Why everyone from OpenAI to SpaceX is building their own chips (and turning up the heat on Nvidia)
TL;DR
Nvidia has dominated AI chips for years, but major customers are trying to reduce dependence on a single supplier. OpenAI is working with Broadcom on a custom inference chip called Jalapeño. Google, Apple, and SpaceX are pursuing similar custom silicon paths. This is not a clean break from Nvidia. It is a hedge: more control, more predictable costs, and chips tuned to specific workloads.
Nauti's Take
This is the normal maturation of a market that has become too expensive to outsource completely. Nvidia remains the benchmark, but the largest AI companies do not want to stay tied forever to one supplier’s GPU availability, margins, and roadmap.
The PR layer is heavy because custom chips are easier to announce than to run at scale. The real test is whether Jalapeño and similar efforts actually lower cost per token or mainly serve as strategic leverage.
Briefingshow
AI compute has become strategic infrastructure, not just procurement. Companies running large models do not want pricing, supply, and performance dictated entirely by Nvidia’s roadmap. Custom chips can lower long-term costs, but they also lock companies into expensive hardware bets that only make sense at massive scale.
Sources