Why everyone from OpenAI to SpaceX is building their own chips (and turning up the heat on Nvidia)
TL;DR
Nvidia still dominates AI accelerators, but its biggest customers are trying to reduce full dependence on one supplier. OpenAI is working with Broadcom on Jalapeño, a custom inference chip aimed at cheaper, more controlled model deployment. Google, Apple, and SpaceX show the same pattern: custom silicon is becoming a hedge against shortages, pricing power, and strategic lock-in.
Nauti's Take
The interesting part is not that Nvidia suddenly looks weak. The interesting part is that AI companies have learned a hard lesson: if you build models but buy compute like a commodity utility, you give up too much control.
Custom chips are expensive, slow, and risky, but at sufficient scale they become almost unavoidable. The PR frames this as revolution; in practice, it is a fight over leverage and margins.
Briefingshow
AI infrastructure is becoming a power game. Companies that build their own chips can control cost, supply, and product roadmaps more tightly, but only if they have massive scale, capital, and hardware talent. For Nvidia, near-term demand stays strong, while the long-term market becomes less comfortable.
Sources