Markets

Meta Platforms is spending like it’s building a new internet

Date Published

Meta Platforms ramps AI spending as layoffs and cloud deals hit

TL;DR

Quick Summary

  • Meta is reportedly planning layoffs of more than 20% while ramping 2026 capex to $115B–$135B, signaling a full-company pivot toward AI infrastructure.
  • A reported $12B, five-year dedicated AI capacity deal with Nebius (NBIS) shows Meta is locking in compute supply, not just “shopping for cloud.”
  • Meta’s AI strategy is increasingly about chips and power—reinforced by a late-February 2026 report of a major AMD (AMD) AI chip deal.

#RealTalk

Meta is spending like compute scarcity is the real enemy—and reorganizing its workforce as if everything else is secondary. That can be strategically smart and socially messy at the same time.

Bottom Line

For investors, today is another reminder that Meta’s 2026 narrative is less about a single app update and more about whether massive AI infrastructure spending can keep strengthening its ad engine while managing the human and financial costs of that buildout.

Meta’s new era: fewer people, more machines

Meta Platforms, Inc. (META) is having a very 2026 kind of Monday: it’s reportedly preparing to cut more than 20% of its workforce while simultaneously leaning harder into AI infrastructure than almost anyone else. The tone isn’t “panic.” It’s “we’re reorganizing the entire company around compute.”

If that sounds cold, it’s also the most honest summary of the last two years in Big Tech: the product is increasingly AI-shaped, the bottleneck is increasingly GPUs (and power), and the companies that win aren’t just the ones with the best models—they’re the ones that can feed those models at scale.

The numbers are loud. Meta has signaled $115 billion to $135 billion in 2026 capital expenditures, largely tied to AI infrastructure. That’s on top of a 2025 spend that was already massive (about $72.2 billion). In other words, Meta is treating AI capacity less like an “initiative” and more like a national infrastructure project—except it’s private, and it lives in data centers.

The Nebius deal: renting a firehose

Today’s biggest plot point is Meta’s reported five-year agreement with Nebius Group (NBIS) for $12 billion of dedicated AI capacity across multiple locations.

Read that again: dedicated capacity. Not “we’ll try to find you space.” Not “best effort.” This is Meta reserving future compute the way airlines reserve gates—because once the AI rush starts, everyone shows up at the same airport.

The meta-story (lowercase m) is that the AI buildout has split tech into two tribes:

  • Companies that own huge infrastructure and keep building it
  • Companies that need huge infrastructure and increasingly rent it

Meta is doing both. It’s building aggressively, and it’s signing deals that effectively say: we’re not waiting in line.

Layoffs and the new cost math

The other headline—planned layoffs of over 20%—lands differently at Meta because it’s happening alongside a hiring market where top AI talent can cost a fortune. That’s not a contradiction; it’s the new corporate math.

Meta ended 2025 with about 78,865 employees. If the reported cuts are in that ballpark, this isn’t a tiny trim. It’s a structural reset: fewer layers, fewer “nice to have” teams, and more dollars redirected to servers, chips, and the people who know how to make them useful.

This is also why the market keeps giving Meta a longer leash than it gives many other “AI spenders.” Meta’s core ad machine still throws off cash, and AI has increasingly been framed as a way to make that machine smarter—better targeting, better creative tools, better measurement, better recommendations. When AI improves the thing that already pays the bills, investors tend to tolerate the scary capex line item.

Chips, power, and the uncomfortable truth about “AI apps”

Meta’s AI story isn’t just chatbots in your DMs. It’s the industrial side: chips, model training, inference, and the electricity bill that comes with all of it.

In late February 2026, the Associated Press reported Meta would buy AI chips from Advanced Micro Devices (AMD) in a deal worth up to $100 billion, with an option that could allow Meta to buy up to a 10% stake in AMD. Whether every dollar of that headline number materializes or not, the direction is clear: Meta is trying to lock down supply in a world where “compute” has become a competitive moat.

So what’s the real market takeaway today?

Meta isn’t pitching a single feature. It’s pitching inevitability: that AI will be woven into ads, messaging, video, and whatever “social” becomes next—and that the company willing to spend uncomfortably large sums to own the infrastructure will be the company still standing when the easy phase ends.

In 2026, Meta is not acting like a social media company with an AI project. It’s acting like an AI infrastructure company that happens to own Instagram.