OpenAI Unveils Custom AI Chip, Outperforming Nvidia’s Top Models
OpenAI has unveiled the first performance test results for its custom-designed inference chip, Jalapeño, which outperforms Nvidia's hardware in energy efficiency.
What happened?
At the Hot Chips conference, OpenAI presented the first performance results for its custom-designed inference chip, Jalapeño. The chip functions as a general accelerator for language models and is designed exclusively for AI inference, not training. According to the company's data, Jalapeño exceeds the performance of Nvidia's Blackwell and Rubin architectures in both throughput per watt and token latency.
Key facts
| Produktnamn | Jalapeño |
|---|---|
| Presentationsdatum | 25 augusti 2026 |
| Prestanda per watt | 1,5x–1,9x högre än konkurrenter |
| Interaktiv prestanda | 2,1x–4,1x högre prestanda |
Why it matters
Inference accounts for an increasingly large share of energy consumption and costs within global AI infrastructure. By developing proprietary hardware accelerators, OpenAI seeks to reduce its dependence on Nvidia's dominance in the chip market and lower operational costs for interactive AI services.
Who is affected?
The development primarily impacts cloud providers, AI companies, and data centres seeking more energy-efficient solutions for large-scale AI inference. Ultimately, users of AI services will benefit from reduced latency and faster response times.
Impact on the EU
The availability of the circuit boards in European data centres has not yet been confirmed, but more efficient inference chips could help lower the energy consumption of AI infrastructure within the EU.
What else you should know
The benchmark tests were presented by OpenAI at the Hot Chips conference. The results are based on the company's internal testing, and independent validation in large-scale production environments is still pending.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Kan Jalapeño användas för att träna AI-modeller?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Assess technical risk: model choice, vendor lock-in, data flow and running cost.
- Update the architecture doc if new APIs or regulations touch production.
- Ensure observability + rollback plan before rolling out to production.
Generated angle — not editorial analysis of "OpenAI Unveils Custom AI Chip, Outperforming Nvidia’s Top Mo"