AI’s math problem deepens ahead of key hyperscaler reports

ANTHONY SAGLIMBENE – CHIEF MARKET STRATEGIST, AMERIPRISE FINANCIAL

WEEKLY MARKET PERSPECTIVES —  October 19, 2026

AI’s math problem deepens ahead of key hyperscaler reports

Pockets of the AI trade came under severe selling pressure last week as investors reassess financing costs, competition, and the returns on infrastructure spending. For example, semiconductor and memory stocks saw weakness, while companies building and operating AI computing capacity faced greater scrutiny over their borrowing needs. During the week, selling extended into software, networking, and technology hardware, putting pressure on the broader market ahead of key tech earnings reports this week and next.

Notably, several concerns are hitting investors at once. Advances in Chinese memory technology are challenging assumptions about the competitive advantages of established suppliers. Industry executives are questioning whether capital will remain readily available to fund expansion. And the yield on CoreWeave’s debt soared last week as its bonds traded at distressed levels. Simply, investors are becoming less willing to finance growth in the current environment without a clearer path to stronger returns.

This graph is shown for illustrative purposes only.

Of course, the current interest rate backdrop is adding a complicating factor to an already complex structure of AI debt and financing agreements. And the hotter-than-expected September inflation reports last week have increased volatility around Federal Reserve tightening odds through year-end. Further, Treasury bond yields saw increased volatility last week as investors moved back and forth between seeking safety amid AI volatility and discounting future inflation and growth dynamics.

As we discussed in September, the AI math problem (i.e., aggressive spending versus unknown future profits) is also receiving more scrutiny on how money moves through the ecosystem. For example, chipmakers can invest across the AI complex, which then uses that capital to buy their processors. Thus, the same company can act as investor, supplier, and leaseholder on a data center, allowing capital committed at one stage to return as revenue at another. These arrangements can accelerate construction, but they also make it harder to distinguish independent customer demand from spending supported by suppliers’ financing. In our view, the Q3 earnings season should help investors better assess whether growing revenue and backlogs are translating into cash to help support the AI expansion, or whether ever-growing financing and debt are needed to support the buildout.

That said, real addressable demand continues to support the AI expansion. For example, Taiwan Semiconductor Manufacturing’s latest quarterly revenue exceeded its preliminary guidance, and management raised its full-year revenue growth outlook while maintaining capital spending plans. The company continues to describe AI demand as “exceeding available capacity”. However, its caution about planning for 2027 centered mainly on macroeconomic uncertainty and consumer-sensitive markets, rather than an identified slowdown in AI demand. Given the company makes most of the world’s advanced chips, its updated pulse check on the AI theme is important to note.

In addition, Anthropic reportedly told prospective investors that Q3 revenue reached $16.8 billion and that the company was profitable on an adjusted basis with positive cash flow. In our view, the update supports the case for adopting paid AI, but it still leaves open questions about development costs, infrastructure payments, and future compute commitments.

Yet, strong demand in “pockets” of AI doesn’t mean every company funding the buildout will earn attractive returns over time. According to FactSet, semiconductor and equipment earnings are expected to grow roughly +130% year-over-year in Q3, compared with just +14% for software. Simply, equipment spending generates revenue for the supplier today before the buyer necessarily earns an attractive return using it. We believe strong semiconductor results can therefore coexist with increasing scrutiny on the companies financing and operating the AI infrastructure buildout. That’s why we believe the upcoming earnings reports and outlooks from the hyperscalers this week and next week could help investors further evaluate the customer side of that equation.

For example, Meta Platforms reports results on Wednesday, followed by Microsoft, Alphabet, and Amazon next week. We believe investors will focus on whether hyperscaler monetization is keeping pace with spending and on how management intends to finance the next stage of expansion. For Microsoft and Amazon, we believe cloud growth should be assessed alongside the cash required to deliver additional capacity. Corporate guidance should explain whether spending addresses demand already committed by customers or builds ahead of anticipated needs. Faster revenue growth would carry greater weight if accompanied by a credible path to improving cash generation, which has been declining in recent quarters.

