Cloud and AI infra already pull in $300B+ a year. Data center vacancy under 1% and they’re power utility constrained. The fiber guys built ahead of demand, these guys are printing money and can’t build new printers fast enough.
But Meta specifically needs returns from AI products to justify the capex. Google and Microsoft eg. have profitable cloud businesses from where they can rent out GPU compute. Meta’s bet is far more risky.
AI build out is more of an extension of datacenter build out though. All the hyperscalers lead AI build out.
Fiber dailed because the telcos overbuilt and demand lagged. When Amazon introduced AWS it succeeded right away because there was lots of demand.
Jeff Bezos Ted Talk 2003 - https://youtu.be/vMKNUylmanQ
Cloud and AI infra already pull in $300B+ a year. Data center vacancy under 1% and they’re power utility constrained. The fiber guys built ahead of demand, these guys are printing money and can’t build new printers fast enough.
But Meta specifically needs returns from AI products to justify the capex. Google and Microsoft eg. have profitable cloud businesses from where they can rent out GPU compute. Meta’s bet is far more risky.
True. But then again they own the consumer side.
If Meta hadn’t invested in AI recommendations a while back they would have lost against TikTok big time.
7 replies →
Is the edge node revenue in the customer/infra graph from investor spend or customer spend? Almost certainly the former, right?