OpenAI's Jalapeño Chip Outperforms Nvidia, AMD, and Google Chips in Key Benchmarks

OpenAI has developed its own custom-designed 'Jalapeño' inference silicon, which outperforms Nvidia's Blackwell systems in throughput per watt and latency on industry benchmarks. The company achieved this breakthrough by leveraging its generative AI models to accelerate hardware engineering, reducing design-to-tapeout timelines to just nine months.

In standard test suites, Jalapeño delivered up to 1.9 times more output per watt and cut end-to-end latency by up to 3.6 times compared to Nvidia's Blackwell. OpenAI will continue purchasing accelerators from Nvidia and other third-party partners while establishing its custom silicon foundation.

The company unveiled the first live performance benchmarks for Jalapeño, demonstrating marked advantages in speed and energy efficiency over Nvidia's flagship architecture. SemiAnalysis reported that Jalapeño excels not only in low-latency scenarios but also in high-throughput scenarios without being tuned for any specific point in the curve.