“This May Be the Fastest ASIC Development Cycle Ever Achieved”
28 June, 2026
OpenAI and Broadcom unveiled Jalapeño, a new AI chip developed in just nine months—what may be the fastest development cycle ever achieved for an advanced ASIC
By Yohai Schwiger
OpenAI and Broadcom have unveiled Jalapeño, the first custom AI accelerator the two companies have developed together. The application-specific integrated circuit (ASIC) is designed for AI inference—the stage in which a trained model generates responses to users—and marks the first step in OpenAI’s strategy to build its own AI computing infrastructure and reduce its dependence on NVIDIA accelerators.
According to the companies, Jalapeño was designed specifically for the workloads of large language models and agentic AI applications. Rather than training models, the chip is optimized for inference, where most of OpenAI’s compute demand now resides. The company says early silicon has demonstrated significant improvements in performance per watt, a critical metric given the enormous cost of operating large-scale AI models.
For OpenAI, the project is part of a broader strategy to control every layer of its AI stack—from the models themselves to the hardware that runs them. As OpenAI President Greg Brockman explained, the company has “a deep understanding of our workloads,” allowing it to build hardware tailored specifically to its own AI infrastructure instead of relying solely on general-purpose accelerators.
The new hardware platform extends well beyond the chip itself. OpenAI said the project combines its own accelerator architecture with Broadcom’s expertise in ASIC implementation, networking and connectivity technologies, alongside Celestica’s capabilities in designing boards, racks and complete data center systems. The announcement signals that OpenAI’s objective is not merely to develop a single processor, but to establish a full computing platform for future generations of AI models.
Following tape-out, Jalapeño is now undergoing silicon validation and early testing ahead of deployment in AI servers built by Celestica and eventually integrated into OpenAI’s data center infrastructure. The company describes Jalapeño as “the first step in a multi-generation compute platform,” suggesting that additional generations of custom AI accelerators are already on the roadmap.
A New Era of Accelerated Chip Development?
Yet the most significant aspect of the announcement may not be the chip itself, but how quickly it was developed.
In their official announcement, OpenAI and Broadcom stated that Jalapeño was developed “from initial design to manufacturing tape-out in just nine months,” adding that the project “represents what may be the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors.”
For the semiconductor industry, that is an extraordinary claim. Developing an advanced chip typically takes between 18 and 24 months—and often longer—because of the complexity of architecture design, verification, optimization and manufacturing preparation. Completing the process in just nine months represents a dramatic reduction in development time.
The companies attribute this accelerated schedule to three key factors: close collaboration between the OpenAI and Broadcom engineering teams, Broadcom’s extensive experience in custom silicon development, and “the use of OpenAI models to accelerate parts of the design and optimization process.” While the companies have not disclosed exactly which engineering tasks were assisted by AI, this marks one of the first public acknowledgments by a leading AI developer that its own models were used to accelerate the development of advanced semiconductor hardware.
Industry analysts believe this could be the broader significance of the announcement. Until now, AI has primarily been associated with software development, content creation and data analysis. Jalapeño suggests that AI is beginning to play a role in designing the very chips on which future AI systems will run. As one post-announcement analysis argued, the real question is not whether Jalapeño is a successful chip, but whether nine-month development cycles could become the new standard for advanced semiconductor design.
If that proves to be the case, the implications could be far-reaching. Shorter development cycles would allow chipmakers to iterate more rapidly, update architectures more frequently and align hardware innovation with the relentless pace of AI model development. In other words, AI may soon accelerate not only software creation, but also the development of the semiconductor infrastructure on which it depends—potentially marking one of the most significant shifts in the chip industry in years.
Posted in: News
