Microsoft (MSFT) has unveiled a new microchip it is developing in-house that will be able to power its artificial intelligence (A.I.) and cloud-computing technologies.
The Seattle-based company announced its “Azure Maia AI Accelerator” chip, which is designed to run generative A.I. technologies, including large language models that are used to train A.I. systems.
Until now, Microsoft has largely relied on microchips supplied by Nvidia (NVDA) to power its A.I. and cloud applications. But now, Microsoft is taking chip development in-house.
Microsoft executives said that the company’s new A.I. chip is manufactured on a five-nanometer process and has 105 billion transistors. It will be rolled out in data centres in early 2024.
Initially, Microsoft’s new chip will be used to power the company’s Bing, Microsoft 365, and Azure OpenAI services.
Microsoft didn’t offer details on the microchip’s performance, saying it will do so once the chip is officially deployed.
Microsoft also introduced a data center computer processor, or CPU, named “Azure Cobalt” that is based on a design from British chipmaker Arm Holdings (ARM).
Microsoft isn’t the first technology company to develop its own microchips. Both Alphabet (GOOG/GOOGL) and Apple (AAPL) make their own microchips.
Currently, Nvidia’s “H100 GPU” microchip is the most popular processor for running A.I. applications. However, Nvidia recently announced a new “H200 Tensor Core GPU” chip that it says is 90% more powerful than the H100 chip and will be available in mid-2024.
Microsoft’s stock has gained 54% this year and currently trades at $369.67 U.S. per share.