Mark Zuckerberg says Meta is planning to invest $60B to $65B in capex in 2025, will end the year with 1.3M+ GPUs, and plans to grow its AI teams “significantly”
Meta Platforms (META.O) will invest $60 billion to $65 billion in capital expenditure in 2025, CEO Mark Zuckerberg said in a Facebook post on Friday.
ReutersJaspreet Singh
Context & Ripple Effects
Meta had already raised the low end of its 2024 capital-expenditure guidance while reporting progress on AI infrastructure. This announcement turns that buildout into a much larger 2025 commitment, pairing spending with a stated target of more than 1.3 million GPUs and significant AI-team growth.
The commitment also foreshadows Meta’s subsequent framing of heavy infrastructure spending as a strategic advantage and, later, its Meta Compute initiative for much larger AI infrastructure. The arc is from incremental guidance increases to compute becoming a top-level corporate capability.
First-order effects
Meta commits $60 billion to $65 billion of 2025 capex, directing substantially more capital toward AI infrastructure than its prior 2024 guidance implied.
Meta’s planned GPU fleet and AI-team expansion give its internal AI efforts a larger dedicated compute base, while making infrastructure execution a central management priority.
Second-order effects
A commitment at this scale strengthens demand signals for the hardware, data-center, and other infrastructure inputs needed to deploy Meta’s planned GPU capacity.
Other large AI developers face a higher strategic benchmark: Meta is explicitly treating infrastructure scale as an advantage, rather than only an operating expense.
Third-order effects
If similar commitments persist, AI competition is likely to become more shaped by access to capital, power, and deployable compute—not solely model development.
Meta’s later move to formalize Meta Compute as a top-level initiative suggests this may evolve from a yearly capex cycle into a lasting organizational and industry-structure shift.
The trend: This is a data point in the AI infrastructure capital cycle, where major platforms are turning compute capacity into a strategic, long-lived competitive asset.
All of this crazy AI spending at Meta and other places reminds me of this Zuckerberg quote from last summer. Everyone is likely overspending by billions of dollars on AI. But better safe than sorry. — www.bloomberg.com/news/article... [embedded post]
New: Mark Zuckerberg wants in on all the AI news this week. — He says Meta will be “growing our AI teams significantly” this year. The company said capital expenditures will be $60-65B this year. For comparison, 2024 capex was $38-40B. — In other words: Spending a ton on AI…
It's sort of funny that every American tech company is bragging about how much money they're spending to build their models, and DeepSeek is just like “yeah we got there with $47 and a refurbished Chromebook”
The latest datapoint that the AI trade remains intact. Zuckerberg's comment that $META will spend $60-$65B in Capex for 2025 exceeds Street estimates of $51B. In the near term, increased investment in AI infrastructure benefits $NVDA and other hardware players. Long term,
quick story. the numbers truly are staggering. wall street doesnt seem to mind the (profligate?) spending, since the stock jumped a tiny bit https://www.nytimes.com/...
Zuckerberg drops new numbers on Meta's data center and AI plans this year. they expect to own 1.3 million GPUs but more significantly, they're hiking capital expenditures to $60-65 billion. HUGE increase of more than 50 percent compared to 40 billion in all of 2024 [image]