Broadcom Climbs as OpenAI Unveils Jalapeño AI Chip

Key Moments

Broadcom (NASDAQ:AVGO) shares advanced 3.4% Wednesday morning after OpenAI revealed the “Jalapeño” custom AI chip.
Jalapeño, described as OpenAI’s first “Intelligence Processor,” was taken from design to production in nine months for large language model inference.
OpenAI intends to roll out the chip at gigawatt scale with data center partners such as Microsoft starting in 2026, using Broadcom’s Tomahawk networking technology.

Market Reaction to the Jalapeño Announcement

Broadcom (NASDAQ:AVGO) traded sharply higher Wednesday morning, with the stock rising 3.4% after OpenAI introduced “Jalapeño,” a custom artificial intelligence processor created in collaboration with the semiconductor company.

OpenAI’s First Custom Intelligence Processor

Positioned as OpenAI’s first “Intelligence Processor,” Jalapeño is a purpose-built accelerator focused on large language model (LLM) inference. According to the announcement, the chip progressed from initial design to production in only nine months, signaling an aggressive execution timeline as OpenAI moves deeper into proprietary hardware as part of a broader full-stack infrastructure approach.

Chip Design, Partners, and Current Workloads

Jalapeño has been engineered from the ground up to meet OpenAI’s specific model requirements. The implementation, rack-level integration, and networking are being supported by Broadcom and Celestica. Engineering samples are already in use on active machine learning workloads in the lab, including OpenAI’s GPT-5.3-Codex-Spark, and are operating at power levels and frequencies targeted for production.

Aspect	Detail
Chip name	“Jalapeño”
Developer partnership	OpenAI and Broadcom, with Celestica for implementation
Primary use	Large language model (LLM) inference
Time from design to production	Nine months
Example workload	OpenAI’s GPT-5.3-Codex-Spark
Deployment timing	Planned starting in 2026

Performance and Architectural Advantages

Preliminary testing in the lab suggests that Jalapeño’s architecture provides a notable performance-per-watt benefit versus existing market offerings. The design emphasizes reduced data movement and a carefully tuned relationship between compute resources, memory, and networking to achieve these gains.

Planned Large-Scale Deployment

OpenAI expects to deploy Jalapeño at gigawatt scale through data center partners, including Microsoft, beginning in 2026. The large-scale rollout is set to utilize Broadcom’s Tomahawk networking silicon, which is intended to support the networking demands of massive AI infrastructure.

Strategic Outlook and Executive Commentary

The collaboration is being framed as a strategic step toward building out the physical backbone needed for future AI growth. Hock Tan, President and CEO of Broadcom, highlighted the long-term nature of the effort, stating, “Our collaboration with OpenAI represents a fundamental commitment to scaling the physical infrastructure required for the next decade of AI,” and added, “This is just the beginning of a multi-generation roadmap.”