OpenAI Unveils Jalapeño, Its First Custom AI Chip

OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom AI accelerator, a processor built for large language model inference that the two companies took from initial design to manufacturing tape-out in nine months.

Broadcom President and CEO Hock Tan and President Charlie Kawwas delivered the chip to OpenAI CEO Sam Altman and President Greg Brockman. Engineering samples are already running ML workloads at production target frequency and power, with GPT-5.3-Codex-Spark among the workloads tested. The design prioritizes efficient data movement and resource balance to achieve realized utilization much closer to theoretical peak. Early testing shows Jalapeño delivers performance per watt substantially better than the current state-of-the-art, and a detailed technical report is expected in the months ahead.

The chip was not adapted from a general-purpose accelerator. OpenAI designed it from scratch around LLM inference, informed by the serving patterns behind ChatGPT and Codex.

"The world is moving to a compute-powered economy," said Greg Brockman, President and Co-Founder of OpenAI. "Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI which 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, described what the design team prioritized.

"Jalapeño was designed from the ground up for LLM inference using detailed insights from our close collaboration with OpenAI researchers," Ho said. "We optimized the architecture around the kernels, memory movement, networking, and serving patterns that matter most for frontier AI models. Based on early testing, Jalapeño will efficiently execute our most important workloads close to the hardware's theoretical limits."

Nine months to tape-out

Custom AI accelerators typically require substantially longer development cycles than the one Jalapeño went through. OpenAI said it completed the process from initial design to tape-out in nine months, a timeline the company believes to be the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors.

That pace relied on deep software-hardware co-development and on the use of OpenAI's own AI models during the chip design and optimization process. Broadcom handled silicon implementation and contributed Tomahawk networking silicon. Celestica managed board and rack system integration.

"Our collaboration with OpenAI represents a fundamental commitment to scaling the physical infrastructure required for the next decade of AI," said Hock Tan, President and CEO of Broadcom. "This is just the beginning of a multi-generation roadmap. 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."

Jalapeño is targeted for initial deployment by the end of 2026 at gigawatt scale with data center partners. The launch places OpenAI in more direct competition with Nvidia, whose graphics processors currently dominate AI training and inference across the industry. OpenAI did not position Jalapeño as a replacement for all external chips but said proprietary silicon would give it greater control over its compute infrastructure.

Enterprise deals and inference demand

Two enterprise agreements announced near the time of the chip reveal give context to the inference demand OpenAI now manages.

OpenAI reached a multi-year agreement with BBVA to deploy ChatGPT Enterprise across the bank's full workforce of 120,000 staff in 25 countries, expanded from 11,000 employees. OpenAI described it as one of the largest generative AI deployments in financial services. The day before that announcement, Visa revealed a strategic partnership with OpenAI for its agentic commerce initiative. According to Visa, the partnership enables secure Visa payment capabilities inside AI-powered shopping experiences.

IPO signals and Amazon's film decision

CEO Sam Altman has indicated OpenAI could pursue an initial public offering within the next year. Coinbase introduced pre-IPO futures tied to OpenAI, with contracts that allow traders to take a position on the company's valuation before any formal market debut.

Amazon withdrew from distributing "Artificial," a film centered on Altman and featuring Elon Musk. Discussions with filmmakers about another distributor continued. The withdrawal came as Amazon deepened its commercial relationship with OpenAI through a multi-billion-dollar investment commitment tied to future milestones.

At the Cannes Lions International Festival of Creativity, OpenAI demonstrated Codex to marketers and promoted advertising opportunities tied to ChatGPT.

Meta Builds Arena App to Rival Prediction Markets | HODL FM NEWS

Meta is developing the Arena app, a standalone prediction market app inspired by Polymarket and Kalshi, as user behavior shifts and competition intensifies.

HODL FM News: The Wildest Ride in the Crypto MarketGunel Ismayilova

Disclaimer: All materials on this site are for informational purposes only. None of the material should be interpreted as investment advice. Please note that, despite the nature of much of the material created and hosted on this website, HODL FM operates as a media and informational platform, not a provider of financial advisory services. The opinions of authors and other contributors are their own and should not be taken as financial advice. If you require advice, HODL FM strongly recommends contacting a qualified industry professional.