NVIDIA Discusses Custom AI Chips with Microsoft, Meta, Amazon

According to reports from nine related companies, NVIDIA in the United States is establishing a new business unit with a focus on designing custom chips, including advanced AI processors, for cloud computing companies and other companies. NVIDIA’s goal is to gain a certain market share in the future growing custom AI chip market and protect itself from the influence of more and more companies interested in seeking alternatives to its products.
According to reports, NVIDIA currently controls about 80% of the high-end AI chip market, and in 2023, NVIDIA’s market share in the high-end AI chip market has more than doubled. It is expected to grow by another 40% in 2024, reaching $1.73 trillion. NVIDIA’s clients include ChatGPT creator OpenAI, Microsoft, Google’s parent company Alphabet, and Meta.

Although NVIDIA can provide multi-purpose AI chips for large companies such as H100 and A100 chips for large model training, it is clear that these tech giants have started developing their own internal chips to meet specific needs. This is obviously not good news for NVIDIA, but for large companies, whether it is to reduce costs, reduce dependence on Nvidia, and improve bargaining power, self-developed AI chips have become a consensus.
Sources have revealed that NVIDIA is attempting to assist these companies in developing customized AI chips, which have flowed into competitors such as Broadcom and Marvell Technology. Industry insiders said in an interview, “If you really want to optimize performance such as power consumption or optimize application costs, then you cannot directly use H100 or A100. You will want to have the correct combination of calculations and the required types of calculations.” This may also be the driving force behind many large companies making choices.
According to sources familiar with the meeting, NVIDIA executives met with representatives from Amazon, Meta, Microsoft, Google, and OpenAI to discuss creating customized chip brackets for them. In addition to data center chips, NVIDIA is also seeking to develop customers in fields such as telecommunications, automotive, and video gaming. NVIDIA stated in 2022 that it will allow third-party customers to integrate some of its home network technology into their own chips.