Microsoft’s AI chips boost Azure efficiency up to 40 percent

Microsoft’s cloud business just got faster and cheaper. On July 29, Microsoft reported that its Azure cloud revenue grew 43% year over year for the fourth quarter of its 2025 fiscal year, beating expectations and accelerating from the previous quarter’s 40% growth. CEO Satya Nadella credited much of that gain to Microsoft’s own AI chips, which are delivering up to 40% better performance per watt compared with using third-party hardware. That means Microsoft can run its large language models more efficiently, cutting power costs and boosting margins in a segment that now drives most of the company’s profit.

Microsoft has long relied on Nvidia’s GPUs to train and run AI models, but over the past two years it has sharply increased its use of internally designed chips. The MAI family of AI accelerators now powers many of Microsoft’s cloud services, including those running its own LLMs and those from partners like OpenAI. Nadella told investors that shifting workloads to these custom chips is already showing up in the numbers. Lower energy use per task translates directly into lower operating costs, and with Azure growing at double-digit rates, even small efficiency gains add up quickly.

The scale of Microsoft’s cloud backlog gives a sense of how much room is left to grow. At the end of the quarter, Microsoft’s committed cloud revenue stood at $678 billion, up 84% from a year earlier. Analysts expect global AI infrastructure spending to reach as much as $1 trillion within three years, and Microsoft’s custom chips are positioning it to capture a larger share of that market. While Microsoft shares have trailed the S&P 500 so far this year—up just 5% versus 12% for the index—the stock surged after the earnings report on hopes that higher margins and sustained demand will drive further gains.

Investors are watching closely to see whether Microsoft can keep expanding its chip advantage. Every percentage point of efficiency improvement on Azure means more profit without a matching rise in costs. If demand for AI services stays strong, Microsoft’s bet on homegrown silicon could turn today’s efficiency gains into tomorrow’s bigger earnings beat.