Microsoft’s ambitious push to develop its own custom AI chips has hit a major roadblock. The company’s next-generation Maia chip, code-named “Braga,” is now expected to enter mass production only in 2026 — at least six months behind the original schedule, according to sources cited by The Information.
Originally slated for rollout in 2025, the delay stems from unanticipated design revisions, staffing constraints, and high employee turnover. Microsoft had aimed to use Braga across its data centers this year to reduce dependency on Nvidia’s powerful but costly chips, particularly the Blackwell series which is already in deployment.
Falling behind in the chip race
When eventually released, the Braga chip is expected to fall short of Nvidia’s Blackwell chip in terms of performance. This puts Microsoft at a disadvantage in the competitive custom AI chip race where peers like Google and Amazon have already made significant strides.
Also read: Nvidia Halves AI Training Time with New Blackwell Chips
Google continues to build momentum with its Tensor Processing Units (TPUs), unveiling its seventh-generation AI chip in April. Amazon, too, announced its Trainium3 chip in December 2024, due to launch later this year. Both companies are leveraging in-house designs to optimize performance and cost for large-scale AI applications.
Microsoft introduced its Maia chip in November 2023 with similar ambitions, but progress has been slower than expected. The chip was seen as a cornerstone of Microsoft’s efforts to reduce reliance on Nvidia and drive AI innovation within its Azure cloud infrastructure.
Bigger picture for AI infrastructure
As the AI boom accelerates, the delay in Braga’s rollout underscores the challenges tech giants face in balancing in-house development with scalability. The chip was central to Microsoft’s strategy of building vertically integrated AI systems to serve both internal services and external enterprise clients.
The news comes at a time when Big Tech is heavily investing in proprietary silicon to power AI workloads — not just to control costs, but to ensure supply chain stability and optimize workloads for specific applications.
While Microsoft continues to work toward eventual mass production, the delay highlights how difficult it is to deliver cutting-edge silicon at scale, particularly when trying to compete with seasoned GPU manufacturers like Nvidia.
