Nvidia is facing delays in the delivery of its highly anticipated B200 AI graphics card for data centers, as reported by The Information citing insider sources. The company has reportedly encountered 'design problems' that have led to the postponement of the release.
While Nvidia had initially planned to launch the B200 chip later this year, the release has now been pushed back 'by at least three months,' according to two sources speaking to The Information. This delay means that large-scale deliveries may not occur until the first quarter of 2025 at the earliest. The decision to delay the release is attributed to 'design flaws discovered late in the production process,' although further details are not provided in the article. Nvidia is currently conducting test production runs to identify and address the underlying issues.
The company has reportedly informed its customers about the postponement this week, a move that could have implications for major tech players like Meta, Google, and Microsoft. These companies are said to have collectively ordered significant quantities of B200 GPUs worth tens of billions of dollars to support their advanced AI models. An Nvidia spokesperson confirmed to The Verge that production of the chip is still expected to ramp up in the latter half of 2024, but declined to comment further on the matter.
The B200 chip, which follows the earlier release of the Blackwell data center chip in March, boasts 208 billion transistors and comprises two interconnected GPU chiplets (B100). The AI-driven GPU is touted to deliver up to two and a half times the performance of its predecessor, the H100, while consuming significantly less energy. With the growing demand for AI applications, Nvidia briefly held the title of the world's most valuable company in June.
English (US)