You're underestimating the cost of compute. This bubble will pop and likely take the market with it (oracle, coreweave, equinix, nvidia), ai capex is deranged. Margins may make sense eventually, but the infrastructure (chips, energy, data centers) needed for compute will remain a monopoly. Its a dotcom replay, and companies that survive (like Amazon) will dominate. Apple is prob in the best position, having not been duped by manic ai investment.
Replies (3)
Apple silicon/closed architecture primed for agentic, which 95% of ai compute is used for.
Intel is now on par with Apple chips. Zig and zag between AMD, Intel and Apple.
That said Google is beating everyone on efficiency with CPUs.
In that world compute hardware becomes important, cost of energy becomes important, data becomes important.
Somewhat agree, though maybe more than the hardware, is the model creation/training. Not sure if there is a good way to fully open-source that though. I think that's what is really creating the lock-in, because as you note, Apple already has a reasonably good localizing solution (though still rather pricy for the plebs).
