# AI capex — X 热门讨论 (2026-09-28 18:02 UTC)

## @intelfabs (John Intel) · 09-28 12:54 · ♥44 ↻4 💬4 TSMC may spend over $100B in capex by 2028.

That's not a company preparing for slower demand.

It's a company preparing for a lot more chips, a lot more packaging, and a lot more AI.

Just like Intel $TSM $INTC https://t.co/1yOXST4iey https://x.com/intelfabs/status/2104555425836470718

## @LaceyPresley (Lacey) · 09-28 05:02 · ♥33 ↻8 💬2 Orbital compute just got a 2028 cost clock from Elon Musk.

Current estimates:

• BofA: ~$165B/GW • WoodMac: ~$170B/GW • Bull case: ~$68B/GW → ~$43B by 2035

“Will be way better than that by 2028.” - Elon

That puts Starship at the center of the economics.

• Lower $/kg • Higher payload • Faster reuse • Higher cadence

If launch cost collapses, orbital compute capex collapses with it.

That makes Starship more than a launch system.

It becomes the cost engine behind orbital AI infrastructure.

The next flights are the thesis being tested in hardware. https://x.com/LaceyPresley/status/2104436672326742501

## @ilzmcfly (McFly) · 09-28 16:06 · ♥33 ↻3 💬2 Keep in mind $IREN has only disclosed $16.7B in contracted capacity, TCV really stands more at $21-$22b right now.

Their exit ARR for 2026 is likely $4.23B now. Total ARR is $4.91B.

IREN also has more IT capabilities in Childress from their 2026 capacity, but is under wraps for now.

They described 2026 as sold out and $4b of ARR, but that is not exactly the case.

IREN is also executing hard behind the scenes on its software stack, with many high tier hires. 2027 will likely be a first year they offer a GPU cluster management on demand through Mirantis and their Nvidia AI Factories. The market is not open to hearing a company with BTC mining background can shift to managing AI Clusters.

IREN also guided $25-$30B of capex FOR fy2027, but did not attach a ARR number or TCV figure to that. ARR will likely be $17b and TCV above $50B.

IREN also still has BTC miners, but will be removed by year end.

IREN stock could move like sideways for so long until they start showing their colors and delivering. I'll wait, if by year end they delivered $4B+ in operational capacity on time, start contracting 2027 at attractive rate/terms, remove BTC miners, disclose their TCV, technical capabilities and the stock still does not react then I will rethink my thesis.

But imo after a year long of volatile sideways movement there should be some sort of relief rally. https://x.com/ilzmcfly/status/2104603693026824192