OpenAI's Custom Chip Embarrasses Nvidia, While Company Vows to Keep Buying From It
OpenAI said Jalapeño delivered up to 4.1 times higher performance per watt on a public benchmark while it pledged to keep buying Nvidia chips.
- On Tuesday, August 25, 2026, OpenAI unveiled Jalapeño, its first custom AI inference chip developed with Broadcom to optimize performance for large language models.
- OpenAI designed Jalapeño to improve efficiency for large language model inference, balancing prefill and decode phases to deliver higher throughput per watt and lower latency than general-purpose GPUs.
- Jalapeño delivered 1.5 to 1.9 times higher peak throughput per kilowatt and 1.7 to 3.6 times lower end-to-end latency than Nvidia's Blackwell-generation GB200 and GB300 systems in benchmark tests.
- Despite unveiling the custom chip, OpenAI stated it will continue to widely deploy Nvidia accelerators for both training and inference, as meeting growing demand requires compute from multiple sources.
- Generation 2 is already deep in development and Generation 3 is taking shape, while analysts call Jalapeño a competitive threat to Nvidia's inference margins as other hyperscalers advance custom silicon.
16 Articles
16 Articles
OpenAI Jalapeno Beats Nvidia Blackwell in First Tests
OpenAI has published the first measured performance results for the OpenAI Jalapeno chip, its custom AI inference processor built with Broadcom, and the numbers show a clear lead over today’s leading Nvidia systems. Tested on the public SemiAnalysis InferenceX benchmark, Jalapeno delivered 1.5 to 1.9 times more AI work per watt at peak throughput and 1.7 to 3.6 times lower end-to-end latency, with even larger gains on interactive workloads. The …
ChatGPT has achieved a significant breakthrough, announcing its success in developing its own AI chipset. This is the company's first, and it aims to reduce its current dependence on Nvidia.
OpenAI's Custom Chip Embarrasses Nvidia, While Company Vows to Keep Buying From It
The post OpenAI’s Custom Chip Embarrasses Nvidia, While Company Vows to Keep Buying From It appeared first on 24/7 Wall St.. OpenAI published a newsroom post on August 25, 2026, announcing its first custom inference chip, called Jalapeño. On a public benchmark, a 700-watt part from OpenAI outperformed Nvidia (NASDAQ:NVDA) accelerators rated at 1,200 and 1,400 watts. The obvious read is that OpenAI has begun replacing its largest supplier. The do…
Coverage Details
Bias Distribution
- 40% of the sources lean Left, 40% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
