarding free cash flow and net income. Despite these looming headwinds, the co-hosts end on a debating note. While one remains highly cautious due to the impending capital expenditure drag, the other, John, takes a more optimistic stance on the stock's long-term prospects.

> “If you're making me choose between buy and sell, I'm choosing buy.”

John argues that Meta's long track record of pulling off positive surprises and the sheer scale of the AI opportunity make it a bet worth taking, provided it is held within a well-diversified portfolio. The episode concludes with a call to the audience to share their thoughts on Meta's massive AI spending in the comments section and to subscribe to the channel.

## Notable quotes > [11:30](https://www.bilibili.com/video/BV14tnv6YEY9?t=690s) The concern that I have from an investment standpoint is I think that investors are currently equating user growth and user adoption with a product like Muse and financial success. — 主持人

> [10:00](https://www.bilibili.com/video/BV14tnv6YEY9?t=600s) I'm unwilling to count meta out of this game right now, especially considering some of the AI talent that it's attracted. — 主持人

## Mentioned Meta Platforms, 马克·扎克伯格, Muse, OpenAI, Anthropic, Reality Labs, Snap, Alphabet, 谷歌, 微软, SpaceX, Facebook, Instagram, John

## Video description 【The Motley Fool】 Meta真的能从AI中赚到钱吗？ Can Meta Actually Profit From AI？

大家好，今天我们来深度探讨Meta在人工智能领域的巨额资本投入究竟能否带来实际回报。作为一家长期依赖广告变现的互联网巨头，Meta在AI浪潮中正面临着前所未有的机遇与财务压力。

广告帝国的高额盈利与历史支撑回溯Meta的历史财务数据，其核心广告业务展现了惊人的印钞能力。从2019年的28亿美元营业利润，一路飙升约4倍，在过去12个月内创下了高达106亿美元的营业利润，总广告收入更是达到了222亿美元。这种利用社交网络规模垄断用户注意力的商业模式极其成功，远超同行竞争对手。此外，人工智能在短期内确实对广告业务起到了正面赋能作用，最新季度的广告定价实现了12%的显著增长，尤其在美国本土市场表现更为优异。

元宇宙业务的亏损与AI转型的抉择相比之下，扎克伯格近年来尝试的第二增长曲线并不顺利。以Reality Labs为例，过去一年该业务录得了高达19亿美元的营业亏损，且硬件产品未能实现大规模商业腾飞。这也让市场对当前Meta在AI领域的数百亿美元资本投入产生了深深的忧虑。与微软或谷歌不同，扎克伯格明确表示不打算将算力对外出租以成为“新云服务商”，而是选择构建自主算力以发展“超级智能”，通过应用和智能服务直接变现。

Muse应用的爆发与变现迷雾作为Meta在AI消费端的重要布局，Muse应用自推出以来展现了强劲的用户增长势头。其累积下载量近期已达1100万次，预计在第三季度财报公布时将突破1500万次。然而，从变现角度来看，Muse目前采取了极具攻击性的补贴策略，免费提供大量周tokens，而最高档的订阅服务定价仅为每月80美元（提供120亿tokens），远低于OpenAI或Anthropic的百元级定价。这表明Meta目前的首要目标是极力争夺用户采用率，而将其真正的商业化变现远远抛在脑后。

自由现金流的挤压与长期财务低谷从现金流角度来看，Meta过去一年虽然迎来了高达130亿美元的经营现金流，但其资本支出在2026年第三季度已经开始超越日常运营支出。考虑到每年高达1300亿至1400亿美元的庞大基础设施建设投入，加上未在资产负债表上完全体现的潜在义务，公司的自由现金流正面临明显的短期下行压力。如果AI应用的商业化进程迟缓，Meta很可能将经历一段财务表现较为痛苦的低谷期。

