The message to shareholders was blunt. Meta Platforms is considering raising tens of billions of dollars through a large sale of new shares to fund artificial intelligence investment, the Financial Times reported, and investors responded by selling: the stock fell more than 5% in a single session.
Meta has raised its 2026 capital expenditure guidance to $145 billion, a figure that has climbed repeatedly over the past year as the company poured money into data centers, chips and the research teams that build its open-source Llama models. The share sale would be the company’s first large equity issuance in years, a step it had previously insisted it would not need.
The reaction was immediate and sharp. Meta shares are down 13% over the past 12 months, the worst performance among the largest U.S. technology companies, and the dilution news added to a story that investors have grown tired of: rising spending, uncertain returns, and a chief executive whose conviction in AI has not wavered despite the stock’s drift.
Mark Zuckerberg has made clear he considers the spending non-negotiable. He has said Meta’s AI investments, including assistants embedded across its apps, an open-source model family used by millions of developers and an increasingly large physical footprint, will define the company’s next decade. The financing plan suggests he is willing to pay for that vision with equity rather than waiting for cash flow to catch up.
The shift in language inside the company has been telling. Where Meta once described its AI infrastructure spending in the language of capital optimization, matching investment to revenue growth, executives now speak of building ahead of demand. The Financial Times reported that the company’s internal plans treat compute capacity as a strategic asset to be secured at any cost, an approach that has pushed the guidance from roughly $60 billion two years ago to $145 billion today.
The parallel with Alphabet is close. The Google parent expanded its equity offering from $80 billion to $85 billion this week, with Berkshire Hathaway taking $10 billion, and the two companies sit on opposite sides of the same coin: both are financing the AI buildout through equity markets at a moment when their stocks are under pressure. The market is being asked to fund the buildout twice, once through capital spending and now through dilution, one portfolio manager said.
Meta’s arithmetic is harder than Alphabet’s in one respect: its AI products are still finding their revenue shape. Advertising remains the company’s engine, and AI has so far improved ad targeting and creative tools more than it has created new revenue lines. Analysts said the $145 billion guidance implies that Zuckerberg sees AI as infrastructure for everything Meta does, with returns measured in competitive position rather than immediate profit.
The share sale also raises governance questions. Zuckerberg controls the company through a special class of stock, so dilution affects other holders more than it affects him. The structure does not require him to carry the same weight of the decision, which makes the financing choice easier for him and harder for the investors who foot the bill.
Investors who bought Meta during its 2022 lows have seen the stock triple and then give back part of the gain, and the decline of the past year has concentrated minds on the spending trajectory. The company’s cash flow remains strong, with Meta still generating tens of billions of dollars annually, which makes the equity sale more striking: a company with that much cash flow does not raise equity lightly.
Meta’s core business remains strong enough to complicate the story. Advertising revenue has grown through the AI enhancements in targeting and creative tools, and the company’s cash flow covers a large share of the spending without new financing. The equity sale, if it proceeds, would top up that flow rather than replace it, which is one reason the market read the news as a signal about the scale of ambition rather than a sign of distress.
The spending is also an insurance policy of sorts against falling behind. Llama’s open-source strategy depends on keeping frontier models competitive, which requires compute on a scale that grows with every generation. Zuckerberg has said repeatedly that the company would rather overspend than fall behind, and the $145 billion figure is the clearest expression of that preference.
For the industry, Meta’s financing adds another block to a wall of capital. The buildout is being paid for with everything available: cash, debt, equity and engineering improvisation. The tent-style data centers the company is testing, the share sale it is weighing and the $145 billion guidance are all symptoms of the same demand for compute that now outruns the industry’s ability to build.
For Zuckerberg, the bet is that the spending will look cheap in hindsight if AI transforms computing and advertising as he expects. For shareholders, the question is simpler: how much dilution, and how long before the returns show up in the income statement. The share sale, if it comes, will put a number on the first part of that question.
