In a move signaling Microsoft’s grand entrance into the AI chip arena, the tech giant is set to launch its custom Azure Maia 100 GPU and the Arm-based Azure Cobalt CPU. The world might need to brace itself for a new era where Microsoft isn’t just battling software wars but is also flexing its silicon muscles.
Azure Maia 100 and Cobalt 100: Microsoft’s Silicon Symphony Scheduled to make their grand debut in 2024, the Azure Maia 100 and Cobalt 100 are like the dynamic duo of Microsoft’s custom silicon saga. With AI at the forefront, these chips are meticulously designed to redefine the cloud infrastructure game and thrust Microsoft into the realm of artificial intelligence.
From Xbox to the Cloud: Microsoft’s Silicon Evolution Microsoft isn’t a newcomer to the world of silicon. With a history of co-engineering chips for Xbox and Surface devices, the company is leveraging its experience to create chips that not only match but potentially surpass industry standards. Rani Borkar, head of Azure hardware systems and infrastructure, emphasizes that these efforts are built on a legacy that began more than 20 years ago with Xbox silicon collaboration.
Cobalt CPU: Arm Powering the Cloud Enter the Azure Cobalt CPU, a 128-core beast designed to navigate the cloud’s diverse workloads. Named after the vibrant blue pigment, this Arm-based CPU is customized for Microsoft’s needs. Borkar notes intentional design choices, allowing control over performance and power consumption per core and per virtual machine. Currently undergoing testing on workloads like Microsoft Teams and SQL server, the Cobalt CPU aims to outperform existing commercial Arm servers, boasting up to 40% better performance.
Maia 100 AI Accelerator: Empowering AI Workloads On the AI front, Microsoft introduces the Maia 100 AI accelerator, named after a bright blue star. This liquid-cooled server processor, manufactured on a 5-nanometer TSMC process, packs 105 billion transistors. With a focus on supporting faster model training and inference times, Maia 100 collaborates with OpenAI to enhance the capabilities of large language model training and inference, marking Microsoft’s commitment to optimizing AI down to the silicon.
Open Collaboration for AI Standardization Microsoft, part of a prestigious group including AMD, Arm, Intel, Meta, Nvidia, and Qualcomm, aims to standardize the next generation of data formats for AI models. Building on the collaborative spirit of the Open Compute Project (OCP), Microsoft adapts entire systems to meet the evolving needs of AI.
Into the Azure Future: Maia and Cobalt Series The Maia 100 and Cobalt 100 names hint at an ongoing series, promising more chapters in Microsoft’s silicon journey. Borkar hints at a continuous evolution, stating, “This is a series, it’s not just 100 and done… but we’re not going to share our roadmaps.” As Microsoft readies itself for the AI revolution, these chips might become the backbone of a future where cloud infrastructure is optimized for every layer.
AI Cloud Pricing and the Copilot Effect While Microsoft remains tight-lipped about the impact on AI cloud service pricing, the introduction of Maia could potentially reshape the cost dynamics for AI cloud users. With Copilot for Microsoft 365 quietly entering the scene at a $30-per-month premium per user, Microsoft’s AI ambitions are starting to take center stage. As the company unveils new Copilot features and rebrands Bing Chat, the Maia chips could become the silent architects balancing the demand for AI chips powering these experiences.
