Microsoft and Meta Report Quarterly Results Wednesday as Investors Weigh AI Spending Plans

The Number That Cuts Both Ways

Wednesday afternoon, after the U.S. stock market closes, Microsoft and Meta will both report quarterly results^[4][16]. Microsoft is closing out its fiscal year, for the quarter that ended June 30. Meta is reporting the second calendar quarter. Investors say the profit line barely matters this time. What matters is the spending line^[5][16].

Here is the tension sitting at the center of both reports. Microsoft's cloud business, Azure, grew 40% last quarter, and revenue overall rose 18% to $82.9 billion^[1]. That is a company hitting on all cylinders. And yet Microsoft's own cloud profit margin fell to 66%, down from 69% a year earlier, because of what the company itself called continued investment in AI infrastructure^[1]. Growth and shrinking margin, in the same report, from the same cause.

Meta's version of that tension is sharper. In April, it raised its 2026 spending plan to as much as $145 billion, up from a prior ceiling of $135 billion^[2]. That is almost double what it spent in 2025. Meta's stock fell more than 6% within hours of the announcement^[2]. The market did not wait for proof the spending would pay off. It reacted to the size of the bet itself.

So the real question for Wednesday is not whether these companies are spending on AI. Everyone agrees they are, and agrees on roughly how much. The question is whether that money is working, and by what measure anyone could tell.

Why the Bill Comes Due Later Than the Spending Does

To understand why this argument is so hard to settle, you need to understand two accounting ideas that both sides use as weapons: depreciation and free cash flow.

When a company buys a data center full of servers, it does not subtract the full cost from this year's profit. Instead it spreads that cost over the years the equipment is expected to last, usually five to six years for servers and 25 to 40 years for buildings^[17]. This is called depreciation. It means a company can spend $100 billion in cash today while only a slice of that cost shows up as an expense on this year's earnings. Profit looks fine even while cash is going out the door fast.

That gap is exactly why bulls and bears can look at the same numbers and reach opposite conclusions. Bulls say this is just normal accounting for anything built to last decades, like a factory or a bridge. Bears say it means today's healthy profit is partly an illusion, because the true cost of this spending is still coming, spread across future years^[17]. Goldman Sachs estimates that depreciation charges across the industry will rise from 7% of revenue in 2022 to 12% by 2027, which bears read as a bill that has already been run up and simply hasn't arrived yet^[17].

Free cash flow is the pressure valve that makes this concrete. It is the cash a business generates from operations, minus what it spends on buildings and equipment. Net income can stay positive while free cash flow turns negative, if the building spree outpaces the cash coming in. When that happens, a company has to raise money from outside, through bonds or new stock, just to keep building^[15]. Big Tech companies issued more than $100 billion in bonds in the first three months of 2026 alone, more than in all of 2025 combined, and Alphabet announced an $80 billion stock sale on June 1 to help fund its AI plans^[17]. That is the evidence bears point to when they say the spending now depends on credit markets staying friendly, a condition no company controls^[15].

One Company Can Show Its Work. The Other Has to Argue It.

This is where Microsoft and Meta split, and it is a structural split, not a matter of which company is managed better.

Azure is a rental business. Outside companies pay Microsoft by the hour to run their computing on Azure's servers. When AI demand rises, it shows up almost immediately as Azure revenue, which is why Microsoft can point to a specific number, 40% growth last quarter, and 39% to 40% growth guided for the quarter now being reported, as direct proof the spending is being used^[1][7]. Microsoft's other argument is about timing: data centers take years to build and power contracts take years to sign, so a company that waits for demand to be proven before it builds has already lost the customer, because you cannot conjure computing capacity on demand^[7].

Meta has no equivalent scoreboard. It does not sell computing power to outside customers. Its payoff has to show up indirectly, through better ad targeting and content recommendations that raise what advertisers are willing to pay^[3]. That is real money, but it is much harder to isolate and prove than a cloud invoice. Meta's own framing is that AI is not a side project for it, it is the engine that already runs its core ad business, and any improvement compounds across billions of existing users without the cost of finding new ones^[3]. Meta also notes that much of its latest cost increase came from outside forces, higher prices for computer chips and added data-center costs tied to future capacity, not from its own choices^[2].

There's also a strategic reason Meta may be willing to spend without a clean scoreboard: it does not want to depend on someone else's AI platform the way it once depended on Apple's mobile operating system, before Apple changed tracking rules and cut into Meta's ad revenue^[3]. Owning the models and the hardware is, in part, a bid to never be that exposed again.

