Microsoft is embarking on developing a cutting-edge network card aimed at bolstering the performance of its Maia AI server chip, potentially diminishing its dependence on chip designer NVIDIA, as reported by The Information.
According to sources familiar with the matter cited in the report, Microsoft CEO Satya Nadella has entrusted Pradeep Sindhu, co-founder of networking gear developer Juniper Networks, to lead the initiative. Sindhu’s involvement stems from Microsoft’s acquisition of his server chip startup, Fungible, last year.
The forthcoming network card resembles NVIDIA’s ConnectX-7 card, typically paired with its graphic processor units (GPUs), per the report’s details.
Although the development of the new equipment may extend beyond a year, its successful fruition could streamline the training of OpenAI models on Microsoft servers and render the process more cost-effective, offering significant advantages.
At the time of reporting, Microsoft had not responded to Reuters’ request for comment on the matter.
Microsoft’s substantial investments in OpenAI, the creator of ChatGPT, have been widely recognized, with the incorporation of OpenAI’s technology into numerous products providing the tech giant with a competitive edge in the artificial intelligence software market.
The introduction of Maia, a chip designed explicitly for powering large language models and facilitating AI computing, in November underscores Microsoft’s commitment to advancing AI capabilities and solidifying its position in the burgeoning AI landscape.
