OpenAI just built its own chip. Broadcom says it cuts inference costs 50%. Your Nvidia contract is now negotiable.

OpenAI and Broadcom just unveiled Jalapeño.

OpenAI's first custom inference chip. Built from scratch in 9 months. The fastest ASIC development cycle in semiconductor history.

Broadcom CEO Hock Tan told Bloomberg the chip delivers 50% cost savings versus typical GPUs.

Read that again. Half the cost to run the same models.

This is not a side project. OpenAI designed the chip around its own LLM architecture. Broadcom handled silicon and networking. Celestica built the racks. Engineering samples are already running GPT-5.3-Codex-Spark workloads in the lab.

The target: 10 gigawatts of custom compute by 2029. That is ten nuclear reactors worth of AI inference capacity. Initial deployment starts at Microsoft and other partners by end of 2026.

Here is what this changes for your enterprise.

Every AI budget projection based on current GPU pricing is now wrong. If inference costs drop 50%, the ROI math on every AI project you shelved because it was too expensive just shifted.

Your cloud provider negotiations have a new lever. OpenAI is no longer dependent on Nvidia for inference. That dependency was pricing power. Now it is leverage for every enterprise negotiating compute contracts.

And the vendor landscape just fractured. OpenAI is becoming a full-stack AI company. Models. Chips. Infrastructure. If your AI strategy assumes your model vendor stays in one lane, throw it out.

Audit your inference cost assumptions this quarter. The pricing floor for enterprise AI is about to drop, and the companies that locked in GPU rates at today's prices just overpaid.

SOURCE: https://www.cnbc.com/2026/06/24/openai-and-broadcom-reveal-jalapeno-first-ai-chip-in-partnership.html
VERIFIED: CNBC (Kif Leswing), Financial Post (Bloomberg), OpenAI press release, The Next Web
SIGNAL: OpenAI going vertical on silicon reshapes compute economics for every enterprise running their models. 50% cost reduction on inference changes the ROI calculus for every AI deployment.