During Meta's second-quarter earnings call on Wednesday, CEO Mark Zuckerberg laid out a central tension in the company's AI infrastructure strategy: how much compute capacity to sell now at a premium versus how much to reserve for developing future models and services. The dilemma comes as Meta's capital expenditures surge, free cash flow plunged 90% year over year, and the company issued a weaker-than-expected revenue forecast, sending the stock down more than 7% in after-hours trading.
The capacity trade-off
Meta is the only major U.S. hyperscaler without a cloud infrastructure business, yet its capex now rivals that of Amazon, Microsoft, and Alphabet. Zuckerberg told investors the company is fielding "a lot of offers for compute at a significant premium over what we paid for it." He described the core decision as a portfolio problem. "A common trade-off that we need to make is around how much do you monetize something today versus develop future assets," Zuckerberg said. "I think that it's always a portfolio."
Zuckerberg described the balancing act as a classic challenge in AI for Product Development: whether to take short-term profit or invest in capabilities that could become far more valuable later. He added, "It would be foolish to basically just sell all of the compute and take a short-term profit."
Meta recently bumped the low end of its 2026 capex guidance by $5 billion, bringing the range to $130 billion to $145 billion. The spending supports massive data center construction as the company races to build out AI infrastructure under new AI chief Alexandr Wang, who joined earlier this year.
Enterprise ambitions and new hires
Zuckerberg acknowledged that selling compute and services to businesses would require building a new muscle. Meta has historically struggled in enterprise markets, earning 98% of its revenue from digital advertising. He said the potential enterprise business isn't just about raw capacity - it also includes APIs, productivity tools, and AI agents the company is developing.
As CNBC reported earlier this month, Anthropic is in preliminary talks to lease computing power from Meta. Zuckerberg did not confirm any deals but said the company is evaluating a range of offers. A former longtime Amazon Web Services senior executive, Dave Brown, is set to join Meta, signaling a more serious push into cloud-like services.
Investor skepticism and the metaverse hangover
Wall Street remains cautious. Brent Thill, an analyst at Jefferies, told CNBC, "I think everyone wants clarity into what he wants to do in the compute business." Meta's AI strategy has been scattershot, leaving it behind OpenAI, Anthropic, and Google in the market for top models and services.
Zuckerberg's track record with big bets also weighs on the conversation. The metaverse pivot, launched in 2021, continues to cost billions each quarter. Reality Labs, the division building VR devices and wearables, lost $4.62 billion in the latest period on just $431 million in revenue. Still, Zuckerberg framed the AI infrastructure investment as a wager he's willing to make. "My personal bet is that the people who invest in this are going to be rewarded and feel very good over time," he said.
Why this matters for IT, development, and product professionals
Meta's balancing act underscores the strategic importance of AI for IT & Development and AI for Product Development in a resource-constrained market. The choices companies make about infrastructure, capacity leasing, and model development directly affect the tools and platforms available to technical teams. Professionals who understand these trade-offs - and the emerging market for AI compute - will be better positioned to plan their own build-versus-buy decisions and anticipate shifts in the enterprise AI landscape.
Your membership also unlocks:
