Summary
On June 24, 2026, OpenAI and Broadcom jointly announced Jalapeno, OpenAI’s first custom AI chip — described as an “intelligence processor” purpose-built for inference. The chip was physically handed to OpenAI CEO Sam Altman and President Greg Brockman by Broadcom CEO Hock Tan, marking a historic shift for a company that has depended almost entirely on Nvidia GPUs since its founding. Unlike general-purpose chips, Jalapeno is designed specifically for the inference step — generating responses from a trained model — and targets the compounding latency problems that emerge when thinking models run inside agentic workflows.
OpenAI’s public claims include performance on par with Nvidia’s Blackwell chips and Google’s TPUs, roughly 50% lower infrastructure cost compared to standard GPU setups, and a development timeline of just nine months from first design to manufacturing tape-out — which Broadcom called potentially the fastest development cycle ever for a chip of this complexity. The video notes that full benchmark data and a technical report are still months away, so these figures remain unverified. The supply chain spans multiple industry partners: TSMC handles fabrication, Broadcom contributes silicon engineering and Tomahawk networking for inter-chip communication inside data centers, and Celestica builds the server boards and racks.
OpenAI also disclosed that its own AI models were used to accelerate parts of the chip design and optimization process. The first Jalapeno servers are expected online by end of 2026, with gigawatt-scale deployment planned alongside Microsoft. OpenAI describes this as the first step in a multi-generation chip platform aimed at controlling the full stack from model to silicon.
📺 Source: TheAIGRID · Published June 25, 2026
🏷️ Format: News Analysis
