OpenAI and Broadcom unveiled Jalapeño on June 24, 2026, a custom AI inference chip designed specifically for accelerating large language model inference workloads [1, 2, 3, 4, 5, 6, 7, 8]. The chip is an application-specific integrated circuit (ASIC) built solely for inference, not for training AI models [2, 3, 4, 8].
Jalapeño's design and manufacturing tape-out took about nine months, one of the fastest ASIC development cycles reported [1, 2, 3, 6, 7, 8]. Broadcom leads chip manufacturing, with Canadian company Celestica assisting on server board and system integration [1, 6, 7].
In early lab tests, OpenAI found Jalapeño delivers substantially better performance per watt compared to current state-of-the-art AI chips [1, 3, 4, 7, 8]. The architecture balances compute, memory, and networking capabilities while reducing data movement to boost efficiency [6, 7]. Richard Ho, OpenAI's hardware chief, said, "We optimized the architecture around the kernels, memory movement, networking and serving patterns that matter most for frontier AI models" [7]. He also stated, "It will be performant on, we think, all kind of future iterations of LLMs" [1].
Broadcom CEO Hock Tan praised the chip's quality, saying, "The chip made by the team is as good as the Blackwell chips made by Nvidia or the tensor processing units designed by Alphabet's Google" [1]. Despite this, OpenAI intends Jalapeño to complement Nvidia GPUs rather than replace them, especially since GPUs handle training workloads [3, 4, 5]. The chip aims to cut inference costs by roughly 50% compared with typical AI GPUs [5].
Greg Brockman, OpenAI’s president, explained the motivation: "We have a deep understanding of the workload. We’ve really been looking for specific workloads that are underserved, [and asking] how can we build something that will be able to accelerate what’s possible?" He added, "By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access" [2, 4].
OpenAI and Broadcom publicly announced their collaboration on custom AI chips in October 2025, following about 18 months of development work before unveiling Jalapeño today [2, 4, 5, 8]. OpenAI plans to deploy the Jalapeño chips in its data centers starting in late 2026 as part of a multi-generation compute platform encompassing hardware, kernels, memory, networking, and deployment systems [1, 2, 3, 4, 5, 6, 8].
Broadcom has set up a chip financing vehicle with Apollo and Blackstone to support OpenAI's chip orders, while OpenAI raised $122 billion earlier in 2026 to finance investments in chips, data centers, and talent [5]. The next step will be rolling out Jalapeño chips in production data centers later this year to boost AI inference capacity at lower costs [1, 2, 3, 5, 8].
