By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panel agrees that rising AI token costs and heavy capex by hyperscalers pose near-term risks, but they disagree on the long-term impact. While some panelists see potential for margin improvement and revenue growth, others warn of stranded asset risk and commoditization of inference.
Risk: Stranded asset risk due to reduced AI workloads and utilization problems
Opportunity: Pricing power and multi-product bundling to sustain revenue growth
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
For a few years now, artificial intelligence (AI) has been the market’s undisputed growth engine, powering massive gains for tech giants and fueling relentless investor optimism. However, a growing number of strategists are beginning to question whether the AI boom can maintain its current pace, with rising AI token costs emerging as a potential headwind for the sector. Among the skeptics is Wells Fargo (WFC) Chief Equity Strategist Ohsung Kwon, who recently warned that the market’s AI-fueled surge may be entering a more challenging phase.
“The sugar high rally is likely over,” Kwon said, describing the recent market pullback as a reminder that “nothing goes up every day, forever.” While he stressed that the current tech selloff was initially driven more by market positioning than weakening fundamentals, he pointed to what he sees as the most immediate risk to AI stocks is the end of “token-maxxing.” The concern centers on rising AI token costs as AI labs scale back subsidies that had previously helped keep usage expenses in check.
As those costs climb, companies ranging from Walmart (WMT) to Uber (UBER) have begun signaling that their AI budgets could be consumed much faster than expected. According to Kwon, any slowdown in AI spending could have significant implications for the broader investment theme that has fueled many of the market’s biggest winners. The challenge arrives at a particularly difficult moment for hyperscalers.
Companies such as Microsoft Corporation (MSFT), Meta Platforms (META), Alphabet (GOOG) (GOOGL), and Amazon (AMZN) are already ramping up capital expenditures amid supply-chain constraints and mounting cost inflation. As infrastructure spending climbs, these companies may be forced to pass higher costs on to customers just as businesses grow increasingly sensitive to AI-related expenses.
Reflecting those concerns, Wells Fargo has moved from a bullish stance in April to a “firmly neutral” outlook this month, with Kwon saying the firm is “unenthused about this market.” So, with concerns about rising costs, slowing demand growth, and mounting spending commitments beginning to surface, here’s a closer look at two hyperscalers that could be particularly vulnerable if the AI trade starts to lose momentum.
Stock #1: Microsoft Corporation
Founded in 1975 by Bill Gates and Paul Allen, Microsoft has evolved from a small software startup into one of the most influential technology companies on the planet. Headquartered in Redmond, Washington, the tech giant has shaped the modern digital era through iconic products such as Windows, Office, Teams, and Xbox, platforms that have become deeply embedded in the way people work, communicate, and entertain themselves.
Over the decades, Microsoft expanded well beyond its software roots, making a bold push into cloud computing with Azure, which has become a cornerstone of digital transformation for businesses worldwide. Today, the company is doubling down on artificial intelligence, weaving AI-powered capabilities throughout its ecosystem in an effort to boost productivity, streamline workflows, and redefine the ways users interact with technology.
Yet despite Microsoft's dominant position and relentless innovation across both cloud and AI, investors have become increasingly cautious. Shares have faced mounting pressure over the past year as the company commits tens of billions of dollars toward expanding its AI infrastructure, raising concerns about the near-term payoff from those massive capital expenditures. The recent broader tech selloff has only added to the stock's struggles, weighing further on investor sentiment.
While Microsoft remains one of the world's most valuable companies, boasting a market capitalization of roughly $2.95 trillion, its stock performance has failed to keep pace with the broader market. Shares have fallen 17.9% over the past year, sharply trailing the 21.27% gain delivered by the broader S&P 500 Index ($SPX). The weakness has persisted in 2026, with MSFT down another 19.8% year-to-date (YTD), compared to the broader market's 6.7% advance.
Microsoft’s fiscal third-quarter 2026 earnings report, released on April 29, showcased a company firing on all cylinders operationally, even as Wall Street focused on a growing concern. The tech giant delivered revenue of $82.9 billion, up an impressive 18% year-over-year (YOY) and comfortably ahead of analysts’ expectations of $81.4 billion. Leading the charge was Azure and other cloud services, which posted remarkable growth of 40% YOY, or 39% in constant currency, underscoring the continued strength of enterprise demand for Microsoft's cloud and AI offerings.
