Meta CEO Mark Zuckerberg acknowledged last week that the company's AI agent development "hasn't really accelerated" over the past four months, a rare admission that sent shares down 5% and raised questions about the $145 billion the company is spending on AI infrastructure this year.
Meta's AI agent development "hasn't really accelerated" over the past four months, Chief Executive Officer Mark Zuckerberg said, a candid admission that erased about $75 billion in market value and cast doubt on the company's $145 billion infrastructure spending plan.
"The trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," Zuckerberg said during a town hall meeting last week, according to a recording obtained by Reuters.
Meta shares fell 5% on the news, bringing the stock's decline from its August all-time high of $796.25 to roughly 25%. The company now trades at 17.7 times forward earnings, the lowest multiple among its hyperscaler peers, compared with Amazon at 25.5 times and Alphabet at 24.9 times.
The admission challenges the core narrative underpinning Meta's valuation — that massive AI spending would translate into new revenue streams. Meta has committed between $125 billion and $145 billion in capital expenditures for fiscal 2026, up from a prior range of $115 billion to $135 billion, with much of that bet riding on AI agents becoming a commercial product.
The $145 Billion Question
Meta's spending plans have drawn scrutiny from Wall Street even before Zuckerberg's admission. The company's capital expenditure for fiscal 2026 is expected to reach as high as $145 billion, a record sum that exceeds the entire market capitalization of most S&P 500 companies. Needham analyst Laura Martin argued in a note that Meta may have overbuilt its AI infrastructure and is now exploring a cloud computing business to absorb excess capacity, though the company has said it remains compute-constrained internally.
The cloud plan, which Meta confirmed to CNBC's Jim Cramer last week, would sell spare computing power to outside companies — a strategy that could generate $20 billion in annual revenue per gigawatt of capacity, according to JPMorgan estimates. But entering a market dominated by Amazon Web Services, Microsoft Azure and Google Cloud, each with decades of infrastructure and enterprise relationships, would be "difficult to enter this late," Martin wrote.
Layoffs and Internal Turmoil
The reorganization effort that accompanied Meta's AI push has also created friction. Thousands of roles were eliminated earlier this year, and Zuckerberg acknowledged during the town hall that the timing of the cuts was miscalculated and not "clean." The company also paused a controversial employee tracking program last month after sensitive data was leaked internally. The program, designed to record everything workers do on their work computers to gather training data for AI, will be made "opt-in" if it resumes, Chief Technology Officer Andrew Bosworth said.
Meta's struggles mirror a broader industry trend. Tech companies across the sector are discovering that replacing human workers with AI agents is far more difficult than anticipated. Ford Motor Co. recently rehired engineers it had let go after AI failed to perform as expected, according to earlier reports.
Zuckerberg said Meta could see benefits from its AI investments in three to six months. But with the stock down 8.5% year to date and the company's next earnings report scheduled for July 29, investors will be watching for evidence that the spending is producing measurable returns. Meta reported fiscal 2025 revenue of about $201 billion and earnings per share of $23.49.
This article is for informational purposes only and does not constitute investment advice.
