I’ve been going down a bit of a rabbit hole lately trying to build an AI-focused portfolio, but not just the usual NVDA/MSFT stuff everyone already talks about.
Instead I’ve been looking more at what actually enables AI to scale, like the stuff that becomes a bottleneck if demand keeps growing.
Right now I’m looking at a few areas:
**Memory / scaling side**
MU
SK Hynix
Samsung
My thinking here is pretty simple, AI workloads seem insanely memory-heavy (especially HBM), and if that keeps ramping, memory might end up being just as important as the GPUs themselves. Not sure if that’s already fully priced in though.
**Power / energy**
BE
This one is more of a “what if” idea. If data centers keep scaling the way people expect, power becomes a real constraint. Bloom seems interesting because of the on-site generation angle, but also feels pretty speculative tbh.
**Less obvious chip stuff**
VICR
From what I understand, as chips get more powerful, power delivery becomes a bigger issue too. VICR popped up a few times when I was digging into that, but I don’t have super high conviction yet.
Overall I’m not really trying to pick which AI company wins, just assuming demand keeps growing and trying to sit somewhere in the supply chain instead.
**Couple things I’m still unsure abou**t:
Is this “bottleneck” angle actually a good way to play AI, or am I just late to it?
Out of these, which ones are actually solid vs just good stories?
Are there any similar names you guys are looking at that aren’t the obvious mega caps?
Would be keen to hear what others are doing, especially if there are risks here I’m missing.