Big Tech's AI Reckoning: This Earnings Week Could Decide The Next Leg Of The Market

For the better part of two years, one question drove nearly every major technology stock: how much is the company spending on AI? Beginning this week, investors may start asking a harder one: when does that spending actually produce a return?

The shift showed up the moment Alphabet and Tesla reported on Wednesday after the close. Both stocks sold off sharply on results that, in almost any other year, would have read as solid. What sank them wasn't revenue, margins, or guidance. It was capital expenditure. Alphabet raised its expected 2026 capex to as much as $205 billion – money going to data centers, networking gear, and custom chips – and instead of rewarding the ambition, investors knocked the stock down about 7 percent. Tesla drew the same treatment: Elon Musk talked up autonomy and robotics, and the market looked past all of it to weakening cash flow and another year of heavy spending, sending the shares down roughly 14 percent. The damage spread from there, with the Magnificent 7 shedding $787 billion in market value and the Nasdaq posting its worst session since April 2025. Wall Street is now sobering up to the fact that the era of unlimited AI spending may be ending.

That makes the coming week unusually consequential. Microsoft, Meta, Amazon, and Apple are all on deck – Meta goes Wednesday after the close – and for perhaps the first time since the boom began, investors care less about who is spending the most and more about who is actually making money from it.

The Charts Say One Thing. The Positioning Says Another.

Technically, the Magnificent 7 look bruised – the group just printed one of its largest down candles in memory and is trading back around levels last seen in September, with several names sitting on long-term trendlines that count as must-hold. Alphabet fell straight into that zone after earnings. Amazon has no clear medium-term trend left. Tesla is testing the line that has supported it since the 2024 lows.

But positioning tells a far less bearish story. Hedge funds spent the run-up cutting Magnificent 7 exposure aggressively, and by the start of earnings season the long/short ratio on the group had fallen to roughly its 25th percentile since 2018. Net exposure has been reset and gross exposure materially reduced. In plain terms, the fast money already left. The group heads into these prints looking under-owned rather than overcrowded – which is usually the setup for an upside surprise, not continued bleeding.

That doesn't resolve the debate. It sharpens it. If the reports are clean, lightly positioned investors have to chase; if the capex guides spook them again, there's less of a crowd left to do the selling. The caveat: Thursday's carnage happened with positioning already this light, which means the sellers were long-onlys, retail, and passive flows – slower money, but stickier once it turns. Either way, the reflexive "next leg down" call is harder to make than the charts alone suggest.

Show Me The Money

The question is no longer whether AI will transform the economy. Most investors still believe it will. The question is narrower and matters far more for valuations: who earns an acceptable return on the enormous amount of capital now being deployed?

Until now the market accepted an implicit bargain – spend today, monetize later, dominate eventually – and that bargain justified the firehose of capital aimed at GPU clusters, custom silicon, nuclear-powered data centers, fiber, and hyperscale cloud. The whole structure rested on one assumption: that demand for AI compute would keep growing fast enough to absorb almost any level of investment. 

The hyperscalers have collectively committed hundreds of billions toward AI infrastructure, and nearly every quarter brings another increase in that guidance. At some point those investments have to start expanding margins rather than shrinking free cash flow. Alphabet just previewed the alternative: free cash flow went negative for the first time since the company came public in 2004 – $44.9 billion of quarterly capex against $39.1 billion of operating cash flow – buybacks have sat at zero for two straight quarters after a 33-quarter run, and June's equity raise means shareholders are now funding the buildout directly. If that becomes the pattern, the market may stop valuing these companies like software firms and start valuing them like capital-intensive utilities – and that distinction is worth a great deal. Software earns premium multiples because each additional dollar of revenue costs almost nothing to produce. A data center packed with hundreds of thousands of power-hungry GPUs is a fundamentally different, and far more expensive, business. And it is utilities, not software companies, that issue equity to pour concrete.

