According to the publication Information, Microsoft, is engaged in the development of its own chips to perform complex tasks of artificial intelligence, which will be used to teach large language models. This will help to avoid dependence on NVIDIA, which currently acts as a key supplier of server and chips.
Work on chips began back in 2019, and some employees of Microsoft and Openai already have access to them to check their effectiveness in processing the latest language models, including GPT-4.
The creation of its chips will allow Microsoft to reduce the costs of promoting AI, as the corporation continues to develop functional capabilities based on artificial intelligence in products such as Bing, Office, Github and other services. Although their AI chips will not be able to completely replace the NVIDIA solutions, this will make it possible to significantly reduce costs.
Microsoft plans to make its AI-chips more accessible to its structures and Openai in 2024, and the company has a roadmap for the development of several future generations of processors. It is worth noting that Amazon, Google and Meta already have their AI chips, but many companies are still relying on NVIDIA solutions to support the latest language models.