Why everyone from OpenAI to SpaceX is building their own chips (and turning up the heat on Nvidia)
TL;DR
OpenAI is working with Broadcom on Jalapeño, a custom inference chip, joining Google, Apple, and SpaceX in the push to reduce reliance on Nvidia. The point is not a clean break from Nvidia. It is a hedge against single-supplier risk around pricing, supply, priority access, and product roadmaps. Custom silicon lets companies tune hardware for specific workloads such as inference, internal AI models, device integration, or specialized data center setups.
Nauti's Take
The Nvidia moat is still deep, but the giants are already building bridges across it. For AI builders, optimizing only for today’s CUDA reality is too narrow.
The next efficiency fight runs through inference costs, custom hardware, and platform control.
Briefingshow
Nvidia still has a huge lead, but the balance changes when its biggest customers start designing chips themselves. Companies with enough volume, infrastructure, and predictable workloads can optimize cost, performance, and supply more directly. That will not automatically make AI cheap, but it weakens Nvidia’s pricing power and makes the chip market more strategic.
Sources