Meta’s Reality Labs division lost another $4 billion last quarter, bringing total losses on the AR/VR unit to a staggering $83.5 billion since 2021, according to TechCrunch AI. That’s an average of $4 billion in losses every single quarter for 21 straight quarters. And as TechCrunch AI reports, the spending is about to get a whole lot worse as Meta pivots its checkbook toward AI infrastructure.
What stands out here is how routine these massive losses have become. Reality Labs bleeding billions is no longer news at Meta. It’s the baseline. The bigger story is what’s coming next on the AI side.
The numbers behind the burn
Meta’s Q1 results were strong on the surface:
- Net income of $26.8 billion, up 61% year-over-year
- Revenue of $56.3 billion, up 33% year-over-year
- Reality Labs loss: $4 billion (again)
- Cumulative Reality Labs losses since 2021: $83.5 billion
The company can clearly afford it. But investors aren’t celebrating. Meta stock dropped more than 5% in after-hours trading once the capex guidance landed.
AI spending eclipses the metaverse bill
Meta now projects 2026 capital expenditures between $125 billion and $145 billion, blowing past both analyst estimates and the company’s own prior guidance. CEO Mark Zuckerberg blamed rising component costs, particularly memory pricing, on the investor call.
“We are increasing our infrastructure capex forecast for this year,” Zuckerberg said. “We are very focused on increasing the efficiency of our investments.”
CFO Susan Li offered an even less reassuring note when asked about 2027.
“We aren’t providing a specific outlook for 2027 capex, and we are, frankly, undergoing a very dynamic planning process,” Li said. “Our experience so far has been that we have continued to underestimate our compute needs.”
Translation: the bill keeps getting bigger, and Meta itself isn’t sure where it tops out.
Why this matters
The pattern here should make every AI watcher pay attention. Meta spent years and tens of billions chasing a metaverse vision that never found product-market fit. Now it’s redirecting that same appetite for massive bets toward AI superintelligence, only at a scale that dwarfs the Reality Labs spend.
A few things to keep in mind:
- Meta poached more than 50 AI researchers and engineers from rivals last year to fuel its push
- That hiring spree produced Muse Spark, the company’s overhauled AI model that shipped earlier this month
- Zuckerberg cited “large increases” in Meta AI usage since the launch
- Compute costs are still climbing faster than Meta can plan for
For practitioners, this is another data point on just how brutal the economics of frontier AI have become. Even a company printing $26.8 billion in quarterly profit is having to defend its spending to nervous shareholders. The capital requirements to compete with OpenAI and Anthropic now rival the GDP of small countries.
What comes next
Expect more capex surprises through 2026, and probably 2027 too. If Meta is openly admitting it underestimates its compute needs, the headline numbers we’re seeing today are a floor, not a ceiling. Investors will keep pressing for an efficiency story. Zuckerberg will keep pointing at usage growth. The burn will keep going.
The question worth watching isn’t whether Meta can afford the spending. It’s whether the AI products that come out the other side justify a bill that’s already eclipsing the metaverse era several times over. Full breakdown available at the original TechCrunch AI report.