Microsoft’s aggressive push into custom silicon is paying off with dramatic efficiency improvements that could reshape the economics of its cloud business. Chief Executive Officer Satya Nadella told investors that running the company’s MAI family of large language models on its proprietary AI chips yields 40% better performance per watt compared to off-the-shelf alternatives, a milestone that signals lower operating costs and potentially fatter margins for its already surging Azure platform.
The disclosure came as Microsoft wrapped up a fiscal fourth quarter that reignited Wall Street’s enthusiasm for the tech giant. Azure revenue jumped 43% year-over-year, accelerating from 40% in the prior quarter. Total cloud revenue climbed 27% to $59.3 billion, accounting for the lion’s share of the company’s $90 billion in quarterly sales. Overall revenue expanded 18% from a year earlier, while operating income rose at the same pace.
“Running MAI models on our custom chips yields 40% better performance per watt,” Nadella said on the July 29 earnings call. The efficiency gain matters because it directly reduces the cost of delivering AI services — Microsoft’s fastest-growing revenue stream. As the company scales inference workloads across millions of servers, every percentage point of power savings translates into tens of millions of dollars in avoided expenses.
Microsoft has historically leaned on external suppliers, particularly Nvidia, whose graphics processing units remain the gold standard for training and running AI models. But the Redmond, Washington-based company has been designing its own accelerators for several years, a strategy that mirrors moves by Amazon Web Services and Google Cloud. By optimizing silicon for its specific MAI model architecture, Microsoft can strip away unnecessary circuitry and tune memory bandwidth to match inference patterns, squeezing more work out of each watt.
The custom-chip push arrives at a pivotal moment. Microsoft ended fiscal 2026 with a $678 billion cloud backlog, up 84% year-over-year, suggesting demand shows no signs of cooling. Some analysts project that global AI infrastructure spending could hit $1 trillion within three years. If Microsoft captures a meaningful share of that spending while simultaneously lowering its per-query cost structure, the combination of volume growth and margin expansion could prove potent.
Investors responded swiftly. Microsoft shares surged roughly 25% in the days following the report, clawing back from a trough below $350 per share toward the $500 level. The stock had been under pressure for months, falling nearly 30% from its 52-week high above $553 as the market rotated toward what it perceived as purer AI plays. Even after the post-earnings rally, shares remained roughly 12% below that peak.
The valuation math still tilts in Microsoft’s favor, according to some analysts. At roughly 27 times trailing earnings, the stock trades at a slight discount to the S&P 500’s multiple of about 29, despite growing considerably faster than the index average. The company also told investors it expects to achieve positive free cash flow in fiscal 2027, a commitment that eased concerns about the enormous capital expenditures required to build out AI infrastructure.
Beyond the chip story, Microsoft’s earnings revealed deepening AI adoption across its product portfolio. Its AI platform Foundry surpassed 100,000 customers, with revenue more than doubling year-over-year. The number of enterprise customers using both Foundry and Fabric, an analytics tool, jumped 60%. The company also introduced Web IQ, a resource that gives AI agents access to real-world intelligence from across the web. OpenAI’s ChatGPT and other assistants already tap into the service, underscoring how tightly Microsoft is woven into the broader AI ecosystem.
Other business lines performed well. LinkedIn, search advertising, and Microsoft 365 Consumer all posted double-digit revenue growth. Xbox content and services were the main soft spot but represent a small fraction of total earnings.
Microsoft’s custom silicon strategy does not mean it is abandoning Nvidia. The company continues to buy Nvidia GPUs in massive quantities to serve customers who demand peak performance or run workloads that are not yet optimized for Microsoft’s chips. But for its own MAI models — which compete with offerings from OpenAI, Anthropic, and Google — the in-house hardware gives Microsoft a lever that pure-play cloud providers cannot easily replicate. If the 40% efficiency figure holds as deployments scale, Microsoft could undercut competitors on price while maintaining or even expanding gross margins in its intelligent cloud segment.
The broader takeaway for investors is that Microsoft’s AI strategy is becoming more vertically integrated. Owning the models, the silicon, the cloud infrastructure, and the application layer creates multiple avenues to capture value as enterprises shift from experimenting with AI to deploying it at scale. The $678 billion backlog suggests that transition is well underway, and the custom-chip efficiency gains provide a concrete reason to believe profitability can keep pace with revenue growth even as capital spending remains elevated.
With the stock still trading below its all-time high and the company delivering accelerating cloud growth alongside improving unit economics, the post-earnings rally may have further room to run — provided Microsoft can sustain the execution it demonstrated in the fourth quarter.