A Stock That Was Down 18% on the Year Jumped 15% in a Single Session

Wednesday, July 29, just after the US market closed. Microsoft posted results for the fourth quarter of its 2026 fiscal year. Up to that moment MSFT was down more than 18% year to date. The largest and most boringly reliable software company on earth was one of the least popular mega-caps of the year, and the reason was simple: it was spending too much money.

The next day, Thursday July 30, the stock closed up more than 15%. Market capitalization landed at $3.35 trillion, and the value added in that one session was roughly $450 billion. According to Reuters' tally, that is the largest single-day market value gain any company has ever recorded. The previous holder was Nvidia, which added $441 billion on April 9, 2025.

It helps to sit with the size of that number for a second. The $450 billion added in one day is more than Intel and AMD combined are worth, and roughly the entire market capitalization of Netflix. What produced it was not a product launch, an acquisition, or a government contract. It was a handful of slides in an earnings deck.

And within those slides, the market really only reacted to two lines. One was the Azure growth rate for the quarter that had just ended: 43%. The other was the Azure growth guidance for the quarter ahead: approximately 45% on a constant-currency basis. Consensus had modeled 40.92%. A four-point gap moved $450 billion.

Why did four points matter that much? Because for the past two years exactly one question has been sitting on hyperscaler valuations: when do the hundreds of billions of dollars being poured into data centers turn into revenue? Microsoft answered that question with numbers for the first time — not by cutting spending, but by showing that demand is growing faster than the spending is.

That distinction is the whole story. Every hyperscaler has been able to say "demand is strong." What Microsoft did on July 29 was attach three specific, auditable figures to that sentence: a 43% growth rate that accelerated instead of decaying, a contracted backlog up 84%, and a forward guide that went up rather than down. Investors had been pricing in the possibility that the acceleration story was over. It wasn't.

The People Who Made That Day Happen

The person standing at the center of it is Amy Hood, Microsoft's chief financial officer since 2013. She is the executive who has absorbed the "why are you spending like this" question on every earnings call for the past several quarters. Back on the fiscal third-quarter call in April 2026, she raised the calendar-2026 capital expenditure outlook to roughly $190 billion, citing surging memory prices. The market did not enjoy that day.

Satya Nadella came in with a sentence engineered to change the frame: "We are advancing the frontier on the cost-to-outcome curve, ensuring every customer can turn tokens into business results." Translation — stop looking at the spending line, start looking at unit economics. And he handed the press the headline he wanted alongside it: "This year, Azure revenue surpassed $100 billion for the first time, and Microsoft 365 Copilot reached over 30 million paid seats, reflecting the confidence customers are placing in us to power their AI transformation."

On the other side of the table was the market — specifically, the investors who had sold the stock down more than 18% since January. Jake Behan of Direxion summarized what they had been waiting to hear: "The key question was whether it could shift the conversation from how much it is spending on AI to what it is earning from those investments, and the results suggested meaningful progress." Brian Mulberry of Zacks Investment Management put it more bluntly, saying Microsoft "struck the tone markets are looking to hear as the key drivers of growth came from the cloud and AI divisions."

There is also a supporting character in this story that almost nobody expected: Anthropic. Microsoft invested $5 billion in Anthropic in November 2025, in the same arrangement under which Anthropic committed to purchase $30 billion of Azure compute. This quarter Microsoft booked a $3.2 billion gain on that Anthropic stake — worth $0.33 of diluted earnings per share. Eight months in, more than half the investment has come back as a paper gain.

The OpenAI side of the ledger was far flatter over the same period. Per the official earnings release, net gains from Microsoft's OpenAI investments increased net income by $480 million ($0.07 per diluted share) in the fourth quarter and by $4,963 million ($0.67 per diluted share) for the full fiscal year. Put those side by side and one quarter of Anthropic gains nearly matched a full year of OpenAI gains. Microsoft's decision to keep a foot in two frontier labs at once paid off this quarter from the foot people were paying less attention to.

What the Release Actually Said

Start with the quarter that already happened. For the three months ended June 30, 2026, revenue was $90.0 billion, up 18% (up 17% in constant currency). Operating income was $40.6 billion, up 18%. GAAP net income was $35.8 billion, up 31%, and GAAP diluted earnings per share came in at $4.81, up 32%. Strip out the $3.2 billion Anthropic gain and the non-GAAP figures land at $35.3 billion in net income (up 22%) and $4.74 per share (up 23%).