当前宏观环境下的三条可执行操作建议理性评估AI资本开支的周期风险：认识到Meta短期内自由现金流可能因大规模AI基建而受压，避免盲目追高带来的短期震荡风险。密切跟踪Muse及核心AI应用变现进展：关注用户下载量向实际付费率和长期广告转化率的落地效率，警惕用户繁荣与财务回报脱节的现象。在多元化资产组合中逢低布局：基于扎克伯格强大的战略执行力和Meta在AI顶尖人才上的吸引力，若股价因短期自由现金流承压出现调整，可在多元化投资组合中分批择机吸纳。

简洁时间轴 [00:00] 探讨Meta巨额AI投资回报率与核心商业模式 [01:36] 回顾Meta广告业务历史营收与AI定价赋能 [02:49] 分析Reality Labs亏损及Meta缺乏第二增长曲线的隐忧 [05:24] 剖析1300亿现金流、资本开支超越运营支出的财务压力 [06:14] 解读AI应用Muse下载量爆发与极其激进的低价策略 [09:09] 讨论AI时代长期变现模式与自由现金流下行周期 [13:09] 评估扎克伯格超级投票权与高额资本开支的执行风险 [15:37] 主讲人总结Meta股票长期投资价值与最终观点

[Source](https://www.bilibili.com/video/BV14tnv6YEY9)

中文简报

Meta资本支出将超运营费用拒售算力令千亿AI投资承压 Meta因拒绝通过对外出租算力对冲基建成本，且Muse等AI产品推行激进免费补贴，导致每年逾千亿美元的AI支出缺乏云业务安全垫，自由现金流恐陷入多年低谷。 The Motley Fool节目主持人探讨了Meta向专有AI基础设施的大规模转型——其囤积算力旨在实现自主智能变现，而非对外作为云服务出租。尽管Meta的核心广告帝国依然极具盈利能力，且正享受短期AI带来的定价红利，但基础设施资本支出预计将在2026年第三季度超过运营费用，可能导致自由现金流陷入多年低谷。与此同时，Meta消费级应用Muse正通过高额补贴Token定价推动用户激增，造成用户增长与实际收入变现之间的显著脱节。由于缺乏Alphabet和微软等同业拥有的云租赁安全垫，Meta在扎克伯格主导下面临更高的执行风险，但主持人们最终也探讨了其无与伦比的分发网络与顶尖人才是否足以支撑长期买入。 Meta资本开支面临历史性拐点：预计在2026年第三季度，其资本支出（CapEx）将正式超过日常运营支出（OpEx），年度基础设施承诺规模高达1300亿至1400亿美元，短期自由现金流将显著承压。 (5:00) 消费级AI应用Muse采用极具侵略性的补贴策略：累计下载量达1100万次且预计Q3达1500万次，但每周向每位用户免费赠送100万tokens，且80美元/月的最高档提供120亿tokens，短期内缺乏产生正向现金流的能力。 (6:14) 核心广告印钞机依然稳固但面临自相蚕食风险：过去12个月广告收入达2220亿美元，AI算法优化拉动最新季度广告单价提升12%，但独立AI工具使用时长与未来自主Agent可能分流传统信息流展示与点击广告。 (1:36) 缺乏云基础设施安全网：与微软和Alphabet可通过向企业租赁GPU算力分摊AI基建成本不同，扎克伯格明确拒绝成为“新云服务商”，完全押注于通过自主应用与超级智能直接变现，执行风险显著提高。 (0:00)

