In an attempt to counter the rising costs of providing AI services, Microsoft (MSFT.O) announced on Wednesday that it will be releasing two custom-designed CPU chips. Following in the footsteps of other digital behemoths, Microsoft intends to internalize key technology in order to reduce the high costs associated with offering AI services.
Maia and Cobalt, the recently revealed chips, are not meant for retail sales; instead, they will be utilized to improve Microsoft’s software subscription services and bolster its Azure cloud computing service. Aimed at corporate software customers and developers looking to create custom AI services, the $30/month “Copilot” service is powered by the Maia chip, which was unveiled at the Ignite developer conference in Seattle. The Maia chip is meant to speed up AI processing processes.
Maia, which is specifically designed to run huge language models, is critical to Microsoft’s Azure OpenAI service, which is the product of a collaboration between Microsoft and the OpenAI designers of ChatGPT. The executives of the company have indicated that they are committed to tackling the rapidly increasing costs by directing the majority of their AI projects through a single set of foundational AI models, for which the Maia chip has been developed.
According to Microsoft’s senior vice president of cloud and AI, Scott Guthrie, “We think this gives us a way that we can provide better solutions to our customers that are faster, lower cost, and higher quality.”
Additionally, Microsoft announced that starting in the upcoming year, cloud services powered by the newest flagship CPUs from Advanced Micro Devices AMD.O and Nvidia (NVDA.O) will be made available to Azure users. Microsoft announced that it is testing AMD CPUs with GPT 4, the most sophisticated model from OpenAI.
Microsoft has positioned itself strategically to reduce costs in response to the industry-wide challenge of high AI service costs. Microsoft can now provide AI services on the cloud thanks to the Maia processor, at least until smartphones and personal computers are powerful enough to do these functions.
The second chip to be introduced, called Cobalt, is positioned as a counter to Microsoft’s main cloud competition, Amazon Web Services (AWS), as well as an internal cost-cutting initiative. Microsoft’s workplace communications application Teams is powered by Cobalt, which was developed as a central processing unit (CPU) using technology from Arm Holdings (O9Ty.F). Cobalt has already undergone testing. Furthermore, by granting access to Cobalt, Microsoft hopes to directly challenge AWS’s “Graviton” line of proprietary CPUs.
According to Rani Borkar, corporate vice president for Azure hardware systems and infrastructure, Taiwan Semiconductor Manufacturing Co. (2330.TW) provides the 5-nanometer technology used in the production of Maia and Cobalt chips. A notable example of this standardization is the choice Microsoft made in their supercomputers designed for OpenAI, choosing Ethernet network cabling over more expensive bespoke Nvidia networking technologies.
Although Microsoft did not provide sufficient technical details to allow direct comparisons with other chips, Ben Bajarin, CEO of Creative Strategies, emphasized Microsoft’s special opportunity to profit from its profitable service-based business model, saying, “Microsoft has a very different kind of core opportunity here because they’re making a lot of money per user for the services.”
While AWS gets ready for its next developer conference later this month, it is also preparing for Microsoft’s efforts to optimize AI services. AWS is committed to innovation and will continue to supply AWS-designed chips in the future for better price-performance that meets client workload requirements.
