Microsoft’s New Network Card Could Reduce Dependence on Nvidia

February 21, 2024 – In a move aimed at enhancing the performance of its in-house Maia AI server chips and potentially reducing its reliance on chip designer Nvidia, Microsoft is reportedly developing a new type of network card, according to sources cited by The Information.

The report states that Microsoft CEO Satya Nadella has tapped Pradeep Sindhu, co-founder of networking equipment developer Juniper Networks (JNPR.N), to lead the network card development project. Microsoft acquired Fungible, a server chip startup founded by Sindhu, last year.

The new network card is said to be similar to Nvidia’s ConnectX-7, which is sold alongside its graphics processing units (GPUs). Development of the device could take more than a year, but if successful, it could reduce the time and cost required for OpenAI to train models on Microsoft’s servers, according to the report.

Microsoft has previously invested billions of dollars in ChatGPT developer OpenAI and integrated its technology into various products, giving the company a competitive edge in selling AI software. Last November, Microsoft unveiled its Maia chip, specifically designed to run large language models and support AI computing.

Leave a Reply