There's something very interesting brewing in the AI race.
You may have noticed a recent wave of "AI doomer-ism" hitting the main stream media; calls for slowdown, needs for safety, an "apocalyptic threat" being imminent.
At the same time, we have emergent capabilities; millennium problems being solved, the hugging face hack, new spatial robotics, biological breakthroughs (Alzheimer's progress!).
So, what's changed recently?
The first thing to understand is that our current AI has two components. The model, and the harness.
The model is the brain - think of it kind of like "IQ" - a measure of your potential intelligence. The model is the general infrastructure for thought, but hasn't actually learned anything. This translates to memory and compute (hardware).
The harness is the school. It's everything you put around the brain to make it useful. For us it's flashcards, teachers, lived experiences. For computers that's the training iterations they do.
For the past year+ we've seen frontier companies focusing on the harness. We essentially gave an average brain an extremely strict study routine with the best teachers. It got better at passing tests, but it's "IQ" never got higher.
Well, that's changing now. Frontier models have switched back to expanding the model instead of the harness (as hardware supply is coming back online).
The frontier labs are ready to expand the hardware again, and I think todays big chip rally was your first sign of it (In case Cap-ex staying high wasn't your first clue).
Other than SOXX and general hardware, what are your plays for focusing on the inference side of the AI trade vs. the software side.
I've made good money timing the previous hardware top and diversifying into software companies setup to benefit (bought both MSFT and META bottom in spades (400k+).
I already have the bulk of my hardware positions in AMAT, CHPX, CHAT, ASML, TSM, etc. Those are all known. I took a smaller position in POET (I see them coming back in the momentum trade with inference in focus). I'd like some more smaller names that may not be on the obvious radar; please no NBIS or other parroted names.