On Wednesday, Microsoft unveiled two custom-designed computing chips, signaling a shift in the tech industry towards in-house technology development, particularly for costly artificial intelligence (AI) services. These chips are not for sale but are intended to enhance Microsoft’s subscription software and Azure cloud computing services.
At the Ignite developer conference in Seattle, Microsoft introduced the Maia chip, designed to accelerate AI computing tasks. It’s the foundation for the $30-a-month “Copilot” service for business software users and developers creating custom AI services. Maia is tailored for large language models like those in Microsoft’s Azure OpenAI service, a result of its collaboration with ChatGPT creator OpenAI.
Given the high costs of AI services, which can be up to ten times more than traditional services, Microsoft plans to streamline its AI efforts through a common set of foundational AI models, with Maia being central to this strategy. Scott Guthrie, the executive vice president of Microsoft’s cloud and AI group, emphasized that Maia will provide faster, lower-cost, and higher-quality solutions.
Furthermore, Microsoft announced it would offer Azure cloud services running on the latest chips from Nvidia and Advanced Micro Devices (AMD) next year, currently testing OpenAI’s GPT 4 on AMD’s chips. Ben Bajarin of Creative Strategies noted that while this doesn’t displace Nvidia, it enables Microsoft to deliver AI services in the cloud until PCs and phones can handle them directly.
Read More: Google Bets on AI Chatbot Bard to Attract Billions More Users
The second chip, Cobalt, is a CPU made with Arm Holdings technology, aimed at reducing internal costs and competing with Amazon Web Services (AWS) “Graviton” chips. Cobalt has been tested with Microsoft Teams and is part of Microsoft’s strategy to offer competitive performance and price-to-performance ratios.
AWS, which plans its developer conference later this month, highlighted its Graviton chip’s 50,000 customers and commitment to ongoing innovation.
Microsoft’s chips, both utilizing 5-nanometer manufacturing technology from Taiwan Semiconductor Manufacturing Co, reflect a trend in big tech firms to internalize key technologies. Maia, particularly, is set apart by its use of standard Ethernet network cabling, moving away from more expensive custom networking solutions previously employed by Microsoft in its supercomputers for OpenAI. This approach underlines Microsoft’s preference for standardization in its technology development.
