Amid an extraordinary surge in AI development, Meta is outspending most of its peers. The company is constructing two enormous data centers, and reports suggest that up to $600 billion could be invested in U.S. infrastructure over the next three years.
While such numbers may not surprise those in Silicon Valley, they are beginning to unsettle investors on Wall Street.
This concern came to the forefront this week when Meta released its quarterly earnings, revealing a $7 billion year-over-year increase in operating costs and nearly $20 billion in capital expenditures. These rising costs are driven by heavy investments in AI personnel and infrastructure, which have yet to generate significant revenue. When pressed for details, Mark Zuckerberg indicated that this level of spending was only the beginning.
“The best course of action is to speed up our efforts to ensure we have the necessary computing resources for both AI research and our new initiatives, and to shift our core business’s computing capabilities,” Zuckerberg explained to analysts. “We believe that once we integrate the new models we’re developing in MSL and introduce truly advanced models with unique features, we’ll unlock a tremendous untapped opportunity.”
If his intention was to calm investors, it failed. By the end of the earnings call, Meta’s stock price had sharply declined. Two days later, the downturn continued, with Meta’s shares falling 12% by Friday’s close, wiping out over $200 billion in market value.
It’s risky to read too much into stock movements, and in pure financial terms, Meta’s quarterly results weren’t disastrous. (A $20 billion profit in a single quarter is still impressive.) However, this was the first time Meta’s aggressive AI investments in both talent and infrastructure had a noticeable effect on its financial results. More troubling was the lack of clarity about what these massive expenditures had actually achieved, aside from building huge data centers and hiring well-paid AI experts.
Analysts questioned Zuckerberg about the rationale behind such heavy AI spending and when it might start generating returns. But the call came at an awkward moment in Meta’s planning cycle, with no defined budget for future investments and no flagship product to anchor revenue projections. As a result, Zuckerberg could only offer broad statements about AI’s potential.
“We’re beginning to see a range of new products based on different content types,” he said during the call. “There are also business applications for these innovations, such as business A … and more advanced models will enhance our core business, improve recommendations across our Family of Apps, and boost advertising performance.”
Meta isn’t alone in pouring billions into AI infrastructure, so it’s worth considering why similar spending hasn’t rattled investors at Google or Nvidia, both of which reported strong quarters. OpenAI is investing comparable amounts, despite having a much smaller financial cushion than Meta.
There are genuine worries about a potential bubble, but if one does form, Meta’s main business may help it weather the storm better than others.
If you ask Sam Altman why he’s investing hundreds of billions in computing power, he’ll say he’s running one of the fastest-growing consumer platforms ever — one that’s already generating $20 billion in annual revenue. While there’s debate about how long that growth can last (a topic for another time), there is a rapidly expanding product at the heart of the OpenAI phenomenon. A soaring ARR figure helps address many concerns.
Meta, on the other hand, lacks such a product, and it’s uncertain where one might emerge from.
Meta’s most advanced AI offering is the Meta AI assistant, which Zuckerberg mentioned now boasts over a billion active users. However, this figure is likely inflated by the three billion users on Facebook and Instagram, and it’s difficult to view the current Meta AI as a true rival to ChatGPT. The Vibes video generator did increase daily active users, but its business impact remains limited.
The most ambitious effort is the Vanguard smart glasses, launched earlier this month. Yet, these glasses seem more like an extension of Meta’s Reality Labs than a serious attempt to leverage LLMs’ capabilities.
In short, these are intriguing trials, not fully realized products.
It’s telling that when questioned about infrastructure spending, Zuckerberg didn’t highlight recent launches, but instead emphasized what’s coming next.
Zuckerberg underscored his enthusiasm for upcoming products, particularly those stemming from the Superintelligence Lab’s new models.
“Meta AI isn’t just an assistant,” he stated. “We plan to develop innovative models and products, and I look forward to sharing more details when we’re ready.”
However, since this was an earnings call rather than a product unveiling, all he could promise was more information “in the coming months.”
Judging by the market’s reaction, that answer is losing its effectiveness.
To be fair, only four months have passed since Zuckerberg reorganized Meta’s AI division, and the new Superintelligence team hasn’t yet had the chance to launch a groundbreaking AI product. Still, as the company pours billions into staying ahead in AI, there’s no clear vision for what role Zuckerberg wants Meta to play in this evolving sector.
Will Meta AI leverage the company’s vast personal data reserves to become a ChatGPT competitor? Is Vibes the first move toward a consumer entertainment strategy, building on Meta’s targeted advertising? Or do Zuckerberg’s mentions of “business AI” hint at a more robust enterprise offering?
At this point, it’s anyone’s guess. Whatever the outcome, Meta is under increasing pressure to find the answer — and soon.
