# AI capex — X 热门讨论 (2026-10-02 14:47 UTC)
## @Maplethorpej (Jerad Maplethorpe) · 10-02 11:27 · ♥160 ↻8 💬89 I’m getting increasingly worried that there’s no ROI in AI.
Yes, it replaces human effort and speeds up work, but it isn’t delivering more revenue.
But here’s the thing…
Human skills are deteriorating.
The AI labs don’t need to outperform humans, they just need to make us dependent.
Once they’ve achieved that, ROI doesn’t matter. AI will be viewed more like electricity: not optional.
But it’s more sinister than that. Once we’re hooked, the dealers can charge whatever they want.
And they’ll have to in order to justify this capex cycle. https://x.com/Maplethorpej/status/2105983124836384774
## @tonmaxis (Maxi) · 10-02 12:06 · ♥32 ↻1 💬20 China has the hardware and supply chain.
USA has strong robotics/model labs and CapEx.
But the biggest missing piece is still real world data.
The goal? 100M hours of it.
We’re still only in the single digit millions.
That might be one of the biggest bottlenecks for embodied AI.
@axisrobotics > 引用 @tonmaxis: Heard @chris_anm01 and @brian_armstrong this part of conversation and can't shake it tbvh
One percent is only seven doublings from 100
Doubling just means twice as good again and again 1, 2, 4, 8, 16, 32, 64, 128
If robotics keeps improving at that pace, the implications get pretty serious.
Do you think robotics gets there in seven doublings, or is 1% too generous?
@axisrobotics https://x.com/tonmaxis/status/2105992741197172904
## @FroehlichThors1 (Thorsten Froehlich) · 10-02 06:51 · ♥31 ↻0 💬4 #AI is a different capex bubble.
Among all large mega capex cycles over the last 170 years, Ai managed to create hardly any jobs - what follows might be worse as it wasn't that great to begin with. https://x.com/FroehlichThors1/status/2105913457992184210