Reading summary

The Motley Fool的主持人探讨了Meta在人工智能领域的巨额资本支出能否为股东带来实质回报：尽管其核心广告业务强劲且独立AI应用Muse用户增长迅猛，但缺乏云计算出租业务的缓冲、激进的算力补贴以及资本支出超越运营支出的拐点，正让Meta面临显著的自由现金流压力与执行风险。核心广告引擎的强劲表现与第二曲线的缺失 (1:36) Meta的核心广告业务依然是庞大的现金奶牛。在过去十二个月中，其广告收入达到2220亿美元，营业利润从2019年的280亿美元跃升约4倍至1060亿美元，并在过去十二个月产生1300亿美元的经营现金流。短期内，AI算法对广告定位和库存优化的赋能带来了积极效果，最新季度的平均广告定价同比上涨12%，美国市场涨幅更高。然而，这凸显了Meta长期以来未能成功打造出第二增长曲线的软肋。作为转型标志的Reality Labs硬件与虚拟现实部门在过去一年中录得19 billion美元的营业亏损，且未能形成实质性收入规模。这笔巨额亏损为当前规模更加庞大的AI资本支出蒙上了投资回报率不佳的阴影。直面消费级AI的变现困境与自我蚕食风险 (0:00) 不同于将算力出租给第三方的“新云”或基础设施模式，扎克伯格的核心理念是拒绝充当批发云房东，坚持将资源投入自建超算基础设施与招揽顶尖AI人才，以图通过消费级应用直接将“智能”变现。旗舰AI应用Muse已取得显著的用户采用率，下载量录得约1100万次，并预计在第三季度财报发布前突破1500万次。但主持人指出，这种繁荣是建立在不可持续的算力补贴之上的：Muse向每位用户提供每周100万个完全免费且未变现的Token；其最高付费档位仅为每月80美元对应120亿个Token，远低于OpenAI和Anthropic等竞品100美元以上却提供更少Token的定价标准。此外，如果用户在Muse等未变现界面中投入过多时间，或者未来自主AI代理承担了搜索与交易功能，传统的信息流展示与按点击付费广告体系可能面临被直接蚕食的结构性风险。资本支出反超拐点与长周期变现滞后 (5:00) Meta的资本配置正逼近前所未有的临界点。受数据中心扩建、定制芯片和电力基础设施推动，预计到2026年第三季度，公司的资本支出（CapEx）将超过其运营支出（OpEx），这还不包括资产负债表外的重大承诺与租赁义务。目前估算Meta每年对基础设施的承诺支出高达1300亿至1400亿美元。市场容易将Muse等产品的早期用户增长误判为财务上的成功。Meta在欧美等成熟市场的社交平台每用户平均收入（ARPU）已超过100美元，但要让AI产品达到同等的变现水平，可能需要耗费4年、5年甚至长达10到15年的漫长周期。即使核心广告业务仍能保持10%至20%的年增长，不断扩大的资本开支仍将导致公司未来数年进入自由现金流显著受挫的财务“低谷期”。缺乏云安全网与单一治理权下的高执行风险 (13:09) 与拥有庞大云基础设施业务的微软和Alphabet相比，Meta缺乏多元化的安全网缓冲。如果Alphabet和微软的消费者AI应用变现不及预期，它们仍能通过向企业出租底层算力和AI工具赚取稳定收入；而缺乏B2B云租赁业务的Meta，其财报在自由现金流和净利润方面对资本投入周期的冲击极度脆弱。这一周期的执行风险显著高于十年前向移动端转型或削减元宇宙支出的阶段。由于扎克伯格持有超级表决权股份，公开市场股东无法通过投票迫使管理层变更战略路线，历史上唯有股价暴跌曾促使其纠偏。一旦巨额AI投资未能转化为利润，普通散户投资者将承担大部分下行风险。围绕Meta股票的前景，节目主持人群体产生了观点分歧：谨慎者因预见到自由现金流将因资本开支而面临剧烈承压而选择观望；主持人John则认为，鉴于Meta强大的用户分发能力、顶尖AI人才储备以及历史上多次化解转型危机的往绩，在保持投资组合多元化的前提下，长期持有仍是值得押注的选择。

