As AI clusters get larger, the challenge isn’t only compute. It’s also moving data between chips quickly enough. That’s why I’m watching four layers: • optical components • networking • interconnect silicon • rack connectivity The opportunity gets more interesting if bandwidth becomes one of the next major AI bottlenecks. https://www.instagram.com/stockswithpawan/p/DeR6W37DrX3/ Not financial advice. DYOR.
636 views
1 Comments
Nate
@hoodnate · 13d
I never thought about the data moving between chips being a bottleneck, I always just focused on the chips themselves. This makes me want to look into how much of AI spending actually goes into networking.
See the full comment section 👀Sign up for the full Blossom experience!