Meta is the cleanest test of whether that spending converts. The company is building visible paths to monetize AI directly rather than only through better ad targeting: consumer and creator subscriptions priced from around $8 to $50 a month, roughly 10 million weekly conversations between users and business agents across its messaging apps – up about tenfold since the start of the year – and a reported move into selling raw compute and hosted-model access, the kind of cloud business that would attach direct revenue to infrastructure investors currently value only through its indirect benefit to advertising.

That said, is the circle jerk just intensifying? Reports indicate Anthropic is in early talks to lease roughly $10 billion of compute from Meta, one AI company renting capacity from another. Google, meanwhile, told analysts it plans to lease third-party capacity of its own – including compute from SpaceX, which now owns xAI and its Memphis data centers. Everybody is simultaneously landlord and tenant. Whether any of it moves the needle against a capex line this size is exactly what the buy side will be listening for.

The Questions Investors Want Answered

Management teams should expect a consistent line of questioning: how much more capex is coming in 2027, when AI products will move revenue in a material way, whether enterprise customers are paying enough to justify the infrastructure, how fast margins can recover, and whether demand is actually keeping pace with all the capacity being built.

The 2027 numbers will be a huge focus. After Alphabet's June equity raise, buy-side expectations for its 2027 capex moved up into the $325-to-$350 billion range, from $275-to-$300 billion before. Amazon is modeled around $300 billion. Meta's 2027 figure is now widely expected in the low-to-mid $200 billions, with the debate centered on whether the company guides to it this quarter at all.

Not to be total doomers: the headline capex number likely overstates the actual cash burden. A growing share of that capacity is partner-financed rather than funded directly off the balance sheet, and internal, lower-cost silicon is starting to bring down the blended cost per unit of compute. Cutting the other way: Alphabet's contracted purchase commitments hit $811 billion at the end of June – up nearly $500 billion in a single quarter, with roughly $200 billion of it due within a year – obligations that sit outside the capex line entirely. The bill is enormous either way. But "how it gets funded" is now as important as "how big it is." 

Why IBM Is The Cautionary Tale

The clearest evidence that this is a spending story, not a demand story, came from the company that spent too little. IBM suffered its largest single-day decline since records begin in 1968 – worse than Black Monday – after missing on revenue.

The reason, per CEO Arvind Krishna's letter to shareholders, was almost poetic: in the final weeks of June, IBM's own customers diverted their quarterly technology budgets toward servers, storage, and memory, racing to lock in supply-constrained AI hardware before prices rose. Large deals slipped. Infrastructure revenue fell.

IBM got punished because its clients were spending on AI infrastructure instead of on IBM. Set that against Alphabet, marked down the same month for spending too much, and the vise comes into focus: there is currently no capital-allocation posture a large-cap technology company can adopt that the tape won't find a reason to sell. Spend and you're burning cash; hesitate and you're being left behind.

Nvidia Still Holds The Center Of Gravity

Nvidia reports later in the season, but every call this week will be judged through the lens of its supply chain. Every dollar Microsoft, Meta, Amazon, or Alphabet commits to infrastructure ultimately feeds demand for Nvidia GPUs and the broader semiconductor complex. The reverse is just as true: any sign the hyperscalers are easing off would ripple straight through chipmakers, equipment suppliers, utilities, data-center developers, and power infrastructure. That's why this stretch matters well beyond a handful of tech tickers. It's effectively a referendum on the largest capital investment cycle in modern corporate history.

None of this necessarily marks the end of the AI boom. The technology remains one of the most significant shifts in decades, and the long-term case is hard to dispute. But markets don't run on long-term visions. They run on cash flow. For two years, investors rewarded companies for announcing ever-larger AI budgets. Beginning this week, they may start rewarding something else entirely: proof that the spending can throw off durable profits.

Alphabet and Tesla fired the opening salvo. The reports from Microsoft, Meta, Amazon, and Apple may decide whether the AI trade regains its footing, or whether Wall Street concludes that "growth at any cost" has finally hit its limit. The charts say the trend is broken. The positioning says the crowd already left. This week we find out which one was right.