Microsoft has begun manufacturing two new chips optimized for artificial intelligence and cloud data infrastructure

The technology company Microsoft began manufacturing Processor designed to work with artificial intelligence (AI)To be used to train large language models (LLM) and avoid the more expensive reliance on Nvidia microchips. Azure Maia AI, as they call it, is optimized to power generative AI and general AI operations.

At the same time, it also established a Windows development company Azure Cobalt, a general-purpose CPU with an ARM architecture – More efficient – ​​Designed to enhance Microsoft Cloud infrastructure. Both chips, made from silicon, are designed to power your Azure data centers.

Both chips were announced at the Microsoft Ignite event, where Scott Guthrie, vice president of the company’s division responsible for artificial intelligence and cloud services, commented on the goals of these launches.

“Microsoft is building The infrastructure needed to support innovation in the field of artificial intelligenceHe explained that we are reimagining every aspect of our data centers to meet the needs of our customers.

Manufacturing Azure Maia AI and Azure Cobalt chips allows the company to have a key part of its infrastructure, where it already runs the integrated circuits, servers, cooling systems and software.

These two new chips It will arrive in Microsoft data centers in early 2024It will be initially installed on servers that support Microsoft Copilot and Azure OpenAI services.

Currently, Microsoft has servers equipped with Nvidia H100 GPUs, and in the future it will offer the same manufacturer’s H200 AI chips and AMD’s Instinct MI300X in some of its devices.