Meta Is Gutting Itself to Fund AI Bets That Aren't Working Yet

Two stories dropped about Meta this week that belong together.

First: Meta delayed the rollout of “Avocado,” its next frontier model, after it failed internal benchmarks. The March release is now May at the earliest. Second: Meta is planning sweeping layoffs affecting roughly 20% of staff as AI infrastructure costs mount.

The company is simultaneously spending $135B+ on AI infrastructure and cutting a fifth of its workforce. Its flagship model isn’t ready. And the cuts aren’t in unrelated divisions — they’re directly tied to the cost of the AI push itself.

This is what happens when you invest in capability without a clear integration path.

Meta has the compute. They have the talent (for now). They have the data. What they apparently don’t have is a model that meets their own quality bar, or a way to fund the attempt without cannibalizing the business that pays for it.

The “AI replacing jobs” narrative doesn’t fit here. These aren’t roles being automated away. These are people being laid off to redirect cash toward GPU clusters and training runs for a model that just failed its own benchmarks. That’s not efficiency. That’s a company eating itself to place a bigger bet.

For anyone building an AI strategy, the lesson is blunt:

• Spending doesn’t equal results. $135B buys a lot of compute. It doesn’t guarantee a model that works. Meta just proved that at a scale no one else can.
• Capability without integration is waste. A frontier model that doesn’t meet benchmarks is an expensive experiment, not a product. The missing piece is always the same — how does this actually connect to revenue?
• Watch what companies do to fund AI, not just what they spend. If the funding source is mass layoffs, that tells you the AI investment isn’t self-sustaining yet. It’s being subsidized by the existing business.

The hype narrative says more AI spending means more AI progress. Meta’s own internal benchmarks disagree.

Before you approve your next AI infrastructure budget, ask the question Meta apparently didn’t: what happens to the rest of the business if this bet takes twice as long to pay off?