精炼阅读摘要

The Motley Fool节目指出，Meta正将数千亿美元投入自建AI超级计算基础设施与人才争夺，但由于缺乏类似微软与Alphabet的云计算转租业务作为财务安全垫，其激进的算力补贴与资本支出恐令自由现金流陷入多年低谷，商业化前景仍面临严峻考验。广告现金牛与第二曲线缺失 (1:36) 节目指出，Meta目前依然高度依赖庞大的数字广告机器。过去十二个月，Meta总广告收入达到2220亿美元，营业利润从2019年的280亿美元暴增至1060亿美元，运营现金流高达1300亿美元。在短期内，AI算法推荐与库存优化进一步推高了广告效率，使其最近一个季度的全球平均广告单价上涨12%，美国本土增幅更为显著。然而，这一无与伦比的盈利能力暴露出Meta难以开拓“第二增长曲线”的长期隐患。最典型的负面案例是Reality Labs，过去一年该部门录得19 billion（190亿）美元的营业亏损，硬件与虚拟现实尝试始终未能形成实质性收入规模。这引发了投资者对Meta当前规模更加庞大的AI资本周期的深层疑虑。自研路线与云出租安全垫缺失 (0:00) 与微软、Alphabet等科技巨头不同，Mark Zuckerberg明确表示无意成为批发算力的云基础设施“房东”（Neo-Cloud模式）。Alphabet与微软即使在消费级AI变现上遭遇延迟，依然能通过向企业租售GPU算力与云服务快速回笼资金；而Meta则选择囤积顶尖AI人才、全资自建内部超级计算基础设施，寄望于未来通过消费级智能应用直接变现。分析指出，缺乏B2B云计算转租这一安全缓冲，意味着Meta的数百亿美元数据中心与定制芯片支出必须完全由自身产品吸收。若消费级AI变现不及预期，这些固定资产与表外租赁承诺将直接转化为利润表与自由现金流的剧烈下挫。资本开支拐点与代币高额补贴 (5:00) Meta的资本配置正逼近前所未有的临界点。根据节目预测，到2026年第三季度，Meta的资本支出（CapEx）将超过其总运营支出（OpEx），年度基础设施承诺规模预计达1300亿至1400亿美元。然而，数千亿美元的投入若仅能让本已高度成熟的广告引擎实现5%左右的边际改善，其资本回报率将难以服众。在消费端，Meta推出的旗舰AI独立应用Muse虽然上线后累计下载量迅速突破1100万并预计在三季报前冲刺1500万，但这一增长在很大程度上依赖激进的算力补贴。Muse不仅向用户每周免费赠送100万个代币，其最高付费档也仅定价每月80美元换取120亿个代币，远低于OpenAI或Anthropic等100美元以上且额度较小的行业定价。分析认为，即便是最高付费层级目前也无法产生正向现金流。变现时间差与内部蚕食风险 (11:30) 节目警告，市场常将应用下载量误等同于财务成功。Meta在成熟市场（如美国）的社交平台每用户平均收入（ARPU）已超过100美元，而通过Muse等AI工具达到同等变现水准（无论通过广告、订阅还是交易抽成）预计需要4到15年的漫长周期。此外，用户在Muse等未商业化界面停留的时间，会直接分流其在Instagram和Facebook信息流上的浏览时长，从而蚕食现有的展示广告与点击付费收入。治理结构进一步放大了执行风险。由于Zuckerberg拥有绝对控股的超级投票权，外部股东无法通过表决迫使其改变战略方向，历史上唯有股价剧烈下跌才能迫使其纠偏。这一局面意味着散户投资者可能需要独自承受未来几年自由现金流下滑带来的估值震荡与财务低谷。主持人在最终评判上存在分歧：谨慎的一方认为在资本开支失控与缺乏云业务对冲下当前估值存在下行风险；而另一位主播John则给出“买入”倾向，认为凭借其庞大的分发优势与顶尖研发团队，Meta仍有望重现2012年向移动端转型时的长期逆袭，但投资者需在分散投资的前提下承受中短期的财务波动。

Source quotes

[11:30] 主持人: The concern that I have from an investment standpoint is I think that investors are currently equating user growth and user adoption with a product like Muse and financial success. [10:00] 主持人: I'm unwilling to count meta out of this game right now, especially considering some of the AI talent that it's attracted.