For Meta Platforms and Alphabet, we believe the discussion should also connect AI investment to advertising performance, engagement, and operating efficiency. According to FactSet, Interactive Media & Services earnings are expected to rise +44% year-over-year in Q3 on revenue growth of +24%. Of course, those estimates encompass the industry’s overall operations rather than AI alone. Yet, we believe management commentary should help investors assess how much AI contributes to those improvements and whether the benefits justify additional investment.

We believe larger backlogs will require more explanation on earnings calls. Multi-year commitments can support future revenue expectations but also require substantial AI spending before additional cash flows arrive. As these companies report, analysts and investors will want more details on delivery schedules, customer payments, and the funding required to fulfill those contracts. Power availability and construction timelines will also come into focus on hyperscaler calls.

In addition, financing capacity could increasingly separate companies across the ecosystem as AI scrutiny grows. In our view, businesses funding expansion from existing operations (which admittedly is increasingly difficult today given the size of spending) should have greater flexibility in a rising-rate environment than those dependent on repeated borrowing or equity issuance. Additionally, wider credit spreads for some AI borrowers (see CoreWeave and Oracle as examples) are already creating differentiation across the ecosystem.

Importantly, we expect analysts to ask the hyperscalers more pointed questions about spending discipline on earnings calls. Given the five major hyperscalers (Microsoft, Alphabet, Amazon, Meta Platforms, and Oracle) are forecast to spend over $1 trillion in capex combined over the next 12 months, we doubt management teams will get a free pass this earnings season from explaining why they can’t slow their spending down in order to better protect their balance sheets. Of course, commentary and color here will likely be consequential for earnings expectations across the AI supply chain, particularly for semiconductors and hardware manufacturers. Any reassessment in hyperscaler spending could also reach beyond technology into industrials, power infrastructure, and the lenders financing the AI buildout.

Bottom line: Last week's AI volatility and the upcoming hyperscaler reports lead back to the same AI math problem. Over the past year, consensus capex estimates for the largest AI spenders have more than doubled. Their revenue estimates have grown by roughly a quarter. In our view, that gap can't stay this wide indefinitely. Either revenue catches up more quickly, or spending plans get reset.

So far, returns have held up. Most of these companies still earn strong returns on their invested capital, which is why investors have been willing to fund the buildout (granted, much of those returns came from money spent before the current wave).

The test now is whether those returns hold (and probably need to rise) as spending continues to climb. Notably, corporate guidance that ties AI spending to visible revenue and delivers steady returns on capital should be rewarded. Spending plans that keep rising without that link will likely meet a tougher reception from investors. We believe AI remains a long-term driver of earnings and economic growth, but investors are becoming more selective about which companies get rewarded from here. For investors, that argues for owning AI exposure selectively rather than broadly, keeping technology concentration in check, and diversifying across market caps, sectors, and regions. Expect volatility around these reports, but we would advise not letting it derail longer-term allocation decisions.

The week ahead:

• Third-quarter earnings ramp up with GE Aerospace, General Motors, RTX, Netflix, and SpaceX on Tuesday. Meta Platforms, Tesla, IBM, and Lam Research report Wednesday, followed by Honeywell, Lockheed Martin, and Union Pacific on Thursday.
• Economic data includes September housing starts and pending home sales on Tuesday, plus flash October PMIs and final October University of Michigan consumer sentiment on Friday.
• The Federal Reserve is now in its quiet period ahead of the October 28th FOMC meeting. Markets have reduced the odds of an October rate hike to roughly 16% but still see a December hike as very likely.

These figures are shown for illustrative purposes only and are not guaranteed. They do not reflect taxes or investment/product fees or expenses, which would reduce the figures shown here. An index is a statistical composite that is not managed. It is not possible to invest directly in an index. Past performance is not a guarantee of future results.