Meta Is Spending $145B on AI While Cutting 10% of Its Workforce. The Market Called It an Existential Bet — Then Subtracted $170B.
Meta is spending $145 billion this year on AI infrastructure, more than the GDP of most countries, while simultaneously cutting 10 percent of its workforce. The two announcements arrived one week apart, and the market's verdict was unambiguous: $170 billion in market capitalization vanished within days. The company calls it an existential bet. The market is less sure.
The bet has a name now. Hatch is Meta's consumer AI agent, a digital assistant designed to act on behalf of users inside Instagram and, eventually, across the wider web: booking appointments, completing purchases, filing expense reports. Powered by Meta's Muse Spark model, the product targets one billion users. That is not a market expansion goal. It is an existence proof: if Hatch works for a billion people, Meta is no longer a social media company with an AI feature. It is an AI company that happens to sell advertising.
The financial case for the bet looks strong in the near term. Meta reported $56.3 billion in Q1 2026 revenue, up 33 percent year-over-year, with capital expenditures of $19.8 billion in the quarter alone Meta Q1 2026 Earnings Call Transcript. Sessions per user on Meta AI rose double-digit percentages after Muse Spark shipped, and the standalone Meta AI app has consistently ranked near the top of app stores Meta Q1 2026 Earnings Call Transcript. Those numbers are real. They are also exactly the numbers you'd expect from a company that has 3.3 billion daily users and can bolt AI onto every surface they already occupy.
The problem is the bolt-on trap. Meta has tried this before. Meta AI shipped to every Facebook and Instagram user in 2023 and ranked in app stores. It did not move the needle on ad revenue or Meta's valuation multiple. The metaverse consumed tens of billions before the pitch changed. Hatch is being sold as different — genuinely autonomous, genuinely useful outside the Meta ecosystem, genuinely capable of becoming the operating system for a user's digital life. Whether that claim holds is the question the next two quarters will answer.
The structural conflict underneath the product announcement is real. Consider a concrete transaction: a user asks Hatch to research a product across five retailers, compare prices, check reviews, and buy the best option. No sponsored listing appeared. No CPA bid was won. No impression was served. The user asked a question and received an answer that led directly to a purchase — with Meta's inference cost generating zero advertising revenue. The same dynamic applies when Hatch rebooks a flight based on calendar context, files an expense report, or reserves a restaurant. These are all actions that, in the current advertising-funded web, would generate clickthroughs, impressions, or affiliate revenue. In an agentic paradigm, they generate none of those things.
That tension is not hypothetical. JPMorgan analyst Doug Anmuth, one of Meta's most persistent bulls on Wall Street, downgraded the stock to Neutral from Overweight on April 30th and cut his price target to $725 from $825. His reasoning, as reported by Investing.com: "full-stack AI competition is intensifying and Meta has a more challenging path to returns on heavy AI capex beyond advertising." That is the market's diagnosis in a single sentence. Session growth metrics flatter the product numbers. They do not answer the revenue-per-transaction question investors are actually asking when they reprice a $145 billion capex program.
Google has spent two decades optimizing for a world where humans initiate queries and see sponsored results. Meta is now describing a world where agents initiate transactions and users act on recommendations without clicking an ad. These are different businesses wearing the same vocabulary. An agent that actually works is worth a lot to users and very little to the ad network that funds Meta unless Meta finds a way to tax the transaction itself rather than the attention around it.
Meta has not disclosed how it plans to monetize Hatch directly. The company is not alone in this uncertainty. Google's AI Mode has faced similar questions about whether conversational commerce replaces or supplements search ads. Perplexity has not solved the monetization problem either. Meta's own business AI division, which now facilitates 10 million conversations a week for small businesses, is not currently generating revenue TechCrunch. CEO Mark Zuckerberg has hinted the pricing model may change. "May change" is not a business model.
What to watch next: whether Hatch escapes the app-store curiosity phase and starts changing retention or transaction metrics in ways that show up in a quarterly report — or whether Meta spends another $145 billion proving that 3 billion users is a lot of users but not a new kind of business.