Key Highlights
- OpenAI partnered with Broadcom to create Jalapeño, a specialized AI chip optimized for inference operations.
- Testing data indicates cost reductions of approximately 50% versus standard AI graphics processors, according to Broadcom’s CEO.
- Development completed in nine months with TSMC manufacturing; rollout expected by late 2026.
- Following the reveal, Broadcom shares gained approximately 2%, while Nvidia experienced a minor decline.
- A multi-generation development strategy exists, with the successor chip scheduled for 2028.
In a significant announcement Wednesday, OpenAI revealed its collaboration with Broadcom on a purpose-built AI processor dubbed Jalapeño. The news propelled Broadcom (AVGO) shares upward by roughly 2%, while Nvidia (NVDA) experienced a slight 0.26% decline.
Jalapeño is engineered exclusively for AI inference tasks—the computational process behind generating outputs in applications such as ChatGPT. As an ASIC (application-specific integrated circuit), it sacrifices versatility for enhanced cost efficiency and optimization for targeted operations.
According to Broadcom CEO Hock Tan, preliminary testing demonstrates approximately 50% cost reductions when compared with conventional AI GPUs. This figure carries substantial weight in a sector where computational expenses represent a persistent challenge.
OpenAI confirmed that prototype units are currently operational in its facilities, meeting targeted performance and power specifications during evaluation with its GPT-5.3-Codex-Spark architecture.
The design phase, completed by OpenAI’s engineering team in roughly nine months with assistance from AI-powered tools, was subsequently handed to TSMC for production.
Richard Ho, OpenAI’s hardware chief, characterized Jalapeño as “a very general purpose device” optimized for large language models while engineered to “address future LLM innovations.”
Tan positioned Jalapeño as competitive with Nvidia’s Blackwell series and Google’s tensor processing units—currently among the industry’s most capable AI accelerators.
Design Philosophy and Purpose
Greg Brockman, OpenAI’s President, positioned the chip within a larger infrastructure initiative. “By designing more of the stack ourselves, we can serve more intelligence with greater efficiency,” he explained.
OpenAI emphasized that Jalapeño represents the inaugural chip in a long-term computing platform. The subsequent generation is slated for 2028, with yearly releases anticipated thereafter.
Production units will be deployed across data centers operated by Microsoft and additional partners. Server system assembly will be handled by Canadian electronics firm Celestica.
Industry-Wide Silicon Strategy
As one of Nvidia’s largest customers, OpenAI faces fierce competition across the AI sector for chip allocation. Creating proprietary silicon provides an alternative pathway for computational resources.
Major tech companies including Amazon, Google, and Microsoft are developing or actively utilizing custom AI processors. Meta has similarly invested in proprietary chip development and deployment. Several providers, notably Amazon and Google, now offer their custom processors to external clients.
The OpenAI-Broadcom chip collaboration was first disclosed in October. Initial projections called for silicon requiring 10 gigawatts of power capacity.
During Wednesday’s announcement, Tan suggested demand has intensified sufficiently that his previous estimate of 1.3 gigawatts in chip deployments for next year “may prove conservative.” He remarked to Bloomberg: “We like to think we can do better because there is a lot of demand.”
Nvidia competitor AMD continues pursuing market share expansion in AI data center infrastructure. Additional players including Qualcomm and Cerebras are similarly advancing into this territory.
OpenAI and Broadcom are targeting initial Jalapeño deployment before 2026 concludes.


