In a quarter defined by mounting anxiety over Silicon Valley’s insatiable AI spending, Microsoft Corp. (MSFT) pulled off a masterstroke that sent its stock soaring and left rivals scrambling for cover. The company didn’t actually slash its infrastructure budget. Instead, it changed the way it accounts for the buildings that house its servers, instantly erasing $150 billion from its projected capital expenditure outlook and convincing Wall Street that the software giant has found a way to win the AI arms race without bleeding out.
The maneuver, buried in the footnotes of Microsoft’s fiscal fourth-quarter earnings report, extended the depreciable life of its data centers and office buildings from 15 years to 25 years. The adjustment also triggered a reclassification of many facilities from finance leases to operating leases. The result: Microsoft’s 2026 calendar-year capex forecast dropped from approximately $190 billion to $175 billion, a paper reduction that electrified investors who have grown increasingly wary of the cash incineration happening across Big Tech.
Shares of Microsoft exploded 15.51% the day after the July 29 earnings release, marking the largest single-day market capitalization gain in history. Over the subsequent three trading sessions, the stock continued its relentless climb, accumulating a gain of more than 25% and adding nearly $1 trillion in market value. The rally pushed Microsoft stock back toward the psychologically critical $500 level, a threshold it has struggled to hold during a volatile year that has seen shares dip into the high $300s before recovering. Year to date, MSFT is up 3.82%, still trailing the S&P 500’s 12.71% gain, but the momentum has shifted decisively in the bulls’ favor.
At the heart of the market’s euphoria is a financial engineering feat that amounts to a “bulletproof vest” for Microsoft’s cash flow statement. By stretching the useful life assumption for long-lived assets, the company reduces its annual depreciation expense, smoothing out near-term pressure on operating margins. More critically, the shift from finance leases to operating leases, while technically increasing the operating cash outflow for rent payments, removes the burden of carrying massive assets on the balance sheet. This means Microsoft avoids potential impairment charges on data centers that could become technologically obsolete long before their physical structures crumble.
The irony is sharp. Microsoft disclosed that two-thirds of its capital spending flows into short-lived assets: CPUs, GPUs, servers, and high-speed networking gear. These are the components that determine the true pace of the AI arms race, and their replacement cycles are accelerating, not lengthening. Nvidia Corp. (NVDA) and Advanced Micro Devices Inc. (AMD) are pushing new chip architectures annually. The accounting change applies only to the concrete shells and cooling systems that house these chips, which the company argues can remain useful for a quarter-century. The 15-year assumption, Microsoft noted, was already conservative relative to industry norms.
This creates what analysts are calling the “AI infrastructure paradox”: the physical envelope lasts longer, but the silicon inside becomes obsolete faster. And since chip spending drives the overwhelming majority of capital outlays, the accounting adjustment does nothing to slow the actual cash leaving Microsoft’s treasury. It simply makes the financial statements look healthier while the spending continues unabated.
“This isn’t Microsoft hitting the brakes on AI investment,” one portfolio manager told clients. “This is Microsoft putting a fresh coat of paint on the car while keeping the accelerator floored.”
The market’s willingness to reward the optics reflects a broader shift in how investors are evaluating the AI buildout. After two years of cheering every billion-dollar infrastructure commitment, Wall Street is now demanding proof that the spending will generate returns. Alphabet Inc. (GOOGL, GOOG) raised its 2026 capex guidance to between $195 billion and $205 billion, pushing quarterly free cash flow negative for the first time. Amazon.com Inc. (AMZN) lacks a dominant business-software entry point, meaning its AI monetization is tethered almost entirely to renting raw compute through Amazon Web Services. Meta Platforms Inc. (META), with no public cloud and no enterprise application suite, must convince advertisers that AI-optimized ad delivery justifies the billions flowing into data centers.
Microsoft’s position is structurally different, and the numbers back it up. Azure and other cloud services revenue surged 43% in constant currency during the quarter, crossing $100 billion in annualized revenue for the first time. AI services contributed more than 11 percentage points to that growth, tangible evidence that capex is converting into top-line expansion. The Productivity and Business Processes segment, which includes Office 365 Commercial and Copilot, grew 14%, with Microsoft 365 Copilot paid seats surpassing 30 million, double the prior quarter’s figure. Against an addressable base of 400 million Office 365 seats, the penetration runway is enormous.
