---
title: "OpenAI's Jalapeño Chip Built for Fast Inference at Scale, Benchmarks Show"
source: "TechCrunch：AI（RSS）"
source_tier: "t2"
url: "https://techcrunch.com/2026/08/25/openais-jalapeno-chip-is-built-for-fast-inference-at-scale-benchmarks-show/"
permalink: "https://feed.astralor.com/items/itm_d1df00f6f18c79cdd6fbadda"
published: "2026-08-25T14:22:04.000Z"
score: 74
channel: "tech"
tags: ["AI inference","Jalapeño","OpenAI","SemiAnalysis","Benchmark"]
---

# OpenAI's Jalapeño Chip Built for Fast Inference at Scale, Benchmarks Show

## AI commentary

> The benchmark numbers for Jalapeño are impressive, but what matters more is why OpenAI chose to release such a signal at this moment. For anyone tracking AI compute costs and chip supply chains, this is a lead that must be followed.

## At a glance

In SemiAnalysis' InferenceX benchmark, OpenAI's self-developed Jalapeño chip surpassed current state-of-the-art AI chips in both tokens per user and throughput per kilowatt. Optimized for large-scale inference workloads, the chip demonstrates superior energy efficiency and concurrency. Although OpenAI has not officially announced it, the data suggests significant competitiveness in data center deployments, potentially reshaping the AI chip market landscape.

## Key points

- Jalapeño delivers more tokens per user and higher throughput per kilowatt on the InferenceX benchmark
- Chip is optimized for large-scale AI inference, emphasizing energy efficiency and concurrency
- Benchmark data comes from SemiAnalysis and has not been officially confirmed by OpenAI
- The chip could become key for OpenAI to cut inference costs and scale compute

## Full text

Tested on SemiAnalysis’ InferenceX benchmark, Jalapeño registered both more tokens per user and more throughput per kilowatt than the currently available state-of-the art.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_d1df00f6f18c79cdd6fbadda