Original transcript

[0:00] Mark Zuckerberg and Meta platforms are investing hundreds of billions of dollars in artificial intelligence, but is that going to pay off? That's the question that we want to ask ourselves today. John, when we think about Meta's investment in AI. Part of it is Muse and all the new products that they're gonna make, but it's a significant amount of money. What I want to start with is the business model. What do we know today about the business model shift? Because I think the thing that's difficult right now is, they make most of their money on advertising. As of today, that does not appear to be the plan in artificial intelligence. So how are they going to make money at least for what we know right now and what are some of those options in the future. We don't know a ton, to be clear, Perfectly Frank meta is investing a lot in attracting top AI talent and And trying to do as much of what it can in-house is, how it's actually monetized we don't have a ton of details what we do know is that Zuckerberg does not necessarily want to rent out the compute. So you think of a Neo-Cloud. This is becoming a Neo-Cloud. This is essentially what SpaceX has recently done, just to kind of put that into context. Exactly. The NeoCloud business, this is... Infrastructure in the cloud that is optimized for AI. inference or training whatever. Zuckerberg does not believe that the long-term business model plan should focus on that rather it should focus on monetizing intelligence. And so it's basically building out its own compute. to develop its own intelligence and then hopefully monetize that. Thank you. Yeah, and let's... Let's start with what the current business model looks like. I'll pull up a chart here. What we have here going back all the way to 2012 is is... Meta's advertising revenue, so this is today, an advertising business and it is extremely profitable. So you can see going from $28 billion in operating income in 2019, That's up about 4x. to 106 billion over the past 12 months now 222 billion dollars worth of advertising revenue so That's where we're going from. The challenge is it's not a direct line from that business model, which is a phenomenal business model, right? You're monetizing attention with advertising. This is what TV has done forever, what radio has done forever. And now Meta has mastered the art of doing that with social media. And once you have that scale, you're able to do it extremely profitably, you're able to do it better than competitors. This is why you can kind of run circles around companies like Snap, for example. But if we look at what they've developed over the past decade, it has not been nearly as successful. I think this would be the critique if you're looking at Mark Zuckerberg is, He has a great business model with the business that was started in 2004. But there's been no kind of second act since then. And the first thing to look at is Reality Labs. The reason the company's name is now Meta Platforms. This business has just been a complete disaster, $19 billion dollars. worth of operating losses over the past year Revenue has never really taken off despite the fact that they now have some glasses that are, you know, mildly compelling. But for the scale of a business like this. It's just not a... really good return on investment and that is fundamentally the question here is are they going to be able to build a product that's a return that's a good return on investment so i think those The contrasting those two businesses is something investors should do before we think about how they're going to make money on Muse. One of the things to consider here is that artificial intelligence so far has has actually been a net positive for that advertising business. You can better optimize and target with your existing advertising inventory with artificial intelligence. And so we actually do see a 12% increase in the most recent quarter for ad pricing. You consider that it's already one of the best in the entire world. And then when you look at it's in the US predominantly, that we're seeing a significant, even better than that 12%, a significant uptick in the pricing. So artificial intelligence is helping that now. Longer term, it's a little bit more of a risk because one If you're interacting with, let's say, Muse, let's say that Muse keeps growing. If you're interacting in that platform with a meta product, it's not actually monetizing that in advertising, at least not right now. It's actually taking attention away. from the Facebook platform or the Instagram platform. So that's actually kind of not good. And then you think about it as well. If the agentic world takes off, then that really starts to cut into the whole advertising thesis. So right now it's a. net positive longer term it could be a bigger risk that's one thing to consider uh, The other thing that we have to think about too is the cash flow. I mean, we're going from a phase of Meta's business that is actually pretty [5:00] mature business you can say this for a lot of the tech companies out there alphabet microsoft you know, we're long past the days of, hey, we're just trying to get to scale, we're trying to reach everybody in the world. We're now to the days of how can we extract more and more and more from the users that we have more attention, more dollars per unit of that attention. So you've got a ton of cash flow coming into the business. That's the good news. You can see here $130 billion in operating cash flow coming in. to meta platforms over the past