Meta's Revenue Grew 33%. Its Stock Fell 8%. The Number That Scared Investors Was $145 Billion.

Meta reported its fastest revenue growth since 2021 on April 29 — a 33 percent year-over-year increase — and its stock fell 8.6 percent the next day. The earnings beat every estimate that mattered. The problem was the number that came after it: a revised 2026 capital expenditure forecast of $125 billion to $145 billion, up from the prior guidance of $115 billion to $135 billion. An additional $10 billion in AI infrastructure spending, announced alongside the best quarterly revenue growth in five years, was enough to erase all of it.

When analysts on the earnings call asked Mark Zuckerberg for signs of return on the AI investment — actual, visible evidence that this unprecedented spending is producing proportional value — he said: "That's a very technical question." It was not a dodge exactly. It was an accurate description of the problem. The ROI on $145 billion in AI infrastructure is genuinely hard to measure, and the places where it is measurable, Meta has not made the habit of quantifying them clearly.

What Happened

Meta's Q1 2026 results were strong by almost every measure. Revenue climbed 33 percent from the prior year, the fastest quarterly growth since 2021. The company beat earnings per share estimates. Advertising remained the engine, with AI-optimized targeting through Meta's Advantage+ system driving measurable results — Advantage+ had reached a $60 billion annual run rate by February 2026.

The capex revision changed the conversation. Meta attributed the $10 billion increase to "higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity." The explanation was technical and brief. The implication — that hardware costs are rising faster than anticipated, at the same time Meta is building more of it — was not elaborated upon.

JPMorgan analyst Doug Anmuth cut his rating on META to Neutral the day after earnings, citing "intensifying full-stack AI competition" and a "more challenging path to returns on heavy AI capex beyond advertising." The downgrade from one of Meta's historically bullish analysts carried weight.

The market's reaction was not to the revenue. It was to the gap between what Meta is spending and what it can show for it.

Reality Labs, Meta's virtual reality and augmented reality division, reported a $4.03 billion operating loss in Q1 — the same division that has now accumulated over $80 billion in cumulative losses since 2020. The Metaverse bet has not paid off. Meta announced layoffs of 8,000 employees in the same period it raised its AI infrastructure budget by $10 billion. The pattern — cut human costs to fund machine investment — is a rational capital allocation decision. It is also the kind of decision that looks different depending on whether the machine investment pays off.

Why the meta capex Number Scared Investors

The $145 billion figure is difficult to evaluate because Meta's AI investment operates in two very different modes, and the earnings report treats them as one number.

The first mode is AI that is already working. Advantage+, Meta's AI-powered advertising optimization system, demonstrably improves ad performance and is now one of the company's most important revenue contributors. The Andromeda ad delivery system, integrated with Llama 4, drove a 24 percent surge in advertising revenue in the quarters before this one. Meta's AI is generating real returns in the advertising business, and the case for that spend is not speculative. It is the most quantifiable AI investment among any major consumer internet company.

The second mode is everything else. Llama, Meta's open-source large language model series, is not directly monetized. Meta AI, the assistant integrated across WhatsApp, Instagram, Facebook, and Messenger, does not have a separate revenue line. The data center infrastructure supporting both the ad AI and the consumer AI products is built together and funded from the same capex bucket. When analysts ask about ROI and Zuckerberg calls it a "very technical question," the honest translation is: we do not have a clean way to separate the infrastructure that is producing returns from the infrastructure that is not yet.

Google's situation is cleaner. Google Cloud grew 63 percent in Q1 2026. That number is a direct measure of how AI investment is converting to enterprise revenue. Google's ad AI improvements also have measurable attribution. Meta's advertising AI has similar attribution internally — but the company has not made a habit of separating it out in earnings calls, which means investor models cannot verify the relationship between spend and return.

The Metaverse Comparison That Won't Go Away

The number Fortune led with in its post-earnings analysis was $83.6 billion — Meta's cumulative Metaverse losses through the end of 2025. Reality Labs is still losing $4 billion per quarter. The AI capex guidance of up to $145 billion in a single year now exceeds the total Metaverse investment over five years.