Growth was broad-based across the business. The Intelligent Cloud segment, home to Azure, server products, GitHub, and Nuance cloud services, generated $34.68 billion in revenue, marking a 30% increase from the prior year. Meanwhile, the Productivity and Business Processes division, which includes Office, LinkedIn, and Dynamics, delivered $35.01 billion in revenue, up 17% YOY. On the bottom line, Microsoft's earnings power remained equally impressive, with adjusted EPS climbing 21% annually to $4.27, easily surpassing Wall Street's consensus estimate of $4.07.
However, the headline-grabbing aspect of the quarter wasn't Microsoft's strong financial performance, it was the enormous price tag attached to its AI ambitions. The company spent a staggering $31.9 billion on capital expenditures during the quarter alone, a level of spending that continued to weigh on free cash flow. More notably, management stunned investors by forecasting roughly $190 billion in CapEx for calendar year 2026, while guiding fourth-quarter spending to exceed an eye-popping $40 billion in a single quarter.
That massive investment plan overshadowed an otherwise stellar report and sparked concerns about how long Microsoft can sustain such aggressive spending before seeing stronger returns. Investors responded cautiously, sending the stock down roughly 4% following the earnings release despite the company's clear top- and bottom-line beat. Still, management highlighted that its AI business surpassed a $37 billion annual revenue run rate during the quarter, soaring 123% YOY. Microsoft also continued rewarding shareholders, returning $10.2 billion through dividends and share repurchases during the third quarter of fiscal 2026.
Despite Microsoft's recent struggles on Wall Street, analysts remain overwhelmingly optimistic about the tech giant's long-term prospects. The stock currently carries a consensus “Strong Buy” rating, reflecting continued confidence in its leadership across cloud computing and AI. Among the 49 analysts covering the stock, 41 rate it a “Strong Buy,” three recommend “Moderate Buy,” and only five remain on the sidelines with a “Hold” rating.
The average analyst price target of $554.28 implies potential upside of 43.12% from current levels, while the Street-high target of $680 points to a potential gain of as much as 75.6%, highlighting analysts' confidence that Microsoft's AI and cloud investments could eventually translate into significant shareholder returns.
Stock #2: Meta Platforms
What started as a college social networking website called Facebook has grown into one of the world's most powerful technology companies. Founded in 2004 by Mark Zuckerberg and headquartered in Menlo Park, California, Meta Platforms has fundamentally reshaped the way billions of people connect, communicate, share content, and engage with the digital world through platforms such as Facebook, Instagram, WhatsApp, and Messenger.
Over the years, Meta has evolved far beyond its social media roots, pouring billions of dollars into technologies it believes will define the next era of computing. From virtual and augmented reality initiatives housed within its Reality Labs division to sophisticated AI systems that power content recommendations, advertising, and user engagement, the company has consistently positioned itself at the forefront of innovation.
Today, Meta sits at the crossroads of social media, digital advertising, artificial intelligence, and immersive technologies, making it one of the most closely watched players in the tech sector. However, despite its aggressive push into AI and its dominant position across multiple digital markets, investors have become increasingly cautious.
While advertising remains Meta's core profit engine, the company's massive investments in AI infrastructure and future technologies have raised questions about spending levels and long-term returns. Currently valued at approximately $1.44 trillion, Meta, much like Microsoft, has struggled to keep pace with the broader market. Shares have fallen 18.94% over the past year and are down another 14.76% so far in 2026, highlighting growing investor concerns even as the company continues to invest heavily in its next chapter of growth.
Meta’s fiscal first-quarter 2026 earnings report, released on April 29, delivered exactly what investors have come to expect from the social media giant. Explosive growth, massive profits, and an even bigger commitment to artificial intelligence. The company crushed Wall Street’s expectations on both revenue and earnings, with sales surging 33% YOY to $56.31 billion from $42.3 billion a year earlier. That marked Meta’s fastest quarterly revenue growth since 2021 and comfortably topped analysts’ estimates of $55.49 billion.