year. What we have seen, and this is not a meta- It's not unique to Meta, but we have seen them now start to spend more. On... this ai build out capex This quarter, the third quarter of 2026, they're going to pass... their operating expenditures With? CapEx... And that doesn't even include the things that are off balance sheet right now. So there is a significant amount of capital that's going into this. with a lot of questions about what that business model is going forward. Yes, it's going to make the existing business a little bit better. But is making that business, you know, 5% better going to be worth burning all of your cash flow in the meantime? Yeah, that's a fantastic point. You look at Mew's wild runaway success story right out of the gate. probably 11 million cumulative downloads as of Last week, by the time it reports third quarter earnings, it could be $15 million already having downloaded Muse. And Muse is actually doing some work for me right now in the background, so I find it useful. The thing is, a million free weekly tokens, it's not monetizing me right now at all. I could pay for the most expensive tier that it has at $80 a month and get $12 billion monthly Muse tokens. $100 tiers. even then aren't giving you this kind of compute. Meta is giving away a ton of compute. To me, even that most expensive tier, it's... not going to be Cash generative for the company. That is still very much a game where it is trying to incentivize adoption and monetize later. And to your point, what is the business model on the other end of this? I don't think that Meta has really shown us that yet. There are little things it said here and there. Actually seeing something concretely and really drawing that line, if you will, to this is where it pays off. I can't personally do it. And that's where this moment in artificial intelligence, if you think of Muse products, it reminds me a lot more of Meta or Facebook as it was known at that point in 2012, in 2013, you could see there's a ton of users. I mean, they were at that point where they're, but they're burning billions of dollars and they're, they're, they're burning many more times that at this point. But Their goal was to attract as many users as possible and then we'll figure out monetization later. The reason that the stock kind of languished in that point in time was early when it was public was it wasn't clear that they were actually going to be able to do what they ended up doing, which is monetize via advertising in the effective way that it did. So the question for investors today from a is this going to pay off standpoint is not. Will people use Muse? I mean, I think that's fundamentally a question. I actually did start using Muse and then I stopped using it and took out, took away all my connections because it started doing things that I didn't ask it to do. started alerting me that I had made a online order for dinner and I was like I asked you to help with the calendar. I didn't I didn't ask you to help with this. We've heard other reports of people having similar things where it's kind of over overstepping the bounds that it had. So that was just a little bit too far for me as a user. So, you know, we don't know what those user metrics look like. That's going to be the other thing is not just how many downloads do you have, but how many actual users are interacting with it on a day to day basis. But if you're in that 2012, 2013 period, you got to figure out monetization pretty quickly, especially when the cost is. 130, 140 billion dollars per year. And Mark Zuckerberg, like you said, said that he doesn't want to rent out this compute. So the question for today is, are they going to be able to do that? Because if you look at just Facebook's, Meta's, free cash flow you go to free cash flow is going in the wrong direction to be a positive return on investment we've got to see that free cash flow turn around and this become a free cash flow machine like it was over the past decade. Do you think they're going to be able to do that? It's hard for me to say that they're not going to because, again – They're going to get the adoption. I feel very strongly about that right now. On the path that they're on, they are going to have the user base in AI that is going to make a lot of companies jealous. And so that is just one lucky break away from actually turning this into a monetizable and strong business. [10:00] I think that AI is still developing so much that it's hard to even know what you don't know. and how these things actually take fruition and actually start to develop business models behind them. I don't think that we've seen everything yet. And so it's really hard to imagine. And like I said, the user base does matter. So I'm unwilling to count meta out of this game right now, especially considering some of the AI talent that it's attracted. Yeah, that is the other thing. They have... I think created a product that was far more compelling than I thought that they probably could have created. And they partly did it by kind of rethinking things from the ground up. Some of the people that they hired have just done a phenomenal job of, you know, coming in and saying, hey, here's how we need to do this and you know. That's both on the model side and then also on the product side with Muse. They're doing some things that are not collecting data that could actually help monetize the rest of their business a little bit better and lay the groundwork for advertising. The concern that I have from an investment standpoint