This reminds me of the chip shortage during the pandemic. Tesla was able to take advantage of this and increase price sharply, which led to record automotive gross margins. This shortage of blades and veins and inability for AI data centers to access the power they need - which of AI at the moment, is not a sustainable long-term competitive advantage for SpaceX. Eventually, supply will catch up and prices per compute will fall.
The key assumption one must make = when this fall take place, and how far do prices drop? The bull case assumes prices keep increasing for at least the next 12 months, and when they do finally fall, they fall to a point where they are still higher than where they are today (or close to it).
$SPCX
The key assumption one must make = when this fall take place, and how far do prices drop? The bull case assumes prices keep increasing for at least the next 12 months, and when they do finally fall, they fall to a point where they are still higher than where they are today (or close to it).
$SPCX
The bottleneck for AI right now isn't chips, it's turbine blades
And this is EXACTLY why Elon is confident they can bring GPUs online faster than the rest of the market
Getting GWs online comes down to power, and suppliers say blades and vanes run 12 to 18 months behind the ause they need advanced casting
While the rest of the market waits, SpaceX has begun casting its own blades with the tech it built for its rocket engines. This is how relentless Elon and his SpaceX team are... they wait for no one!
But Nvidia wont sell its chips unless the company is able to energize them, otherwise it becomes a liability for the industry. So companies that can energize chips the fastest get access to the chips first
And that's exactly Elons plan to reach 13 GW by the end of next year. If they can't get the power, Jensen won't sell them the chips. If they can, they are likely to buy up 40% of the entire Nvidia GPU supply in the next 5 quarters
As I've been saying for a long-time now, the biggest moat in any company for the future is access to compute. $SPCX appears to have the magic formula to get it faster than anyone else!
15
2
51
5,514
@TSLAFanMtln is rushing to build as much compute as possible right now, because it can get paid for by other companies within 6-9months at current spot prices
Then, in 2 years from now as pricing begins to stabilize he can re route that same compute to Grok, FSD, robotaxis and humanoids as they scale up
So all that compute he’ll need in a couple years is just getting paid for by other companies now
Then, in 2 years from now as pricing begins to stabilize he can re route that same compute to Grok, FSD, robotaxis and humanoids as they scale up
So all that compute he’ll need in a couple years is just getting paid for by other companies now
Oct 24, 2026 · 4:41 PM UTC
8
11
176
17,521
@KyleReidhead @TSLAFanMtlPU)를 짓는 이유는, **현재 스팟 가격이 워낙 높아서** 다른 회사에 6~9개월만 임대해도 건설 비용 상당 부분을 회수할 수 있기 때문입니다.
- 지금 AI 기업들은 칩과 전력이 부족해서 높은 가격을 주고라도 컴퓨트를 사려 합니다. 이 임대 수익이 바로 “다른 기업이 비용을 대는” 부분입니다.
- 2년 뒤 가격이 안정되고 공급이 늘어나면, 그때 그 같은 컴퓨트를 Grok, FSD, 로보택시, 옵티머스 등 자기 회사 용도로 돌리는 겁니다.
- 결과적으로 미래의 자체 수요 물량을 미리 확보해두면서, 그 건설 비용은 지금 급하게 컴퓨트를 찾는 다른 회사들의 임대료로 상당 부분 충당하는 구조라는 분석입니다.
- 지금 AI 기업들은 칩과 전력이 부족해서 높은 가격을 주고라도 컴퓨트를 사려 합니다. 이 임대 수익이 바로 “다른 기업이 비용을 대는” 부분입니다.
- 2년 뒤 가격이 안정되고 공급이 늘어나면, 그때 그 같은 컴퓨트를 Grok, FSD, 로보택시, 옵티머스 등 자기 회사 용도로 돌리는 겁니다.
- 결과적으로 미래의 자체 수요 물량을 미리 확보해두면서, 그 건설 비용은 지금 급하게 컴퓨트를 찾는 다른 회사들의 임대료로 상당 부분 충당하는 구조라는 분석입니다.
2
0
0
63
@KyleReidhead @TSLAFanMtl Lot of delays and denials in several large scale projects
- could turn out to be longer than a couple years AND at higher rental prices
- could turn out to be longer than a couple years AND at higher rental prices
0
0
0
245
@KyleReidhead @TSLAFanMtl I was thinking the same thing. We are looking at a few years of high prices for compute and he is taking advantage of it and building out as fast as possible.
0
0
5
1,174