More importantly, Microsoft is shifting Copilot’s pricing model from pure seat-based subscriptions to a hybrid that includes consumption-based charges. This means revenue growth can come from both adding new users and increasing the spend of existing ones, a dual engine that supports margin expansion as adoption deepens. The company’s commercial remaining performance obligations, a measure of contracted future revenue, hit $678 billion, up 84% year over year. That backlog provides visibility that no other hyperscaler can match.
The bull case at $500 per share rests on three pillars. First, the $678 billion RPO converts into recognized revenue over the coming quarters, giving Microsoft rare top-line predictability at its scale. Second, Copilot’s early penetration and evolving pricing model position it to become a material profit contributor as enterprise adoption moves beyond pilot programs. Third, the AI revenue run rate reached $37 billion, up 123%, suggesting monetization is accelerating faster than depreciation, setting the stage for free cash flow to re-accelerate once the heaviest capex period passes.
Full-year capital expenditures hit $115.95 billion, up 79.62%, crushing fourth-quarter free cash flow by 23.19%. Cash and equivalents dropped nearly 31% year over year. Bears point to two active securities class actions alleging misleading Copilot disclosures, and prediction markets assign only a 48% probability that MSFT finishes the week above $500. The stock trades at a trailing P/E of 27 and a forward multiple near 25, which bulls argue is modest for a business compounding revenue at 17.7% with a 45.1% operating margin. The consensus analyst price target sits at $562.73, with 57 analysts covering the stock: 14 Strong Buy, 40 Buy, 3 Hold, and zero Sell ratings.
Yet even as Microsoft’s external narrative centers on disciplined spending, an internal memo obtained by the press reveals a different kind of cost consciousness taking hold. Jay Parikh, executive vice president of Microsoft’s CoreAI engineering group, instructed employees on August 4 to default to OpenAI’s flagship GPT-5.6 Sol model when using GitHub Copilot. The directive marked a shift away from Anthropic’s Claude models, which had been the primary default for internal coding work.
“Internally, shifting more workloads to OpenAI models helps us get greater value from our token investment,” Parikh wrote in the memo, according to CNBC. He framed the change as a break from what he called “tokenmaxxing,” a period when developers were encouraged to run up large AI processing bills without rigorous scrutiny of the output. “Tokenmaxxing is not what we are optimizing for. I want all of us focused on maximizing outcomes that move the needle for our customers and our business.”
The directive is notable because Microsoft has built its own AI programming model and offers cloud customers access to more than 11,000 models, including Anthropic’s. As of July 2026, Microsoft divisions began operating under formal AI token budget targets for the first time, with monthly spending per engineer ranging from hundreds to several thousand dollars. Parikh acknowledged that CoreAI has not yet set formal token budgets for individual teams but encouraged staff to document both successful and unsuccessful uses of AI spending.
The internal pivot toward OpenAI carries financial logic beyond token economics. OpenAI completed a corporate restructuring nine months ago that extended Microsoft’s intellectual property rights through 2032. Microsoft said separately in April it had stopped revenue-sharing payments to OpenAI, changing the financial shape of a partnership that once ran on shared profits. OpenAI-related investment gains contributed $4.963 billion to Microsoft’s net income for the full fiscal year. Microsoft also invested up to $5 billion in Anthropic, with Anthropic agreeing to spend on Azure in return, but that relationship carries no public IP commitments.
GitHub Copilot, which runs on models from Anthropic, Google, Moonshot AI, xAI, and Microsoft’s own, has scaled to 50 million users. By directing internal traffic toward OpenAI, Microsoft captures more value from its earliest and deepest AI bet while simultaneously signaling to the market that it is serious about cost discipline.
The broader AI spending landscape is shifting beneath the hyperscalers. Cheaper open-weight models, many from Chinese labs, are quietly eroding the dominance of frontier models from OpenAI and Anthropic. They cost less to run, giving budget-conscious enterprise teams a real alternative in a market that used to have only one direction: spend more. Microsoft’s internal memo makes explicit what many investors have begun to suspect: even the company building the infrastructure believes more AI spending is not automatically better spending.
For Wall Street, the combination of accounting finesse, a dominant enterprise distribution channel, and an emerging culture of internal cost discipline has made Microsoft the default safe haven in an AI landscape increasingly defined by risk. If the arms race forces every hyperscaler to endure a period of cash flow stress, the one that bleeds least, heals fastest, and can most clearly trace spending to revenue will command the premium. At $500, that is exactly the bet Microsoft bulls are making.