OpenAI (AI Company) and Broadcom (Semiconductor Company) have jointly unveiled Jalapeño, an AI inference accelerator developed specifically for large language model workloads, marking OpenAI's first venture into custom silicon and the beginning of what the two companies describe as a multi-generation compute partnership targeting gigawatt-scale deployment.

A New Chip Purpose-Built for LLM Inference

Jalapeño is described by the companies as a blank-slate design for modern LLM inference rather than a general-purpose accelerator adapted from earlier AI workloads.

OpenAI says the chip was architected around its understanding of LLM fundamentals, drawing on the company's roadmap of models, kernels, serving systems, and product requirements. The chip is designed with flexibility to work with all LLMs guided by OpenAI's insights into the inference needs of current and future AI models across the industry.

Engineering samples of the chip are already running machine learning workloads in the lab at production target frequency and power, including GPT-5.3-Codex-Spark.

While OpenAI says it is still measuring final performance figures, early testing indicates that Jalapeño will deliver performance per watt substantially better than current state-of-the-art accelerators. The company says a detailed technical report on performance will be presented in the coming months.

The architecture is designed to reduce data movement and balance compute, memory, and networking resources to achieve realized utilization much closer to theoretical peak performance. Broadcom's silicon implementation and networking technologies, including Tomahawk networking silicon, are part of the platform's path to large-scale production.

Nine Months from Design to Tape-Out

One of the more striking claims surrounding the chip's development is the speed at which it was produced. Jalapeño was co-developed from initial design to manufacturing tape-out in just nine months, a timeline the companies say represents what they believe to be the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors.

OpenAI attributes that pace to deep software-hardware co-development with its engineering teams, Broadcom's silicon implementation expertise, and the use of OpenAI's own models to accelerate parts of the design and optimization process.

Partners Broadcom and Celestica helped industrialize the platform through chip implementation, board and rack system integration, high-performance networking, and scalable production systems.

The chip was formally delivered to OpenAI CEO Sam Altman and President Greg Brockman by Broadcom President and CEO Hock Tan and President Charlie Kawwas, an event OpenAI described as marking an important step in its strategy to build the full stack behind its models and products.

Full-Stack Strategy Moves Into Hardware

The launch represents a significant expansion of OpenAI's infrastructure ambitions. The company, long known as a model developer and consumer product provider, is now positioning itself as a hardware developer as well.

OpenAI frames this as a full-stack advantage: by operating across chip architecture, kernels, memory systems, networking, scheduling, deployment systems, and product experience, each layer can be optimized around the same goal of making its models faster, more reliable, and more affordable for users.

Greg Brockman, President and Co-Founder of OpenAI, framed the chip in broader economic terms.

"The world is moving to a compute-powered economy," he said. "Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI that is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems. By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access."

Richard Ho, who leads OpenAI's hardware program, emphasized that the chip reflects close collaboration between hardware and research teams. "We optimized the architecture around the kernels, memory movement, networking, and serving patterns that matter most for frontier AI models," Ho said. "Based on early testing, Jalapeño will efficiently execute our most important workloads close to the hardware's theoretical limits."

Gigawatt-Scale Deployment Planned with Microsoft and Others

Broadcom CEO Hock Tan characterized the partnership as a long-term infrastructure commitment rather than a single product effort. "This is just the beginning of a multi-generation roadmap," Tan said.

"By co-developing our industry-leading silicon directly with OpenAI, we are enabling the deployment of gigawatt-scale data centers with Microsoft and other partners beginning in 2026."

The chip is intended to be deployed at a gigawatt scale with data center partners over multiple generations, according to the announcement. The goal, as described by OpenAI, is to combine the power and throughput of today's leading AI accelerators with latency closer to the fastest specialized inference systems, making Jalapeño suited for interactive LLM products at scale. The use cases informing the design span ChatGPT, Codex, the API, and future agentic products.

Infrastructure as a Competitive Flywheel

OpenAI described the chip as a reinforcing mechanism for its broader business and research ambitions. Better infrastructure drives compute efficiency, the company said, which in turn enables better training and serving, ultimately powering more capable AI models.

Those models become better products for people, developers, and businesses, driving more usage, more customers, and more revenue, which allows reinvestment into the next generation of infrastructure. OpenAI framed this cycle as the mechanism by which it intends to make intelligence more capable, more reliable, and less expensive over time.

Track Every Semiconductor Megaproject Before Your Competitors Even Know It Exists

In an industry where fabrication facility announcements can shift supply chains overnight, the difference between winning and losing often comes down to who had the intelligence first. Semiconductor manufacturers, equipment suppliers, and construction firms operating across markets worldwide cannot afford to rely on fragmented sources and delayed reporting when billion-dollar opportunities are on the line.

The Global Project Tracking (GPT) platform by Blackridge Research consolidates the full lifecycle of semiconductor infrastructure projects into a single, continuously updated intelligence resource. From early-stage government incentive announcements to shovels hitting the ground on advanced node fabs, the platform gives your team the visibility needed to act with confidence and precision.

Whether you are mapping expansion activity across legacy chip plants or monitoring the latest advanced packaging developments across markets worldwide, the GPT platform structures complex project data into actionable intelligence your business development and strategy teams can use immediately.

• Upcoming Projects

• Tender Notices

• Contract Awards

• Projects Under Construction

• Completed Projects

See how semiconductor industry leaders are using the GPT platform to stay ahead of the project pipeline. Book a Free Demo with the Blackridge Research team today.