The impressive performance highlights how CEO Mark Zuckerberg’s AI investments are already strengthening the company’s advertising machine, even before generating meaningful standalone AI revenue. The growth story extended well beyond the top line. Family Daily Active People (DAP) reached 3.56 billion on average during March 2026, up 4% from a year ago. Meanwhile, ad impressions delivered across Meta’s Family of Apps jumped 19% YOY, underscoring the growing effectiveness of the company’s AI-powered recommendation and advertising systems.
Profitability was even more impressive. Earnings per share soared 62% from the prior year to $10.44, while adjusted EPS of $7.31 easily surpassed Wall Street’s consensus forecast of $6.71. Yet despite the blockbuster quarter, investors found a reason to worry. Rather than celebrating Meta’s accelerating growth, the market focused on the enormous costs associated with the company’s AI ambitions. Shares slid 8.55% on April 30 as investors digested another sharp increase in spending plans.
Capital expenditures, including principal payments on finance leases, totaled a staggering $19.84 billion during the quarter, reflecting Meta’s aggressive buildout of AI infrastructure and data center capacity. The company also returned $1.35 billion to shareholders through dividends and dividend-equivalent payments while maintaining a fortress-like balance sheet with $81.18 billion in cash, cash equivalents, and marketable securities as of March 31, 2026.
However, the biggest concern was management’s updated outlook. Meta reaffirmed its full-year 2026 expense guidance of $162 billion to $169 billion but raised its capital expenditure forecast to between $125 billion and $145 billion, up from its previous range of $115 billion to $135 billion. Management cited higher component costs and additional data center investments needed to support future AI capacity as the primary drivers of the increase.
Even with Meta's recent struggles on the stock chart, Wall Street's conviction in the social media and AI giant remains remarkably strong. The stock currently carries a consensus “Strong Buy” rating. Among the 54 analysts covering the stock, 42 recommend “Strong Buy,” two rate it a “Moderate Buy,” nine maintain a “Hold,” and only one analyst has issued a “Strong Sell” rating.
That overwhelming optimism is reflected in analysts' price targets. The average target of $826.78 implies potential upside of 47.6% from current levels, while the Street-high target of $1,015 suggests Meta could surge as much as 81.2%, highlighting Wall Street's belief that the recent weakness may be temporary rather than a sign of deteriorating fundamentals.
On the date of publication, Anushka Mukherji did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on Barchart.com
Four leading AI models discuss this article
"MSFT and META capex commitments of $190B and up to $145B risk multiple compression unless AI monetization outpaces the token-cost headwind already visible in customer commentary."
Wells Fargo flags rising token costs pressuring enterprise budgets just as MSFT guides $190B 2026 capex and META lifts its range to $125-145B. Both stocks already lag the S&P 500 by wide margins (MSFT -17.9% past year, META -18.9%), and the 4% and 8.55% post-earnings drops show investors are pricing in slower ROI on that spend. Yet Azure grew 40% and Meta revenue 33% with AI already lifting ad impressions 19%, so any demand cliff would have to overcome still-accelerating usage metrics.
Enterprises facing unsubsidized token pricing could simply throttle usage or delay projects far more than current guidance assumes, turning the $37B MSFT AI run-rate and Meta's ad gains into temporary blips rather than durable growth.
"Long-run AI-driven cloud demand and platform monetization should absorb higher token costs, preserving the uptrend."
While Wells Fargo flags a near-term risk from rising AI token costs and heavier capex by hyperscalers, this reads like a cautionary macro note rather than a durable thesis on demand. The strongest case against a 'death knell' scenario is that AI is becoming embedded in mission-critical workflows, with enterprise budgets expanding for efficiency gains, new revenue models, and stickier cloud-native workloads. Even if unit costs rise, pricing power, multi-product bundling, and efficiency gains can sustain revenue growth and potentially improve margins over time as data-center utilization improves. The piece understates accelerants like deeper platform monetization and rising AI workloads that broaden long-run demand.
Token-cost pressure could choke usage growth, forcing pricing concessions or slower capex. That would dent near-term momentum and margin resilience.
"The current market selloff in hyperscalers is a temporary reaction to front-loaded infrastructure costs that ignores the long-term operating leverage these firms will capture as inference efficiency improves."