The segment split shows exactly where the growth lives. Intelligent Cloud was $39.3 billion, up 32%. Productivity and Business Processes was $37.8 billion, up 14%. More Personal Computing was $12.9 billion, down 4% — the only line moving backward. Microsoft Cloud as a whole came in at $59.3 billion, up 27%. And Azure and other cloud services grew 43%, the fastest rate since early 2022.

But the numbers that actually moved the stock were forward-looking. First, the roughly 45% constant-currency Azure guide for the coming quarter. Re-acceleration in a cloud business of this size is genuinely unusual — as the denominator grows, growth rates normally drift down, not up. Second, commercial remaining performance obligation, meaning revenue that is under contract but not yet recognized, hit $678 billion, up 84%. That is not a pipeline estimate. That is signed paper.

Hood's explanation of the re-acceleration on the call is the part worth reading twice. It did not come primarily from new data centers coming online. It came from running the existing fleet harder: efficiency gains across the CPU and GPU fleet, process improvements that shortened the lead time to bring new capacity live, and a 50% reduction in GPU dock-to-live time. Because supply is short, those gains converted to revenue inside the same quarter. Hood's phrasing: "Demand continues to exceed available supply. When we can make efficiency gains, they are quickly monetized."

Metric FY26 Q4 Year over year Note
Revenue $90.0B +18% +17% constant currency
Operating income $40.6B +18% 45% operating margin
GAAP diluted EPS $4.81 +32% Includes $0.33 Anthropic gain
Microsoft Cloud $59.3B +27% —
Azure and other cloud — +43% Fastest since early 2022
Intelligent Cloud $39.3B +32% —
More Personal Computing $12.9B −4% Only declining segment
Commercial RPO $678B +84% Contracted, unrecognized
Capex + finance leases $41B +69% Quarterly
Free cash flow $19.6B — Quarterly
Next-quarter Azure guide +45% cc — Consensus was 40.92%

The capital expenditure line needs to be read carefully, because the headline is misleading. Capex plus finance leases for the quarter was $41 billion, up 69%. Yet the calendar-2026 outlook came down from the roughly $190 billion guided in April to about $175 billion. That looks like a spending cut and it is not one. Effective in fiscal 2027, Microsoft extended the estimated useful lives of its data centers and office buildings from 15 years to 25 years, and will classify more future data-center leases as operating leases rather than finance leases. Hood said it plainly on the call: "Outside of this useful life impact, our calendar year 2026 CapEx investment expectations remain unchanged." The physical investment is identical. Only the accounting bucket moved.

The physical footprint numbers back that up. Microsoft brought 31 new data centers online in the quarter and 88 across the fiscal year, and management said it intends to roughly double total capacity within two years. First-quarter fiscal 2027 capital expenditure was guided above $50 billion, including the lease reclassification effect, and Hood said the company still expects to be free cash flow positive throughout fiscal 2027. Free cash flow in the fourth quarter was $19.6 billion.

One more operating datapoint that got less attention than it deserved: Foundry customers running at a one-trillion-token annualized rate grew fourfold year over year, and Microsoft 365 Copilot net seat additions more than doubled sequentially to push past 30 million paid seats. EY alone deployed the top-tier suite to 400,000 employees. Those are the metrics that make the $678 billion backlog feel less abstract.

What Each Party Walked Away With

For Microsoft's leadership, the prize is control of the narrative. For two years hyperscaler CFOs have had to play defense on spending every quarter, and the pattern held regardless of how good the rest of the numbers looked — one capex line could sink the stock. This time the order reversed. Microsoft kept the spending level intact and attached three sentences to it: demand exceeds supply, efficiency gains monetize immediately, contracted backlog is up 84%. Suddenly capex reads as a leading indicator of revenue rather than a risk. That is exactly what Nadella's "turn tokens into business results" framing was built to do.

For shareholders the math is more layered than the headline suggests. Yes, $450 billion appeared in a day. But the stock had been down more than 18% year to date going into it, so this was closer to recovering lost ground than setting a new high. At least nine brokerages raised price targets, taking the mean to $560.90. Still, it is worth holding onto the fact that $0.33 of the $4.81 in GAAP diluted EPS came from marking up the Anthropic stake. That is a change in carrying value, not cash through the door, and it can reverse.

