Although much of the attention has centred on Meta’s (NASDAQ:META) newly unveiled Muse Spark 1.1 AI model and its shift toward a paid developer strategy, investors are increasingly focused on a far less visible development that could have a much greater impact on the company’s long-term value.
According to an internal company memo first reported by Reuters, Meta’s AI infrastructure buildout is proving to be significantly more cost-efficient than previously expected. The findings prompted BofA Securities analyst Justin Post to reaffirm his Buy rating on Meta Platforms with a price target of $835.00.
Why Investors Are Paying Attention
For months, one of the biggest concerns surrounding Meta has been the enormous capital investment required to expand its AI capabilities. Developing advanced artificial intelligence demands massive data centre capacity and substantial electricity consumption.
However, the Reuters report suggests Meta is delivering that expansion far more efficiently than the market anticipated.
BofA highlighted several key points from the internal memo:
- Massive Capacity Expansion: Meta is targeting an additional 14 Gigawatts (GW) of computing capacity during 2026 and 2027. The memo states that 1GW has already been deployed in 2026, with another 5.5GW expected to come online during the second half of the year.
- Lower-Than-Expected Costs: BofA had previously estimated Meta’s infrastructure buildout would cost roughly $45 billion per GW. Based on the capacity figures outlined in the memo and Meta’s projected $145 billion capital expenditure programme, the actual cost appears to be closer to $22 billion per GW.
BofA Sees Significant Upside
Justin Post highlighted the importance of the findings in a note to clients:
“The 6.5MW 2026 capacity growth in the memo is well above BofAe at 2.6GW, and if 2026 capacity estimates in the memo are even close to accurate, Meta may have engineered significant cost savings to get capacity cost per MW well below our and Street expectations.”
If those estimates prove accurate, Meta would be expanding its AI infrastructure at roughly half the cost previously assumed by Wall Street.
That represents a notable shift in the investment narrative. A major concern among bearish investors has been that Meta’s AI spending would consume enormous amounts of capital without producing attractive returns.
Instead, the new analysis suggests those investments may generate much stronger economics than expected.
As Post explained:
“We think building MW of AI capacity at below $30bn per GW could have significant positive economics relative to our estimates for Amazon and Google annual Cloud revenues per GW at $10-16bn or recent SpaceX capacity deals that could range from $40-50bn per year per GW.”
Custom AI Chips Add Longer-Term Potential
Reuters also reported that Meta intends to begin manufacturing its custom AI chip, codenamed Iris, later this year following successful testing. The chip will complement the company’s purchases of GPUs and will be produced alongside partners Broadcom and TSMC.
While investors have welcomed news of the Iris programme, BofA believes it is not responsible for the cost improvements reflected in the 2026 capacity estimates.
Instead, the efficiency gains appear to be coming from Meta’s existing infrastructure strategy, with the custom chip roadmap representing additional upside over the coming years.
Reuters reported that Meta plans to introduce new custom chips approximately every six months through 2027 while securing multi-year supply agreements with key manufacturing partners, including Broadcom and TSMC.
Post said:
“Given that Iris is just starting to be manufactured in September, it seems unlikely that the chip is driving significant capacity cost savings in 2026, making the Reuters reported capacity GW estimates possibly less likely. However, we see reported progress with chip development as a big positive for Meta (given Cloud margin contribution from TPUs and Trainium), and likely supportive of Meta CEO’s optimism on returns on capacity investment.”
Infrastructure, Not Software, Is Driving the Bull Case
While headlines have focused on Meta’s latest AI software announcements, investors appear to be responding more strongly to evidence that the company is building AI infrastructure far more efficiently than expected.
By significantly reducing projected computing capacity costs while simultaneously developing its own custom chip ecosystem, Meta is strengthening the investment case that its sizeable AI spending programme can deliver substantial long-term returns.