Stifel Raised Its Microsoft Target — The Same Analyst Who Flagged the Capex Problem

Stifel raised its price target on Microsoft to $530, up from $450, and the analyst who did it — Brad Reback — is the same one who downgraded the stock to Hold in February, cutting the target to $392 and warning about "ballooning AI capital expenditures" and Azure deceleration.

The turnaround is the point. Stifel didn't change its mind because the underlying economics shifted. It changed its mind because Microsoft changed the story.

Two days ago, Microsoft announced it would start disclosing Azure revenue in dollars for the first time — a concession to years of Wall Street pressure. In the same announcement, the company renamed its core business unit "Agents and Infra," merging its cloud and productivity segments to make AI sound like the center of gravity rather than one growth stream among many. Then, on this side of that reveal, Stifel cites "accelerating momentum" for Copilot, customers moving from proof-of-concept to full-scale deployment, and lifts the target above where the stock trades.

The analyst who identified the capex problem is now pricing past it because the company learned to reframe it.

Let's look at the numbers behind the reframing.

In the June quarter, Azure revenue reached $29.4 billion under the new structure, up 42% year-over-year. That's real growth. Azure crossed $100 billion in cumulative annual revenue for the first time during fiscal 2026, and Microsoft is guiding for 44% to 45% Azure growth in the upcoming quarter. Copilot hit 30 million paid seats — up 10 million in a single quarter, from 20 million in March. At $30 per user per month, that implies roughly $10 billion in annualized revenue.

The counterweight is in the cash flow statement. Microsoft's trailing twelve-month capital expenditure sits at $115.9 billion. The company guided toward a $190 billion annual run rate for fiscal 2026. Operating cash flow TTM is $182.9 billion, which sounds fine until you subtract capex and get free cash flow of $66.99 billion — down 6.5% year-over-year. The company that built its valuation on near-zero marginal costs for software distribution is now spending more on physical infrastructure in a single year than most S&P 500 companies generate in annual revenue.

The return on that spending has not been demonstrated at scale.

Microsoft's AI business — Azure AI consumption plus Copilot — reached a $37 billion annual revenue run rate in the March quarter, up 123% year-over-year. That number is impressive until you put it next to the spending curve. $37 billion in AI revenue against $115.9 billion in TTM capex means the company is investing roughly three dollars in infrastructure for every dollar of AI revenue it collects. The ratio improves with time if adoption holds, but GPU hardware has a useful life of roughly six years, and Microsoft has already committed those allocations. The investment is sunk, not discretionary.

The other detail that changes the reading: 45% of Microsoft's Azure remaining performance obligations — the backlog of contracted future revenue — comes from OpenAI. The company recorded a $7.6 billion gain on its 27% OpenAI stake last quarter and has committed $13 billion in direct funding. If OpenAI diversifies its cloud providers through its separate $110 billion funding round, that creates a concentrated revenue risk that the "Agents and Infra" rebrand does not eliminate.

There's a separate tension in the Copilot numbers themselves. The 30 million paid seats represent roughly 6.5% of Microsoft's total commercial Microsoft 365 seat base of about 464 million. Management touts that the number of enterprise customers with more than 50,000 seats increased more than 7x year-over-year and that time to "high usage" has compressed from months to days. But 44.2% of users who stop using Copilot cite distrust of answers as their reason, and the accuracy net promoter score was -19.8 as of January — meaning users were more likely to distrust the output than recommend it. The product is scaling, but the quality gap between marketing and daily reliability is a deployment constraint.

None of this is negative evidence. It's unproven evidence.

The question for the investor is not whether Microsoft's AI strategy is a failure. The question is what the $3.79 trillion market capitalization is paying for. Microsoft trades at a trailing P/E of 28.3x, roughly 1.6x Alphabet's 17.2x and 1.4x Amazon's 20.6x. That premium requires the assumption that the AI investment will compound at a rate higher than its peers and that the capex cycle will produce durable margin expansion rather than a permanent structural shift toward heavier, lower-margin economics.

Microsoft's gross margin sits at 68.3%, operating margin at 46.8%, and ROIC at 28.3% — all elite numbers. But the transition from asset-light software to capital-intensive AI infrastructure changes the cost structure at the margin. Every additional GPU rack in an Azure data center adds to COGS in a way that an additional Office 365 license never did. The company is trying to run both businesses simultaneously: maintaining the software margin story while executing the cloud infrastructure story.

The segment rebrand is a signal about which story management wants priced in. "Agents and Infra" bundles the high-margin Copilot growth with the lower-margin Azure infrastructure spend, creating a composite that looks like AI-led acceleration rather than infrastructure investment chasing software monetization. Microsoft will also stop showing costs and operating margins for the old three-segment structure, which means the old comparables disappear from earnings calls.

That's the practical consequence: the accounting change removes the transparency that let investors track whether the software and infrastructure margins were diverging.

Stifel's shift from lone bear to bullish target-holder is the most readable signal in this week's coverage. The firm that called out the exact capex-to-cash-flow problem investors feared is now citing the exact deployment metrics Microsoft highlighted on the same earnings call. It's not that Stifel was wrong in February or is wrong today — it's that the company successfully shifted the frame from "is this spending justified?" to "is this momentum real?"

The momentum is real. The justification is still being built.

At $510 per share, Microsoft is not trading at the depressed valuation it hit in June near $349, when the market was still pricing the capex concern. The stock has recovered more than 45% from that low. The $530 target gives only 4% implied upside. The sell-side consensus median across 47 analysts is $600, with the range stretching from $400 to $870. These are not prices that reflect execution risk. They reflect the assumption that the execution will work.

The investor who bought at $400 in February is sitting on a comfortable gain. The investor who is considering the stock now carries the question of whether the $190 billion spending cycle will deliver the margins the P/E requires, or whether the company will settle into a permanently more capital-intensive profile that trades at a lower multiple. The segment rebrand makes that question harder to track, not easier.

That's what the story change was designed to do.