Microsoft is staging an artificial-intelligence comeback that could power the stock to new highs, some on Wall Street believe.

Shares of Microsoft are up 30% since the company reported impressive fiscal fourth-quarter results at the end of July, putting the stock into positive territory for the year.

It’s just the beginning of a much larger rally, according to Bernstein analyst Mark Moerdler. “Even with the relatively large recent move the stock is still trading at a meaningful discount to its prior valuation multiple range,” he wrote in a Monday note.

Shares of Microsoft now trade at 25x forward earnings, a significant discount from the 33x multiple they commanded a year ago. Moerdler raised his price target to $660 from $647 previously, implying a 27x forward multiple and room for the stock to run more than 30% higher from current levels.

For the most recent quarter, Microsoft spent $41 billion on capital expenditures and finance leases, a lower amount than expected. The company also extended the useful life of its data centers to 25 years from 15, decreasing long-term AI costs.

Still, investors remain concerned that Microsoft is overbuilding its AI capacity, which Moerdler believes is a “misunderstanding” of Microsoft’s business trajectory.

Microsoft’s future lease obligations increased 255% in the latest fiscal year to $329.1 billion, but Moerdler points out that these leases are being rolled out over a seven-year period from 2027 to 2033, with lease terms between one and 20 years. This results in a “reasonable mid-teens growth of long-term lease expenses” in line with historic norms, he wrote. Additionally, many of these contracts contain conditional requirements or cancellation clauses in case build-outs stall.

Microsoft has also given itself flexibility with its AI hardware purchasing strategy. For fiscal 2027, Microsoft committed $169 billion in purchase agreements for chips, power contracts, cooling systems and other AI infrastructure needs. However, commitments for fiscal 2028 and beyond are just $25 billion.

“This would mean that if AI was the bubble that some argue, then Microsoft may be stuck with data centers but not AI-specific hardware, which is the far larger cost of AI,” Moerdler wrote. In that scenario, Microsoft could utilize its data-center capacity for its cloud business and internal use, as Moerdler believes Microsoft has built its data centers to be compatible with both AI and non-AI workloads.

Bernstein estimates that AI accounted for roughly 17% of Microsoft’s Commercial Cloud revenue in fiscal 2026, meaning that traditional cloud operations are still a prominent growth driver.

“Microsoft is not building too fast, but rather taking a surprisingly measured approach given the demand signals they are receiving and their ability to easily pivot facilities to meet demand,” Moerdler added.