OpenAI Unveils Jalapeño Chip, Surpassing Nvidia in AI Performance Benchmarks

For years, Nvidia has been the undisputed leader in AI hardware, especially for running large language models. Companies building advanced AI have relied on Nvidia's GPUs to power their services, but the high costs and supply constraints have become a growing concern.

OpenAI, best known for its language models, began developing its own custom chip to reduce reliance on outside suppliers. According to The New Stack, the project moved quickly, with OpenAI and Broadcom collaborating to design and build the chip in less than a year. The process also leveraged artificial intelligence to rewrite parts of the chip's code, an unusual step in hardware development.

On August 25, 2026, OpenAI publicly introduced the Jalapeño chip at the Hot Chips conference. Multiple outlets, including TechCrunch and The Next Web, reported that OpenAI shared benchmark results showing Jalapeño surpassing Nvidia's top Blackwell processors in both efficiency and latency. OpenAI invited third parties to benchmark the chip, signaling confidence in its performance.

SemiAnalysis, as cited by Techmeme, provided a detailed breakdown: Jalapeño beat not just Nvidia's chips but also those from AMD and Google on several open-source AI models. Digital Trends and Interesting Engineering highlighted how the chip delivered faster responses and better energy efficiency, with tests showing up to 1.9 times more efficiency than Nvidia's hardware.

As the news spread, industry observers noted the potential impact on Nvidia's market share. CNBC International and Bloomberg Technology both described Jalapeño as a direct threat to Nvidia's margins, especially as other tech giants like Google, AWS, and Meta are also working on custom silicon.

Forbes and other outlets emphasized the importance of Jalapeño's efficiency, pointing out that power consumption is now a limiting factor for large-scale AI deployments. Reports from August 29 and 30, including the Times of India and DEV Community, noted that while Jalapeño has not yet fully replaced Nvidia in OpenAI's operations, it could reshape purchasing strategies across the industry.

OpenAI plans to deploy Jalapeño by the end of 2026, according to several sources including DEV Community. The chip's strong benchmark results have heightened competitive tensions in the custom silicon market. OpenAI still relies on Nvidia and other partners for some AI computing needs, but the announcement marks a clear step toward more independence.

If Jalapeño performs as expected in real-world deployments, it could reduce costs and speed up services across OpenAI's platform. The chip's introduction has already prompted industry-wide discussion about the future of AI hardware and the growing trend of tech companies designing their own processors.

OpenAI must now prove that Jalapeño's benchmark results translate into real-world gains at scale. Industry analysts will be watching closely to see if other companies follow OpenAI's lead or if Nvidia responds with new advances of its own. The next year will be key in determining whether Jalapeño can shift the balance of power in the AI chip market.