OpenAI Unveils ‘Jalapeño,’ Its First Custom AI Chip — A Direct Shot at Nvidia
OpenAI introduced Jalapeño, its first in-house AI chip, co-developed with Broadcom and built for inference — a direct move to cut its reliance on Nvidia.
TL;DR — OpenAI unveiled Jalapeño, its first custom AI chip — co-developed with Broadcom and built for inference — claiming a record ~9-month design cycle and better performance-per-watt, in a direct move to reduce Nvidia dependence.
OpenAI is no longer just renting compute — it is designing it. On June 24, 2026, the company introduced Jalapeño, its first in-house AI chip, built with Broadcom.
The details
Jalapeño is OpenAI’s first custom silicon, designed in-house and manufactured with Broadcom. OpenAI says it is purpose-built for inference — running its models efficiently at scale — and that, in early testing, it delivers better performance-per-watt than the current state of the art. The company claims a roughly 9-month design cycle, which it calls one of the fastest ever for a high-performance AI ASIC.
The chip builds on OpenAI and Broadcom’s October 2025 agreement to develop custom accelerator racks totaling 10 gigawatts of compute capacity, with rollout beginning later in 2026.
The custom-silicon race
OpenAI joins every other hyperscaler in building its own AI chips to cut reliance on Nvidia.
| Company | Custom AI chip | Partner |
|---|---|---|
| OpenAI | Jalapeño (new) | Broadcom |
| TPU | Broadcom | |
| Amazon | Trainium / Inferentia | Annapurna (in-house) |
| Microsoft | Maia | in-house |
| Meta | MTIA | in-house |
What they said
"Jalapeño is part of our long-term, full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, and more affordable for people and businesses." — Greg Brockman, President, OpenAI
Why it matters
- Inference is where the volume is. Optimizing for inference targets the fastest-growing, most cost-sensitive part of AI workloads.
- Nvidia dependence is the strategic risk. Custom silicon gives OpenAI leverage on price, supply, and roadmap.
- Speed is the flex. A ~9-month ASIC cycle, if it holds, compresses a process that usually takes years.
FAQ
What is OpenAI’s Jalapeño chip?
Jalapeño is OpenAI’s first custom AI chip, co-developed with Broadcom and optimized for inference — running OpenAI’s models efficiently. OpenAI says early testing shows better performance-per-watt than current state-of-the-art chips.
Is OpenAI dropping Nvidia?
No. Jalapeño complements OpenAI’s broader compute strategy rather than replacing it, but it reduces reliance on any single supplier and gives OpenAI more control over cost and supply — joining Google, Amazon, Microsoft and Meta in building in-house AI silicon.
Sources
- TechCrunch — OpenAI unveils its first custom chip built by Broadcom
- Yahoo Finance — OpenAI and Broadcom announce first custom AI chip
Image: “1515 Third Street” by Coolcaesar, CC BY 4.0, via Wikimedia Commons.
← Back to all posts