Leopold Aschenbrenner appears to have become aggressively long CLSK as part of a broader “AI infrastructure bottleneck” thesis centered on power availability and data-center capacity.
Source:
Fund Update: 10,635,739 CLEANSPARK (CLSK) shares added to Situational Awareness LP portfolio
Paper trading today resulted in a crazy 5.54% return. Win rates 7/9 trades today and 7/8 days. Currently using an API linked to Alpaca. Once I'm confident in the results and the allocation is optimized, it will be time to turn on the "money printer". XD
​
At a certain scale, markets seem less about isolated decisions and more about process quality.
Reading about Otonomii AI and the limited beta made me think more about how institutional workflows might be evolving.
I’m trying to figure out whether AI infra is still one of the best long-term themes in 2026, or if the market is starting to price in slowing growth / margin compression.
Names like ARM, Fabrinet, and Teradyne all have real AI exposure:
ARM → AI CPUs / power-efficient compute
Fabrinet → optical networking / photonics
Teradyne → AI chip testing
But despite strong earnings and AI demand, these stocks have been selling off hard lately.
My current take:
AI demand itself still seems very real
Inference + agentic AI probably increases infrastructure demand long term
But the market may now care more about durable margins and bottlenecks instead of “anything AI”
It feels like we’re moving from:
“AI buildout euphoria”
to
“Which companies actually keep pricing power?”
I’m wondering:
Are these pullbacks opportunities?
Or are these names becoming cyclical semiconductor/hardware plays again?
Which AI infra layers do you think still have the best long-term economics?
Is software/observability becoming a better AI bet than hardware now?
Curious what people here think, especially anyone following hyperscaler capex, networking, or inference trends closely.