Jakarta, INTI - Microsoft has officially launched its latest chip, the Maia 200, a specialized processor designed to accelerate and optimize artificial intelligence (AI) inference. The company describes the Maia 200 as the silicon backbone for running large-scale AI models with lower power consumption.
The Maia 200 succeeds the Maia 100, which was introduced in 2023, and brings significant technical improvements. Microsoft equipped the Maia 200 with over 100 billion transistors, delivering performance of more than 10 petaflops at 4-bit precision and approximately 5 petaflops at 8-bit precision, far surpassing its predecessor.
Inference is the stage of computation when an AI model is executed to generate outputs, distinct from the model training process. As the AI business grows, inference costs have become a major component of operational expenses.
Microsoft Aims to Reduce AI Inference Costs
Technology companies are now racing to find ways to reduce these costs without compromising performance, and Microsoft hopes the Maia 200 can meet this challenge.
“Pratically, a single Maia 200 node can easily run today’s largest models, with ample capacity remaining for even bigger models in the future,” said Scott Guthrie, Executive Vice President of Cloud + AI at Microsoft, in an official statement.
The launch of Maia 200 also reflects a broader trend in the tech industry toward developing proprietary chips. This move aims to reduce dependence on Nvidia GPUs, which have long been the backbone of AI computing.
Google offers its TPU through cloud services, while Amazon has developed the AI accelerator chip Trainium, with the latest Trainium3 released in December 2025. These chips help offload some of the computational burden from Nvidia GPUs and reduce hardware costs.
Conclusion
With the Maia 200, Microsoft positions itself at the forefront of the emerging AI chip market. By combining high performance, energy efficiency, and scalability, the chip not only reduces operational costs but also supports the deployment of increasingly larger AI models, marking a significant step toward proprietary AI infrastructure.
Read more: Driving Economic Growth, Deputy Minister Stresses AI Adoption in Strategic Sectors