Story
Meta Shares Climb as Leaked Memo Reveals AI Infrastructure Costs Are Half of Estimates
Summary
Meta Platforms is seeing its stock rise not on new AI models, but on a report revealing the company is building its AI compute capacity at nearly half the cost previously estimated by Wall Street, according to analysis from BofA Securities.
Meta Platforms (NASDAQ: META) shares are gaining traction due to a newly revealed, significant cost advantage in its artificial intelligence infrastructure build-out, rather than the launch of its latest AI model. An analysis by BofA Securities, based on an internal company memo first reported by Reuters, suggests Meta is developing its massive AI capacity at a far lower cost than Wall Street had anticipated.
The Efficiency Revelation
The core of the investor optimism stems from Meta's capital expenditure (capex) efficiency. For months, a key concern for investors—the "bear case"—was the immense cost associated with building the necessary data centers and compute power for its AI ambitions. The memo indicates Meta is managing these costs far more effectively than expected.
According to the BofA analysis of the Reuters report:
- Capacity Goal: Meta is working to add a total of 14 Gigawatts (GW) of compute capacity during 2026 and 2027.
- Cost Discrepancy: BofA previously estimated a build-out cost of approximately $45 billion per GW. However, based on the memo's figures and Meta's projected capex, the actual cost is tracking closer to $22 billion per GW.
In a note to clients, BofA Securities analyst Justin Post reiterated a Buy rating and an $835.00 price target on Meta. Post highlighted that the reported capacity growth for 2026 is "well above" the bank's own estimates, suggesting Meta has "engineered significant cost savings."
Market Impact and Context
AdThis development fundamentally alters the financial outlook for Meta's AI investments. By potentially halving the cost to build its AI infrastructure, the company dramatically improves its potential return on investment (ROI) and profitability relative to competitors. This directly counters concerns that its AI spending would be an unrestrained cash drain.
Post noted that building AI capacity at under $30 billion per GW could yield "significant positive economics" compared to cloud revenue estimates for rivals like Amazon and Google. The news suggests Meta has developed a hyper-efficient, vertically integrated system for its AI expansion.
Custom Chip Strategy
The Reuters report also detailed Meta's plans to manufacture its own custom chip, code-named "Iris," with partners Broadcom and TSMC starting this fall. However, BofA pointed out that because manufacturing has not yet begun, the Iris chip is not the source of the current cost savings.
This implies Meta's cost efficiencies are already being realized through other means, with the custom chip pipeline set to provide an additional long-term advantage. The company reportedly plans to introduce new custom chips roughly every six months through 2027, securing a strategic, cost-effective supply chain for its future AI needs.