META
Meta’s weird little AI side hustle
Evercore ISI is basically asking: what if Meta’s monster AI buildout isn’t just a money pit, but also a future revenue machine? Analyst Mark Mahaney lifted his price target to $860 from $820 and argued Meta could one day rake in as much as $22 billion a year by renting out excess compute capacity.
That’s not a “Meta becomes a cloud company” thesis. It’s more like: Meta buys the really expensive sports car, then rents it out on weekends when it’s not cruising down the highway.
The math is doing a lot here
The bullish case hinges on two things:
- Meta keeps building out capacity fast, with plans to roughly double compute from 7 gigawatts in 2026 to about 14 gigawatts in 2027
- GPU rental prices stay rich enough in 2027 to make surplus capacity lucrative
Evercore says even half a gigawatt of commercialized capacity could mean about $11 billion in annual gross revenue. At a full gigawatt, the estimate jumps to $22 billion — which the analyst says could add as much as $4.32 in annual EPS.
Why investors should care
This matters because Meta already plans to spend a ridiculous $130 billion to $145 billion this year. So the market has been squinting at the bill and wondering: are they overbuilding, or just building ahead of demand like a very expensive chess move?
There’s a catch, of course. Meta still needs tons of compute for its own AI ambitions, so the $22 billion figure is more of a “call option” than a clean forecast. In other words: nice upside if it works, but don’t mortgage the house on it.
The broader AI trade stays messy
Meta is also a giant buyer of outside capacity, with roughly $35 billion in commitments to CoreWeave and up to $27 billion pledged to Nebius. So the company could end up playing both sides of the AI infrastructure trade: buying where it’s short, renting out where it’s long.
And Nvidia? It gets dragged into the story because GPU pricing is the whole ballgame. If H100 rental rates stay hot, Meta’s little compute landlord fantasy looks a lot more plausible.
Big picture: this is still a Wall Street model, not a done deal. But if Meta can turn spare AI power into real cash, the market may start treating its capex bill less like a warning sign and more like a very expensive option on the future.