The market is currently mispricing the transition from 'AI experimentation' to 'AI infrastructure amortization.' While Wells Fargo is right to highlight the 'token-maxxing' cost wall, they overlook the massive operating leverage inherent in these hyperscalers. Microsoft (MSFT) and Meta (META) are not just spending on CapEx; they are building proprietary Moats that will eventually commoditize the very compute costs currently squeezing their margins. The 19.8% YTD drop in MSFT is a classic overreaction to temporary margin compression. When these firms shift from building data centers to optimizing inference efficiency, the current 'death knell' narrative will flip, likely driving a massive re-rating as free cash flow recovers in late 2027.
If AI model costs fail to scale down through software optimization, these firms face a permanent structural shift toward lower margins that makes their current valuation multiples unsustainable.
"Rising AI costs are a demand-side headwind for customers, not yet a supply-side margin squeeze for hyperscalers, but capex ROI remains unvalidated and represents the real execution risk."
The article conflates two separate problems and overstates their severity. Rising AI token costs are real, but they're a *customer* problem (Walmart, Uber), not a hyperscaler problem—those costs flow to end-users, not to MSFT or META's P&L. The actual concern is capex intensity: MSFT's $190B 2026 guidance and META's $125-145B represent 15-20% of revenue, historically unsustainable. But the article ignores that both companies are generating $37B+ and $56B+ quarterly revenues with 30%+ cloud/AI growth. The math isn't broken yet—it's just unproven. Wells Fargo's pivot to 'firmly neutral' is a positioning call, not evidence of fundamental deterioration.
If token costs force customers to cut AI budgets sharply, hyperscalers lose pricing power on inference workloads and face utilization pressure on newly built capacity—turning capex into stranded assets faster than revenue can grow.
"Customer throttling from token costs will hit hyperscaler utilization faster than revenue offsets the capex build-out."
Claude correctly separates customer token costs from hyperscaler P&L but misses the direct feedback loop: if Walmart or Uber throttle AI workloads to stay under budget, Azure and Meta's inference utilization falls on the very capacity built with $190B and $125-145B capex. That turns the 40% Azure and 33% revenue growth into a utilization problem, not just a customer one, raising stranded-asset risk by late 2026 rather than 2027.
"Token-cost pressure need not erode asset value if hyperscalers monetize AI via APIs/services and redeploy capacity, offsetting utilization drag."
The stranded-asset risk is plausible but overstated if you assume linear scaling. Hyperscalers monetize AI through APIs, software, and managed inference, which can preserve margin even with token-cost pressure. Also, capacity can be redeployed to high-margin enterprise workloads or edge deployments, mitigating late-2026 utilization drag. The real risk is persistent near-term margin compression from token costs and capex intensity, not a wholesale collapse in asset value; beware assuming a clean re-rating only by 2027.
"Hyperscalers are building depreciating utilities rather than moats, risking long-term margin compression as inference becomes a commodity."
Gemini’s 'moat' thesis ignores the commoditization of inference. If software optimization is the only path to margin recovery, we are betting on a race to the bottom where hyperscalers compete on price against open-source models like Llama. This isn't a moat; it's a margin-crushing cycle. If MSFT and META cannot demonstrate proprietary value-add beyond raw compute, they aren't building moats—they are building expensive, depreciating utilities that will struggle to command premium multiples.
"Inference commoditization threatens margins only if bundled software value-add fails to justify premium pricing; that's testable by 2026 earnings, not inevitable."
Gemini's commoditization risk is real, but the panel hasn't addressed pricing power asymmetry: MSFT bundles inference into enterprise suites (Copilot, M365); META monetizes via ad targeting. Both extract value *above* raw token cost. Open-source Llama doesn't compete there. The stranded-asset risk materializes only if enterprises abandon AI entirely—plausible but requires demand destruction, not just cost pressure. That's the actual binary.
The panel agrees that rising AI token costs and heavy capex by hyperscalers pose near-term risks, but they disagree on the long-term impact. While some panelists see potential for margin improvement and revenue growth, others warn of stranded asset risk and commoditization of inference.
Pricing power and multi-product bundling to sustain revenue growth
Stranded asset risk due to reduced AI workloads and utilization problems