To fuel AI initiatives for OpenAI and its own internal teams, the software giant is apparently constructing its own machine learning hardware.
According to a story from The Information, Microsoft has been quietly creating its own artificial intelligence (AI) chips to combat the escalating expenses of research for internal and OpenAI initiatives. Microsoft's recently unveiled hardware project, which has reportedly been in the works since 2019, seems to be intended to lessen the Redmond, Washington company's dependency on Nvidia's GPUs.
In light of the growing market shortage, the Nvidia H100, one of the more well-liked GPUs for training machine learning systems, can cost as much as $40,000 on reseller services like eBay. Due to these exorbitant prices, numerous Big Tech firms have been forced to create their own hardware, with Meta, Google, and Amazon all creating machine-learning processors in recent years. Microsoft hasn't yet made any public comments, so details are still limited, but according to The Information, the chips are being developed under the code name "Athena"—possibly a tribute to the Greek goddess of war as the generative AI arms race heats up. The article also states that developers from OpenAI and Microsoft's internal machine learning team are currently testing the new processors.
Sam Altman, co-founder and CEO of OpenAI, recently informed a crowd at the Massachusetts Institute of Technology that the infrastructure and design that led the company from GPT-1 to GPT-4 is "played out" and will need to be rethought. While it is currently only speculative how OpenAI intends to use Microsoft's AI chips: This follows a busy news cycle for the AI industry, in which Amazon just entered as a (relatively) new contender with the launch of its first in-house models as part of the Bedrock AI infrastructure rollout.
Then on April 17, during an interview with Fox News' Tucker Carlson, internet tycoon and world's richest man Elon Musk revealed the upcoming release of TruthGPT, a purported "truth-seeking" massive language model created to combat ChatGPT's apparent left-wing bias.