OpenAI introduced a specialized chip Jalapeño for running LLM

Company OpenAI officially introduced its first specialized processor, named Jalapeño. The announcement took place on June 24, 2026, marking the developer's transition into designing its own hardware for high-performance computing.

The key feature of the project was the implementation timeline: the full cycle of chip development from concept to a ready engineering sample took only 9 months. For comparison, standard design cycles for similar semiconductor solutions in the industry usually take around 3 years. Such a high development speed was made possible by optimizing design processes and integrating the company's internal AI tools.

The Jalapeño processor is designed specifically to accelerate inference of large language models. The chip's architecture is focused on minimizing latency when processing tokens and increasing energy efficiency when performing matrix operations, which is critical for scaling the company's cloud services.

The release of its own processor allows OpenAI to reduce its dependence on third-party computing power providers and optimize the operating costs of its models. The transition to specialized equipment is part of the strategy to create a vertically integrated infrastructure capable of supporting future generations of neural networks with increased requirements for computing resources.