Much of the commentary around Meta Platforms' (META 3.87%) disastrous third-quarter earnings report focuses on the prolific spending related to the metaverse. The Reality Labs segment, responsible for the company's virtual reality and metaverse initiatives, posted an operating loss of $3.67 billion and generated just $285 million of revenue.

Meta expects operating expenses tied to Reality Labs to surge further in 2023. This outlook is a big reason why shares of Meta had tanked more than 20% by Thursday afternoon. The company's aggressive plans to invest in its data centers got a bit less attention.

Betting big on AI

Meta needs plenty of compute capacity to handle the billions of users who scroll through Facebook and Instagram each month. But your standard data center server isn't going to be able to handle artificial intelligence (AI) workloads efficiently. Beefy graphics cards or specialized chips tailor-made for AI are required to run AI workloads at scale.

Meta has decided that it needs to upgrade its data center footprint to support next-gen AI hardware. The company wants to make greater use of AI to improve how it serves content and ads to its users. "We expect these investments to provide us a technology advantage and unlock meaningful improvements across many of our key initiatives, including Feed, Reels and Ads," said CFO Dave Wehner during the third-quarter earnings call.

On top of buying expensive servers and networking equipment for its existing data centers, Meta is also expanding its data center footprint. The company believes this will give it greater flexibility, which is expected to generate cost savings over time.

Meta plans to spend between $32 billion and $33 billion on capital expenditures this year, ramping up to a range of $34 billion to $39 billion in 2023. Beyond next year, the company's spending will be guided by the return on these investments. For reference, the company spent just $19.2 billion on capex in 2021.

This investment cycle is different

This isn't Meta's first major investment cycle. The last one, in 2018-2019, saw capex as a percentage of revenue surge just like it's doing now.

Meta's annual capex and capex as a percentage of revenue.

Data source: Meta and Yahoo! Finance. Chart by author. Figures for 2022 and 2023 are based on the average of Meta's capex guidance and the average analyst estimate for revenue.

The difference this time is that Meta has stopped growing. The core advertising business is facing various headwinds, including privacy changes Apple made on iOS devices, a slowdown in advertiser demand, and the surging popularity of newer apps like TikTok.

Investing in AI should help Meta serve more relevant content and ads, which would make the company's apps more appealing for advertisers. But all the AI in the world isn't going to make Facebook popular with younger users again.

Meta's total revenue slumped 4% year over year in the third quarter. Ad impressions rose 17%, but the average price per ad tumbled 18%. Investing in AI to improve how much Meta can generate per ad impression makes sense, but the scale of these investments should be at least a little worrying for investors. Meta isn't investing to support its growth like the last time around, it's investing to restart it. That's a much riskier proposition.

To its credit, Meta is doing the opposite of what many companies would do in this situation. Instead of slashing costs to prop up the bottom line, Meta is plowing cash back into its business. The problem is that these investments don't address the core issues facing Meta's advertising business. What good are more relevant ads and content if people are increasingly turning to competing platforms like TikTok?

Despite Meta stock tumbling more than 70% from its peak and reaching a price-to-sales ratio below 2.5, it's a risky bet. Valuation doesn't matter when the future is clouded with uncertainty. To me, Meta seems like the most likely mega-cap tech company to not exist in a meaningful way 20 years down the road. I can't put odds on that, so I simply won't bother with the stock at all, no matter how cheap it looks.