A key component of Microsoft’s strategy to power its artificial intelligence ambitions and reduce reliance on external silicon providers has reportedly encountered a significant setback. The company’s next-generation custom AI chip, code-named Braga, is facing a delay of at least six months in its mass production timeline, according to a recent report.
Production Timeline Shifted
The delay means that mass production of the Maia AI chip, as it is formally known, is now anticipated to commence in 2026, a notable shift from the original target of 2025. This development was reported by The Information on Friday, June 27, 2025, citing insights from three individuals purportedly involved in the chip’s development effort. The postponement could have implications for Microsoft’s internal infrastructure plans, as the company had initially intended to integrate and utilize the Braga chip within its sprawling data centers starting in 2025.
Performance and Competitive Landscape
When the Braga chip eventually does reach mass production, its expected performance characteristics are another point of consideration. According to the report, the chip’s performance is projected to be significantly less than that of Nvidia’s Blackwell chip. Nvidia, currently the dominant player in the high-performance AI chip market, released its Blackwell architecture in late 2024, setting a benchmark for processing capabilities in the AI space.
This performance disparity highlights the challenges major cloud providers face in developing in-house silicon that can directly compete with the bleeding edge offered by specialized chip manufacturers like Nvidia, particularly on their initial generations of custom designs.
Factors Contributing to the Delay
The reasons behind the reported six-month delay are multifaceted. The Information’s report indicated that several factors have contributed to pushing back the mass production schedule for the Braga chip. These include unanticipated changes to the chip’s design during the development process, which often necessitate revisions and testing that consume considerable time.
Furthermore, the report cited staffing constraints and a high rate of turnover within the teams working on the Braga project as significant hurdles. Developing complex, cutting-edge silicon requires highly specialized talent, and challenges in attracting, retaining, and managing these teams can directly impact project timelines.
Strategic Importance of Custom Silicon
This setback occurs within the broader context of Microsoft’s strategic imperative to reduce its dependence on expensive, commercially available chips, most notably those from Nvidia. The rising costs associated with procuring vast quantities of high-end GPUs for AI training and inference have become a substantial expense for cloud providers.
Developing custom processors for both AI operations and general-purpose applications allows companies like Microsoft to potentially optimize performance for their specific workloads, gain greater control over their supply chain, and ultimately aim for cost efficiencies at scale. The Braga chip is a crucial part of this long-term strategic vision.
Cloud Provider Silicon Arms Race
Microsoft is not alone in pursuing an in-house silicon strategy. Other major cloud providers are also heavily invested in developing their own custom chips to power their infrastructure and differentiate their services.
Google (an Alphabet company), for instance, has seen notable success with its Tensor Processing Units (TPUs), designed specifically for machine learning workloads. Google unveiled its seventh-generation AI chip in April 2025, demonstrating continued iteration and advancement in its custom silicon line.
Similarly, Amazon introduced its Trainium3 AI chip in December 2024. This chip, designed for training deep learning models, is expected to become available for use later in 2025, further intensifying the competition in the cloud AI infrastructure market.
The reported delay for Microsoft’s Braga chip underscores the complexities inherent in designing and manufacturing advanced semiconductors, even for companies with vast resources. While the strategic rationale for pursuing custom silicon remains strong among major cloud providers, bringing these sophisticated processors to market on ambitious timelines presents significant engineering and operational challenges.