is I think that investors are currently equating investors. user growth and user adoption with a product like Muse And financial success. And I think there's going to be a long lag between those two things. Even if Muse does figure out a business model, does get, you know, let's say a billion users, we're a long way from a point where they're going to be able to monetize it anywhere near the level that they currently have. I mean, you're talking about, I think, over $100 per user is the revenue that Meta is generating in developed markets like the U.S., That's a lot of revenue to be generating from a product like Muse, whether you're doing it through subscriptions or whether you're doing it through advertising kind of, you know, or transaction volume, things that are kind of behind the scenes. we're a long ways, four, five, 10, 15 years away from that happening. So I think it's very possible that we go through a trough where the numbers actually don't look that good. Where the core business is maybe doing fine, maybe it's growing at 10% to 20% year over year, maybe those operating margins are getting better. But your cash flow and Alphabet's in the same boat here. So it's a lot of different companies that are kind of looking at similar things. similar trajectories of their business, but short term that free cash flow is going to go negative. The obligations that they have that are off balance sheet are going to start to rear their ugly heads every once in a while. We're going to have questions about should they be investing just like we had with the Reality Labs business. I think from an investment standpoint, the sentiment is extremely strong with Meta today. And that's in large part because of the success of Muse, but that is not financial success. That is kind of like media success, user success. Hey, we're getting a lot of use out of this. Well, do you want to pay for it? Meh, no, just like you said, probably not. And so that's what I think is actually going to happen over the next few years is Not only Meta, but a lot of these companies are going to go, hey, this is going to be bad for our bottom line. Short term, trust us that it's going to be good for the bottom line long term. And we don't yet know what that business model is going to be. It's such a fair point here that I want to cede to you, and I want to reiterate what I said a moment ago. I am willing to give Meta the benefit of the doubt and say this is going to work out as a long-term investment for shareholders. At the same time, I will say that the execution risk is much higher here than it's been in the past. You referenced the Facebook build out back in the day, spending a lot of money that it wasn't making back at the time. Okay, that was one thing. much higher when it moved into the metaverse Even then, it was able to eventually say, okay, this isn't working. We're going to pull back, and we're going to start returning money to shareholders. We're going to start paying a dividend, for example. That was a move that was very well rewarded, and even though they were spending a ton, it was still – Fixable from a financial standpoint, you have now an order of magnitude more than what you were investing 10 years ago into CapEx. So the execution risk much, much higher here. It could be a lot more painful for shareholders if it doesn't figure this out. Yeah, and at the end of the day, as a shareholder in meta platforms, you are just following along with what Mark Zuckerberg wants to do. There is no way to vote him out as CEO. There is no way. Except for the stock price going down, that seems to be the only thing that kind of gets a reaction from management at this point, which maybe is okay. I'm not saying that this is a terrible thing. because this has worked out extremely well, but just something to keep in mind. that he has, you know, super voting rights. And... If this doesn't work out, He is still going to be extremely wealthy. So are all of these other people that he's hired to come on and build some of these AI products. you know as an investor this is not a stock that i'm buying today i want your thoughts on that in the biggest reason I think is we are going to go through this valley of financial performance. They do not have the rent [15:00] business. to fall back on like. Alphabet has for example or like Microsoft has where, you know what, maybe we're not going to put... so much into building using building our own tools our own artificial intelligence tools because we're going to be able to make so much money renting this stuff out to other people I think that's been a detriment to Alphabet's technology progress and product progress but financially it's probably going to be good for them. That may flip eventually, but in the meantime we're probably going to see you know some pretty rough uh quarters from at least a free cash flow and maybe even an income perspective as some of these things play out but john what do you think about meta platform stock today If you're making me choose between buy and sell, I'm choosing buy. I think that there's enough opportunity here and enough history of – surprises that I'm willing to take that bet. in a diversified portfolio. Let us know what you think. What is the payoff from all of this AI spending going to be for meta platforms? Leave your comments in the comment section below. Don't forget to subscribe to the Motley Cools channel here on YouTube. Thanks for watching, everybody. See you here next time. *Ding* . Thank you.

Meta第三季资本支出将超运营费用 拒售算力令千亿AI投资承压