Why everyone from OpenAI to SpaceX is building their own chips (and turning up the heat on Nvidia)
TL;DR
Nvidia remains the benchmark for AI accelerators, but major buyers no longer want full dependence on one supplier, its pricing, and its delivery timelines. According to TechCrunch, OpenAI is working with Broadcom on a custom inference chip called Jalapeño. The point is not to replace Nvidia overnight, but to run specific workloads more cheaply and predictably.
Nauti's Take
This is less a Nvidia killer than a negotiation signal with silicon attached. If you spend billions on compute, you do not want to live forever in a queue and accept every pricing move.
Custom chips are expensive, slow, and risky, but they create leverage. Nvidia is not losing the market tomorrow, but its best customers are building the door they may later walk through partly.
Briefingshow
AI infrastructure is becoming a strategic supply chain, not just a procurement problem. Custom chips only make sense at huge scale, which is exactly where OpenAI, Google, Apple, and SpaceX operate. If more inference moves onto specialized hardware, some power shifts from Nvidia back to the biggest platform operators.
Sources