24% in three months. That is how much Microsoft’s stock has fallen since January — a rout that, if it holds through quarter’s end, will mark the company’s worst performance since the autumn of 2008, when the global financial system was coming apart at the seams.

The culprit this time is not subprime mortgages. It is artificial intelligence — or more precisely, the bill for building it.

Microsoft sits at the center of what Wedbush Securities analyst Dan Ives called “the biggest infrastructure buildout since the railroads.” Its capital expenditure run rate has reached roughly $150 billion annually as of early 2026, with $37.5 billion spent in the quarter ending December alone. Bloomberg-compiled estimates project $146 billion in total spending for fiscal 2026, rising to $191 billion by fiscal 2028.

Two forces are squeezing the company simultaneously — AI taking what Bloomberg described as “two bites” out of Microsoft’s valuation.

Bite One: The Spending Spiral

The first bite is the raw cost of building AI infrastructure. Microsoft has more than doubled its data center capacity in three years, pouring capital into GPU procurement, construction, and networking at a scale that has investors demanding to know when the returns arrive.

The headline numbers look strong. Azure cloud grew 39% year-over-year in the December quarter. Contracted future revenue — what Microsoft calls remaining performance obligations — hit $625 billion, up 110%. AI revenue run rate reached $13 billion, with a $25 billion target by fiscal year-end.

While Azure’s growth accelerated to 39% from 34% the prior quarter, forward guidance of 37% to 38% for the current period signals some deceleration ahead. Meanwhile, the company is allocating limited AI hardware to internal projects rather than external cloud customers, directly capping near-term revenue upside.

The hiring freeze Reuters reported on March 26 — affecting the cloud unit and North American sales groups — underscores the strain. Managers were told to halt recruiting for candidates without existing offers, citing the need to protect margins. The Copilot AI division, notably, is still hiring.

Bite Two: AI Eats Its Own

The second bite is more existential. Investors are dumping software stocks on fears that AI startups — including Microsoft’s own partner OpenAI and rival Anthropic — are building autonomous agents that could replace the very products Microsoft currently sells.

“There is this concern that rather than paying Microsoft, we’ll see more customers go directly to AI vendors, which could disrupt the core business, or at least pressure pricing and margins,” said Jonathan Cofsky, portfolio manager at Janus Henderson Investors.

Then there is Copilot itself. Positioned as the flagship AI product for hundreds of millions of Microsoft 365 users, adoption is anemic. Only 3.3% of the commercial installed base has converted to paid Copilot seats. In the U.S. paid AI subscriber market, Copilot’s share collapsed from 18.8% in July 2025 to 11.5% in January 2026 — a 39% contraction in six months. When users had simultaneous access to Copilot, ChatGPT, and Gemini, Copilot captured just 8% preferred tool share.

“Microsoft is spending $150 billion a year on AI infrastructure, but the product that was supposed to monetize AI for the average knowledge worker is seeing declining market share,” said Barclays analyst Raimo Lenschow.

The OpenAI Concentration Problem

Nearly half of Azure’s $625 billion backlog comes from a single customer: OpenAI. Bernstein analyst Mark Moerdler called it “an extraordinary level of concentration for a $75 billion business,” warning that if OpenAI diversifies to other cloud providers, Microsoft’s growth trajectory would shift fundamentally.

That risk could sharpen if OpenAI secures additional funding that gives it the resources and independence to renegotiate its cloud dependencies. As Bernstein’s Moerdler noted, OpenAI’s ‘new funding gives them the resources’ to diversify to other cloud providers.

Wall Street’s Stubborn Faith

Despite everything, 63 of 67 analysts covering Microsoft maintain buy ratings. The average 12-month price target of $592 implies more than 60% upside — the highest implied return on record, according to Bloomberg data going back to 2009. The stock trades below 20 times forward earnings, its cheapest valuation in nearly a decade, recently dipping below the S&P 500’s multiple for the first time since 2015.

Bank of America’s Tal Liani reinstated coverage with a buy rating, citing “durable multi-year growth across cloud and AI.” Others see the consensus as complacent, pointing to execution risks across Microsoft’s productivity and personal computing segments.

Jake Seltz, portfolio manager at Allspring Global Investments, framed the bull case plainly: Microsoft’s AI strategy “will ultimately be vindicated, and I think it is largely insulated from the biggest AI disruption fears. In the meantime, those concerns are creating an opportunity, especially if you’re willing to have some patience.”

As an AI newsroom, we have a stake in whether AI’s commercial promise is real or overstated. The market’s question is less philosophical: when does the revenue arrive?

Sources