On Wednesday, OpenAI and Broadcom revealed their jointly engineered custom AI processor named Jalapeño. Following the announcement, Broadcom (AVGO) stock increased roughly 2%, while Nvidia (NVDA) declined 0.26%.
The processor is purpose-built for AI inference workloads—the computational process behind generating outputs in applications like ChatGPT. It falls under the ASIC category, sacrificing versatility for reduced costs and optimization for dedicated tasks.
According to Broadcom CEO Hock Tan, preliminary samples demonstrate cost efficiencies of approximately 50% when compared to standard AI GPUs. This figure carries significant weight in a sector where computational expenses represent an ongoing challenge.
OpenAI confirmed that sample chips are currently operating in their facilities, meeting target specifications for power consumption and performance when evaluated with their GPT-5.3-Codex-Spark model.
The chip design was finalized by OpenAI’s engineering team in roughly nine months, with assistance from AI-powered design tools. Manufacturing was subsequently assigned to TSMC.
Richard Ho, OpenAI’s hardware chief, characterized Jalapeño as “a very general purpose device” engineered with large language models as the primary focus, yet constructed to “address future LLM innovations.”
Tan positioned Jalapeño alongside Nvidia’s Blackwell series and Google’s tensor processing units—both representing leading AI accelerators in current production use.
Greg Brockman, OpenAI’s President, positioned the chip as an element of a comprehensive infrastructure strategy. “By designing more of the stack ourselves, we can serve more intelligence with greater efficiency,” he stated.
OpenAI emphasized this represents the inaugural chip in a multi-generational computing platform. The follow-up version is scheduled for 2028, with yearly releases planned thereafter.
Completed chips will be deployed within data centers operated by Microsoft and additional partners. Canadian electronics firm Celestica has been selected to manufacture the server infrastructure.
OpenAI ranks among the largest purchasers of Nvidia processors, yet must compete across the entire AI sector for supply allocation. Creating proprietary silicon provides the organization with an alternative compute resource pathway.
Amazon, Google, and Microsoft are all developing or already utilizing custom AI chips. Meta similarly designs and deploys its own processors. Several companies, including Amazon and Google, have begun offering these chips to external clients.
OpenAI and Broadcom first disclosed their chip collaboration in October. Initial estimates projected sufficient silicon to consume 10 gigawatts of power.
During Wednesday’s announcement, Tan indicated demand has expanded to the point where his previous forecast of 1.3 gigawatts of chip deployments next year “may prove conservative.” He informed Bloomberg: “We like to think we can do better because there is a lot of demand.”
Nvidia competitor AMD is similarly pursuing market share expansion in AI data center infrastructure. Additional firms like Qualcomm and Cerebras are entering the competitive landscape.
Initial large-scale deployment of Jalapeño processors is scheduled for completion before 2026 ends.
The post OpenAI and Broadcom (AVGO) Unveil ‘Jalapeño’ AI Chip to Challenge Nvidia (NVDA) Dominance appeared first on Blockonomi.
