OpenAI Unveils Jalapeño Inference Chip

OpenAI just unveiled Jalapeño, its first custom AI chip built with Broadcom, aiming to slash the cost and power needed to run massive language models like ChatGPT. This isn’t about training AI from scratch, but about making the day-to-day use—called inference—cheaper and faster for millions of users. If the chip delivers as promised, OpenAI will have a leg up on cost and speed just as it quietly prepares for an IPO, and could put serious pressure on rivals like Anthropic and Google.

But here’s the catch: pulling this off means orchestrating huge new data centers with Microsoft and convincing big enterprise buyers that OpenAI can handle both scale and cost control. OpenAI is already rolling out new spend controls and analytics for ChatGPT Enterprise, landing marquee customers like Samsung and the U.S. Department of Defense. The question isn’t just whether companies will use AI, but whether they can track—and justify—what they’re spending. Meanwhile, OpenAI’s new ad offerings and a fresh partnership with Getty Images show it’s chasing ad dollars hard, but as the assistant blends chat, coding, and ads, privacy and governance become hotter issues.

Featuring insights from Forrester, Gartner, Fortune, PYMNTS, and Getty Images.

Powered by Apisod.com