Microsoft is reportedly venturing into the development of a new network card that aims to enhance the performance of its Maia AI server chip, potentially reducing the company’s dependence on Nvidia, according to a report from The Information on Tuesday.
To lead this effort, Microsoft CEO Satya Nadella has reportedly enlisted Pradeep Sindhu, who co-founded Juniper Networks and currently heads Fungible, a server chip startup that Microsoft acquired last year.
The envisioned network card is said to share similarities with Nvidia’s ConnectX-7 card, which the chip manufacturer sells alongside its graphic processor units. This move by Microsoft signifies a strategic push to gain more control over its hardware components and potentially decrease reliance on external chip designers.
The development of this new network card is expected to be a complex process, likely taking over a year to materialize. If successful, it could have significant implications for Microsoft, including reducing the time required for OpenAI to train its models on Microsoft servers and potentially lowering associated costs.
Microsoft, which has invested significantly in OpenAI, has incorporated the organization’s technology into various products. This move into hardware development aligns with Microsoft’s broader strategy to have more control over the technology stack and enhance its capabilities in artificial intelligence.
The introduction of Maia, a chip specifically designed to run large language models and support AI computing, in November marked a significant step for Microsoft in the field of artificial intelligence hardware.