Analysts had penciled in about $87.71 billion in Microsoft revenue and $4.24 in earnings per share for Wednesday, but multiple outlets agreed the stock reaction would turn on Azure's growth number and Microsoft's first comments on 2027 spending, not on whether it beat those estimates^[7][8]. Microsoft shares are down about 21% this year, one of the weaker performers among large tech stocks, on doubts that all this spending is translating into a bigger lead over competitors^[8]. Meta was trading near $590 heading into Wednesday, with some analysts warning another spending increase could trigger a second selloff like April's^[19].

The Money Is Real. So Is the Doubt About Where It Goes.

Zoom out and the scale gets harder to argue with either way. Across Alphabet, Microsoft, Amazon and Meta combined, capital spending is projected at roughly $724 billion in 2026, climbing toward $950 billion in 2027^[5]. Most of that is already locked in through multi-year contracts for land, chips and power, so it can't be quickly reversed even if investor sentiment sours^[5].

Skeptical investors don't argue that AI is fake or worthless. Their argument is narrower and, they'd say, more measurable: that current earnings look strong partly because of the depreciation gap described above, and that the real signal is in cash, not profit. Fortune described the run-up to this earnings week as a market already "in revolt" over AI spending, after Alphabet and Tesla shares fell sharply the week before on the same worry^[5][15]. Fund managers are judged on returns measured in quarters, not the years it might take a data center to pay for itself, which gives them a direct incentive to punish spending now rather than wait^[5].

The Bill Nobody Asked to Split

There is a separate dispute that has nothing to do with shareholders at all, and it has grown large enough to actually slow construction down.

When a single data center needs new power lines or new generation capacity, utilities typically spread that cost across all their customers, not just the data center. Critics say that means an ordinary household ends up subsidizing a private facility it gets no benefit from. Time reported that electricity costs in some areas near data centers have risen as much as 267% compared with five years earlier, citing Bloomberg-reported data^[11]. More than 300 U.S. cities, towns and counties have now passed bans or moratoriums on new data-center construction^[11]. On July 14, New York Governor Kathy Hochul signed an executive order creating a statewide pause on new hyperscale data centers, and Pennsylvania, Michigan and South Carolina are weighing similar moves^[11]. A bill introduced in Congress in June 2026, the Ratepayer Protection Act, would force large customers like Microsoft and Meta to cover the full cost of grid upgrades built to serve them, rather than spreading it to everyone else^[11].

This opposition is no longer just noise. In the first three months of 2026, 75 major data-center projects worth more than $130 billion were delayed or canceled, partly because of this organized local pushback^[11]. That means the headline capex numbers from Microsoft, Meta and their peers represent an upper limit on what gets built, not a guarantee.

Two Versions of "Falling Behind"

A fourth group reframes the entire argument as a matter of national competitiveness rather than shareholder return, and it points in the opposite direction from the ratepayer complaint. This camp, prominent in right-leaning commentary, argues that computing power and the electricity behind it are strategic assets now, the way steel or shipping capacity once were, and that whoever holds the most of them shapes markets and military capability for decades^[13]. A Fox News opinion piece argues Beijing has committed more than $125 billion to AI and the energy systems supporting it, and isn't constrained by the same quarterly investor patience that disciplines U.S. companies^[13]. In this framing, local moratoriums aren't a legitimate cost dispute, they're an obstacle to American capacity, and the fix is faster permitting and clearer rules rather than bans^[14]. The Trump administration released a 28-page AI action plan on July 23 aimed at speeding up permitting, opening federal land for data centers, and expediting grid connections^[13].

The numbers support the idea that the U.S. currently holds a wide lead: Goldman Sachs estimates put U.S. AI capital spending near $700 billion in 2026, against roughly $105 billion in China^[9]. Chinese outlets covering the same gap, including the Alibaba-owned South China Morning Post, treat the American number less as a return-on-investment question and more as a competitive bar Chinese firms like Alibaba and Tencent now have to answer, even as those two companies are reportedly taking more cautious approaches than their U.S. counterparts^[9][10].

Coverage of all this splits about as cleanly along political and national lines as the arguments themselves. Right-leaning opinion writing at Fox News and the Washington Examiner largely skips the free-cash-flow debate entirely and frames the story as a race the U.S. cannot afford to lose^[13][14]. Left-leaning and centrist outlets like Time and Consumer Reports center the story on ratepayers, water use and land, with less space given to the competitive case for building fast^[11][18]. Financial outlets like Bloomberg and Fortune tend to write from the investor's chair, treating market skepticism as the neutral starting point rather than one view among several^[4][5]. None of this makes any single fact wrong. It does mean that which fields of this story a reader sees first often depends less on the news and more on which outlet they opened.

What happens Wednesday won't resolve any of it. Azure's growth rate and Meta's next capex number will move both stocks by the close of trading. But the underlying arguments, over cash, over China, over who pays for the power grid, will still be running the next morning, no matter what the earnings say.