Meta orders eight reactors that don't exist yet, delivery promised by 2032

A sales sheet for hardware that hasn't powered on

TerraPower and Meta announced a deal for up to eight Natrium advanced reactors to power Meta's AI data centers: two firm units plus rights to six more, 345 MW of baseload each, scaling to 2.8 GW baseload and 4 GW with storage across the fleet. CEO Chris Levesque framed it as a maturity milestone: "we have completed our design, established our supply chain, and cleared key regulatory milestones." First delivery is targeted for 2032 at roughly $6,000/kW, about $2.1B per reactor.

The Natrium design is sodium-cooled and ships with a molten-salt thermal battery: 345 MW steady, boosted to 500 MW for over five hours on demand. TerraPower's first one, in Kemmerer, Wyoming, is still under construction, completion expected around 2030. So Meta is buying a series of a machine whose prototype isn't finished. The thread zeroed in on exactly that gap.

"Given that they haven't actually built one…"

The most pointed objections targeted the performance numbers, not the technology.

"A dual Natrium reactor site can provide 690 MW of reliable 24/7 365 power. Given that they haven't actually built one, asserting the performance seems inappropriate, especially the uptime which IIRC is far, far higher than is typical for proven designs, let alone a new one." — an engineer questioning the advertised uptime

A reply granted that mature plants do hit 90%+ capacity factors, so the ceiling isn't fantasy. The dispute is claiming it before unit one runs. On the schedule, a skeptic predicted the whole thing quietly disappears:

"My guess is this deal is quietly swept under the rug when the first reactor fails to go fully online." — a commenter on the 8-reactor build-out

That drew a substantive correction: Wyoming is a demonstration of the exact production design, and it has NRC approval, the commenter argued, asking what extra regulatory hurdles the eight plants would actually face.

Is this real money or a press release?

The deal's wording left even supporters unsure how much Meta is committing.

"The wording there implies some upfront money from Meta, and that this isn't just a PPA like we normally see. But with no numbers attached it's hard to know if it's a serious investment or just PR fluff." — a reader parsing the funding language

Underneath ran a louder argument about who should own reactors at all. One commenter wanted nuclear kept utility-owned and tightly regulated, away from "outside interests." The counter, from a commenter who works on related committees: Meta will resell the capacity and harvest local tax incentives, and most of the nuclear supply chain is already private anyway.

The pattern, not the press release

Strip the gigawatts and this is an AI capex story wearing a reactor's clothes. Meta has no flagship AI product the way it sells ads, yet it is underwriting baseload power for data centers a decade out. One voice put the trade plainly: even private nuclear that accelerates cheap, clean energy beats burning more coal. The honest read is that a 2032 date on a first-of-its-series reactor is a bet, and Meta is large enough to make bets that read as inevitabilities until a prototype slips.