Anthropic quietly did well here too. A $3.2 billion mark-up on Microsoft's books means Anthropic's valuation moved sharply higher over the period, and its $30 billion Azure purchase commitment very likely contributed to the backlog Microsoft is now waving around. In a market where the company selling compute and the company buying compute hold each other's equity, one side's valuation increase shows up on the other side's income statement.

For enterprise buyers the benefit is more ambiguous. Copilot crossing 30 million paid seats, with net adds more than doubling quarter over quarter, gives internal champions a data point in the "should we standardize on this" argument. Reference deployments at genuine scale — EY rolling it out to 400,000 people — mean the pilot-to-production path has been walked by someone else first. But there is a cost to that. When the vendor's CFO says on a public call that demand exceeds supply, your negotiating position has just gotten worse.

How Days Like This Have Ended Before

The lineage of record single-day gains is instructive. In 2022 Apple and Amazon each set the mark at roughly $190 billion. On February 2, 2024, Meta broke it by adding about $197 billion in a session. Three weeks later, on February 22, Nvidia broke that with roughly $277 billion. On April 9, 2025, Nvidia broke its own record with $441 billion. And now Microsoft with $450 billion. The record has more than doubled in four years mostly because the companies themselves have.

The success case worth studying is that February 2024 Meta day. Meta paired its first-ever dividend and a $50 billion buyback with evidence that its AI investment was measurably improving ad targeting. The message was not "we will spend less." It was "what we already spent is already coming back." The stock kept climbing from there. Microsoft's logical structure this week is nearly identical: do not cut the spending, show the output of the spending.

The failure case comes from the same company. Exactly two years earlier, on February 3, 2022, Meta lost roughly $232 billion in a single day — still the record single-day decline for any US company. The cause was a combination the market cannot tolerate: enormous confirmed spending on Reality Labs alongside user metrics that were rolling over. Spending confirmed, output unconfirmed. The same firm demonstrated, from both directions two years apart, that what the market pays for is not the size of the infrastructure bet but the verified demand behind it.

The longer-arc cautionary tale is Cisco in 2000. Riding the entirely true observation that internet traffic was multiplying every year, Cisco became the most valuable company in the world in March 2000. The traffic growth was real. The problem was that the telecom carriers buying the gear were buying it with borrowed money, and when the financing dried up the orders vanished. Whether Azure's $678 billion backlog is durable depends on the same variable — how much of it sits with a handful of very large customers, including OpenAI, and whether those customers can keep raising capital across the contract term. A contract is a contract. Cash is cash.

What Amazon and Google Put on the Same Exam

This day is more revealing because the competitors sat the same test the same week. On July 30, the day Microsoft's stock made history, Amazon reported second-quarter 2026 results. AWS revenue was $42.2 billion, up 37%, which Amazon described as its fastest growth in 18 quarters, at a $169 billion annualized run rate. Total net sales were $200.6 billion, up 20%, with operating income of $27.5 billion versus $19.2 billion a year earlier. But purchases of property and equipment ran $54.2 billion in the quarter alone, and Andy Jassy raised the full-year 2026 capital spending outlook from roughly $200 billion to about $220 billion.

Google went a week earlier. On July 22, Alphabet reported second-quarter revenue of $119.8 billion, up 24%, with Google Cloud revenue of $24.8 billion, up 82%. Cloud operating income jumped to $8.8 billion from $2.8 billion a year prior. On growth rate alone, Google Cloud is not close to second — it is far ahead of both rivals. And the stock fell that day anyway, because Alphabet lifted its full-year capex guidance to $195–205 billion and reported $44.9 billion of capital expenditure in the quarter, roughly double the year-ago figure.

Three companies, one week, broadly the same story, and completely divergent market reactions. Rank them by growth and it goes Google Cloud at 82%, Azure at 43%, AWS at 37%. The market did not reward that ordering. Microsoft was the only one that delivered the specific combination of "spending plan unchanged, contracted backlog up 84%, efficiency gains monetized on contact." Alphabet delivered explosive growth stapled to a capex raise. Amazon delivered accelerating growth stapled to a $220 billion capex number.

That is precisely where the competitors are running their numbers right now. A growth rate is no longer a defense on its own. The market is not asking how fast you are growing; it is asking how much you burned to get there and how much of the resulting demand is locked into signed contracts. Google has already started disclosing cloud backlog and will almost certainly push that figure further forward in its narrative. Amazon will lean on AWS re-acceleration and a 36.8% operating margin, but that $220 billion absolute spend will keep dragging on the story until it can show a comparable contracted-revenue anchor.

