An ongoing shortage of compute could provide Meta with the opportunity to sell its capacity for a big premium

Meta is targeting 14 gigawatts of total compute capacity by 2027 as it aggressively expands its data-center footprint.

Shares of Meta Platforms are once again getting squeezed by the company's elevated artificial-intelligence spending plans. After briefly getting a boost from a new model launch in early July, Meta's stock is down 6% in the past month.

But Meta (META) is sitting on what could be a $22 billion annual gross revenue opportunity starting in 2027, according to Evercore ISI analyst Mark Mahaney.

In a Monday note, Mahaney raised his price target on Meta to $860 from $820 previously, pointing to potential upside from the company selling its excess AI compute capacity to external buyers. The new price target reflects a gain of over 50% from the $559 level that Meta shares closed at on Monday.

The stock was rising 1.1% in recent morning trading Tuesday.

Meta is the only hyperscaler without a public cloud business, but recent reports suggest the company could be rolling out a new Meta Compute division to monetize its data-center capacity beyond internal use. The company is aiming to bring 14 gigawatts of capacity online by 2027 and planning to manufacture its own Iris chips.

"In a market where almost every scaled operator is sold out, Meta is a very rare potential merchant seller of surplus capacity," Mahaney wrote.

While Mahaney doesn't anticipate Meta becoming a full-fledged neocloud business such as Nebius (NBIS) or CoreWeave (CRWV), he believes Meta Compute will be a "tactical monetization of a deliberately thin slice of potential surplus capacity."

Mahaney said management appears to be purposely keeping expectations low, with CEO Mark Zuckerberg saying on the recent earnings call that Meta is getting offers to buy compute at a premium, but that "it would be foolish to basically just sell all of the compute and take a short-term profit," because "there will continue to be a significantly higher margin on selling intelligence rather than selling compute directly."

If Meta were to sell its excess capacity, putting just 0.5 gigawatts to 1 gigawatt on the market could bring in between $11 billion and $22 billion of external revenue per year, according to Mahaney's estimates. Leasing out 1 gigawatt would only take up 7% of Meta's total 14-gigawatt capacity goal for 2027 while adding up to $4.32 in per-share earnings, he added.

Investors should think of a potential compute business as a "call option," according to Mahaney. "We would use compute optionality as a reason to own the AI infrastructure play in Meta, but we would not underwrite a 'full blown' hyperscaler compute annuity stream," Mahaney wrote.

Meta shares are trading at 17 times analyst earnings-per-share estimates over the next 12 months, within 10% of their trailing three-year trough valuation multiple, according to Mahaney. He believes the company is "fully capable" of showing a return on investment from its AI spending plans.

"Over the last two years, Meta has decisively proven that it can effectively deploy AI to materially improve its customer experience," Mahaney said. And the possibility of a Meta Compute division is "still on the table."

-Christine Ji