The comparison is not entirely fair. The Metaverse was a consumer hardware and platform bet that assumed a future product category that did not materialize on schedule. AI infrastructure is being built on top of a category — large language models and AI assistants — that has already demonstrated product-market fit at scale. Meta's AI is embedded in products used by more than three billion people. The Metaverse never reached that kind of distribution.

But the comparison is not entirely unfair either. The investor concern is not whether AI will matter. It is whether Meta's particular version of the AI bet — $145 billion in infrastructure spending on top of an already-profitable advertising business — will produce returns proportional to the investment, or whether it will produce Reality Labs results at twice the scale.

Economist Dean Baker put it directly: the question is whether Zuckerberg is set to repeat the pattern of large commitments to unproven directions, or whether the AI infrastructure spend is categorically different. The honest answer is that both interpretations of the data are defensible right now, which is why the stock reacted the way it did.

What Meta's AI Is Actually Doing

The most important thing that got lost in the capex coverage is that Meta's AI is already generating substantial revenue. Advantage+ is not a future bet. It is a present contributor with a measurable run rate. Meta's advertising AI has helped the company close the gap with Google in digital advertising and has positioned it to potentially surpass Google as the largest global advertising platform by the end of 2026.

The Llama model series, released as open source, does not generate direct revenue but serves a different strategic function: it positions Meta in the AI developer ecosystem, reduces dependence on proprietary model providers, and generates goodwill that translates into enterprise and API partnerships. Llama 4's integration with the Andromeda ad system has been directly attributed to the 24 percent ad revenue surge in early 2026.

What Meta does not have is a narrative that ties the $145 billion to a specific, non-advertising revenue outcome. Google can point to Cloud. Microsoft can point to Azure and Copilot enterprise licenses. Meta's AI investments outside advertising are either built into an advertising system that analysts already understand, or they are bets on future products — Meta AI, AI-powered social experiences, AI-generated content — that do not yet have separate P&L lines.

Until those lines exist, the meta capex question will be answered by inference and analogy rather than data. Investors will compare it to the Metaverse because the Metaverse is the last time Zuckerberg committed resources at this scale to something with a long feedback loop. The advertising AI success does not make that comparison disappear — it just makes the case that this time the underlying infrastructure has already proven one important use case.

What Happens Next

The $145 billion guidance sets a floor for what investors will expect Meta's AI to return. If 2026 ends with advertising AI continuing to outperform — and Meta's revenue growth sustaining above 25 percent — the capex narrative shifts from "Metaverse redux" to "disciplined infrastructure investment with measurable ROI." If advertising growth decelerates while costs accelerate, the comparison holds and the stock reflects it.

The short-term variable is Reality Labs. A division losing $4 billion per quarter while the company announces $145 billion in other capital spending is a constant reminder that Meta has a history of large commitments that do not produce proportional returns. As long as Reality Labs is losing money, the burden of proof on AI capex is higher than it would be for a company without that track record.

Zuckerberg's "very technical question" response will age well or poorly depending on what the next two earnings reports show. If the $145 billion capex produces a measurable step-change in revenue beyond advertising — enterprise AI tools, AI-powered subscriptions, developer platform revenue from Llama integrations — the question will look like appropriate CFO caution about over-promising. If the revenue mix stays the same while spending grows, it will look like the same thing it sounded like.

The most optimistic read of Meta's position is that it has already proven AI infrastructure works — inside the advertising engine, at $60 billion annual run rate — and is now extending that capability to a broader set of products with a longer feedback loop. That is a credible story. It is not the story Zuckerberg told on the earnings call, and the gap between the credible story and the one he told is why the stock fell 8 percent on the best revenue quarter in five years.

For companies and teams relying on AI tools to do real work, Meta's situation illustrates a pattern that applies at every scale: the infrastructure decisions that are easiest to justify are the ones with clear, measurable outputs. Meta's Advantage+ investment is justified because the output is ad revenue. The rest of the $145 billion is justified by a vision of where AI is going — which is a reasonable basis for investment but a difficult one to defend in a quarterly earnings call. Building knowledge management workflows that actually capture what AI produces, not just what it costs, is the same problem at a different scale. The measurement gap is the real risk.