There is a fourth competitive axis that surfaced quietly in these results. Microsoft booking $3.2 billion from its Anthropic stake confirms that the structure where cloud providers sell compute and simultaneously own equity in the companies buying that compute is now fully entrenched. Amazon is in Anthropic. Google is in Anthropic. Nvidia is in nearly everyone. In that structure cloud revenue and investment mark-ups ride the same cycle — twice as good on the way up, twice as bad on the way down. Nobody has yet had to report a quarter where both lines move the wrong way at once.

And Nvidia sits above all of it. Every dollar of the $175 billion, $220 billion, and $195–205 billion these three companies plan to spend this year flows substantially through the same supplier. A quarter this strong for Azure is, mechanically, a strong forward signal for Nvidia's order book — which is another way of saying the entire complex is now correlated to a degree that did not exist three years ago.

So What Actually Changes for You

If you are a developer — the practical change is capacity and price. Hood said publicly that demand exceeds available supply, and Microsoft said it plans to roughly double capacity within two years. Read those together and the conclusion is unavoidable: certain regions and certain GPU SKUs will stay tight for several more quarters, and your room to negotiate discounts is narrow. If you have a large training or inference workload planned, moving your reserved-capacity commitments earlier is rational, and building region flexibility into your architecture is now a genuine cost-control lever rather than a nice-to-have.

If you are an investor — separate three things in this print. First, 43% Azure growth and a 45% forward guide are operating results. Second, the $3.2 billion Anthropic gain is a mark-to-market adjustment, not cash — $0.33 of the $4.81 GAAP EPS came from there. Third, the capex outlook falling from $190 billion to $175 billion is not a spending cut; the company explicitly attributed it to the useful-life extension and lease reclassification. Blend all three into one "earnings beat" and you raise your odds of being disappointed next quarter, when the Anthropic mark could move either way and the reported capex number resets on the new accounting basis.

If you own the IT budget at a company — your negotiating environment just got worse. When a supplier tells a public earnings call that efficiency gains are monetized immediately, that is a supplier telling you it has no reason to discount. With Copilot past 30 million paid seats and 400,000-employee reference deployments in the market, the internal "we will fall behind" pressure rises at the same time. When you sign, be explicit about seat-commitment duration, ceilings on usage-based charges, and the terms under which you can move workloads between regions.

If you just saw this in a headline — this day does not change your life. But one thing is worth carrying. Add up what the largest US technology companies have said they will spend on data centers this year and it exceeds the annual budget of a mid-sized country: roughly $175 billion at Microsoft, about $220 billion at Amazon, and $195–205 billion at Alphabet. If that money keeps producing returns, the software you use keeps changing. If demand rolls over at some point, the index moves with it. The market has bet on the first outcome, and the evidence it is betting on is Microsoft's $678 billion of contracted backlog.

If you hold an index fund or a pension — this is your story too, whether or not you own the stock directly. Microsoft is roughly 6% of the S&P 500 on its own. When $450 billion moves in a day, the index moves with it, and that works identically in the other direction. You do not have to pick individual stocks to have this concentration already sitting inside your portfolio.

🥄 Three Things You're Probably Wondering

— What does "added $450 billion in a day" actually mean? No money arrived in Microsoft's bank account. Shares outstanding multiplied by share price got that much bigger, and there is no implication that all of it could be sold at that price. What does move for real is buyback capacity, the purchasing power of stock used in acquisitions, and the value of employee equity compensation.

— Capex guidance dropped from $190 billion to $175 billion. Did Microsoft cut spending? No. Amy Hood said directly on the call that "outside of this useful life impact, our calendar year 2026 CapEx investment expectations remain unchanged." Extending the estimated useful life of data centers and buildings from 15 to 25 years and reclassifying leases lowered the reported figure; it did not lower the number of servers and GPUs being bought. Guidance for the coming quarter is above $50 billion.

— Can I take the $678 billion backlog at face value? It is firmer than a forecast because it is contracted. But the 84% increase is reported to include very large deals, OpenAI among them, and the disclosure alone does not tell you the term length or whether those customers can keep funding themselves across it. The real test over the next several quarters is the pace at which contracts convert to recognized revenue and the balance-sheet strength of the counterparties. Too early to call.

Further Reading

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!