---
format: "aidr-story-markdown/v1"
id: "f00d4d12ad795b12a4a4b4ba132318bf1910d4a2e4c2c380d1942037db47a380"
canonical_url: "https://aidr.today/f00d4d12?lang=en"
title: "OpenAI Jalapeno Beats Nvidia Blackwell on Power and Speed to Lower Data Center Costs"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-25T14:52:31.000Z"
category: "Chips"
topics: ["openai","nvidia","chips","inference","infra"]
source_urls: ["https://huggingnews.com/ai/openai-jalapeno-beats-nvidia-blackwell-on-power-and-speed-to-lower-data-adb8fc62","https://x.com/EdLudlow/status/2092251479793184938","https://x.com/DeItaone/status/2092251884895527366","https://x.com/wallstengine/status/2092252412711215158","https://x.com/StockSavvyShay/status/2092252634057236891","https://news.ycombinator.com/item?id=49434378","https://newsletter.semianalysis.com/p/openai-jalapeno-better-than-nvidia"]
summary: "The AI startup announced that its new Broadcom-designed processor delivered faster response times and greater power efficiency than Nvidia's GB300 in internal tests for AI inference. Named Jalapeno, the chip runs at 700 watts and aims to reduce the infrastructure costs associated with deploying large-scale models. OpenAI plans to deploy the hardware later this year to support its own models. The company stated that the performance lead expanded on larger workloads, including Moonshot’s Kimi model, though Jalapeno was not tested against Nvidia’s upcoming Vera Rubin generation. A second-generation chip is currently approaching tape-out, and work has already begun on a third generation."
---

# OpenAI Jalapeno Beats Nvidia Blackwell on Power and Speed to Lower Data Center Costs

> [Open the canonical story](<https://aidr.today/f00d4d12?lang=en>)

**Published:** 2026-08-25T14:52:31.000Z
**Category:** Chips
**Topics:** openai, nvidia, chips, inference, infra

## Summary

The AI startup announced that its new Broadcom\-designed processor delivered faster response times and greater power efficiency than Nvidia's GB300 in internal tests for AI inference\. Named Jalapeno, the chip runs at 700 watts and aims to reduce the infrastructure costs associated with deploying large\-scale models\. OpenAI plans to deploy the hardware later this year to support its own models\. The company stated that the performance lead expanded on larger workloads, including Moonshot’s Kimi model, though Jalapeno was not tested against Nvidia’s upcoming Vera Rubin generation\. A second\-generation chip is currently approaching tape\-out, and work has already begun on a third generation\.

## Sources

- [Story source](<https://huggingnews.com/ai/openai-jalapeno-beats-nvidia-blackwell-on-power-and-speed-to-lower-data-adb8fc62>)
- [Story source](<https://x.com/EdLudlow/status/2092251479793184938>)
- [Supporting source](<https://x.com/DeItaone/status/2092251884895527366>)
- [Supporting source](<https://x.com/wallstengine/status/2092252412711215158>)
- [Supporting source](<https://x.com/StockSavvyShay/status/2092252634057236891>)
- [Discussion](<https://news.ycombinator.com/item?id=49434378>)
- [Story source](<https://newsletter.semianalysis.com/p/openai-jalapeno-better-than-nvidia>)