Meta's $125B CapEx bet worries investors who fear a metaverse repeat

Meta's $125B CapEx Bet Looks Like a Metaverse Repeat—Without the Revenue Story

Meta is burning investor confidence by announcing $125 billion to $145 billion in annual CapEx, and the market reaction is sharp: stock down 10% in five days. The sell-off reflects a specific fear: that Mark Zuckerberg is repeating the metaverse playbook—massive spend with no clear revenue path—just with AI infrastructure instead.

The backdrop makes the timing painful. Meta's core ad business is accelerating at 33% year-over-year in Q1, delivering the kind of margin and impression growth that earned the company its $1.5 trillion valuation. That cash machine is what makes the CapEx announcement sting. Investors see a beautiful, proven business being starved of capital to fund an uncertain infrastructure bet.

The compute oversupply risk

Meta's specific vulnerability is structural. Unlike Google, Amazon, or Microsoft, Meta has no cloud business to absorb excess GPU capacity if its AI spend doesn't translate into revenue. If the company builds out $125 billion in compute and can't fully deploy it, there's no secondary market to sell into—no AWS instance to resell, no Azure customer to offload to. The company has no clear answer for what happens with stranded capacity.

The broader compute market is already showing strain. xAI, despite being positioned as a frontier AI player, reportedly struggles to drive demand for the compute it already has. Elon Musk's solution—partnering with Cursor, which has massive developer demand for AI coding—sidesteps the problem by matching GPU-rich players with GPU-poor applications. Meta has no equivalent dance partner.

The consumer AI story doesn't move the needle yet

Meta AI now ranks in the top five on the App Store, a meaningful achievement in what may be the most competitive category since search. But investors aren't pricing in the app as a meaningful revenue driver. Consumer LLMs are commoditizing fast, and Meta's model, while solid, hasn't produced the kind of differentiated demand that justifies the infrastructure spend.

Zuckerberg has been talking about "personal super intelligence" for nearly ten months—vague language that could mean AI in Ray-Bans or a consumer LLM competitor. The ambiguity itself signals the problem: there's no clear commercial thesis yet, just spend in anticipation of one.

The credibility problem

This is what ties the metaverse comparison. Zuckerberg spent years on a bet that didn't deliver material returns and arguably damaged investor trust in his capital allocation judgment. The CapEx announcement feels like a replay: "all the spending and no revenue acceleration associated with that spending," as one analyst puts it.

The irony sharpens the critique. Meta has historically been one of the world's most effective businesses at converting AI investment into revenue—its ad targeting, ranking, and recommendation systems are second to none. Yet investors aren't extending that credit to the new AI infrastructure plan. The company is sitting on $200 billion in annualized revenue growing at 33% annually, "one of the greatest businesses in all of human history," and not getting full credit for either that or its existing AI work.

That's the real disconnect. Meta has a track record of AI execution. It also has a recent track record of massive speculative bets that didn't pay off. Investors are choosing to remember the latter.