By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel is largely bearish on the $700B AI capex thesis, citing potential latency in AI adoption, margin compression due to competition, cyclicality of AI capex, and geopolitical risks such as supply chain disruptions in Taiwan.
Risk: Geopolitical risks, particularly supply chain disruptions in Taiwan, and the cyclical nature of AI capex.
Opportunity: Strong demand for AI infrastructure, supporting growth for companies like Nvidia and Broadcom.
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 →
Tech giants aim to spend nearly $700 billion in the AI build-out this year alone.
Demand for capacity is high and has resulted in explosive revenue growth for chip designers and cloud services companies.
Technology stocks, specifically the Magnificent Seven, powered the S&P 500 higher over the past three years. And though investors rotated out of some of these players earlier this year, that move may have been temporary. It's important to remember that the artificial intelligence (AI) story that drove the gains in recent years is far from over. In fact, it may just be getting started as AI today makes its way from the research lab into real-world applications.
This year represents a key moment in the story as major tech companies pour investment into infrastructure build-out -- not to serve anticipated demand, but instead to keep up with current customer commitments. Tech giants Amazon (NASDAQ: AMZN), Alphabet, Microsoft, and Meta Platforms plan on spending almost $700 billion on this ramp-up in 2026.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Against this backdrop, my prediction is that AI infrastructure stocks will crush the S&P 500 this year. Let's check out the full story.
We'll start by talking about AI infrastructure. What is it exactly? It includes all of the elements that make AI training and use possible, from chips to servers, networking equipment, and complete data centers. The tech giants I mentioned above are infrastructure players, as are chip giant Nvidia and networking powerhouse Broadcom. And these are just a few examples.
Now, let's consider what's happened so far in this AI boom. A few years ago, investors began snapping up AI stocks as they realized the technology's potential. Applied in a variety of ways and across industries and into our daily lives, AI could help companies and individuals save time and money -- and become more innovative. All of this is fantastic news for corporate earnings. The winners will be companies that develop and sell AI, as well as those that use it, and investors aimed to bet on these players as early as possible.
The first stages of AI involved the training of large language models, and infrastructure companies clearly benefited as customers rushed to them for chips, servers, and other products and services. This has helped companies such as Nvidia and Broadcom report revenue that exploded higher quarter after quarter.
But training isn't a one-time event. It's a task that will continue, and on top of that, these freshly trained models will then be put to work on real-world problems. Here, they need chips, networking equipment, and more to help them through the "thinking" process so that they can solve these problems.
As these models are applied more frequently and in more sophisticated and precise ways, companies will need more of the products and services offered by a wide range of infrastructure companies.
This whole process is starting to gain momentum right now. Amazon recently offered an example, as it spoke of its spending plans -- the company forecasts capital spending of $200 billion this year to serve the needs of Amazon Web Services (AWS) customers. AWS is the world's biggest cloud services provider, offering both AI and non-AI products and services. The company says it aims to monetize a good deal of this investment over the next two years, given current agreements with customers.
Recent messages from chipmakers, networking companies, and other data center giants have been the same: Demand for AI capacity is soaring. This suggests that these companies will continue to report explosive growth in the quarters to come. And reports we've seen so far this quarter support this idea. Alphabet, Amazon, and Meta pushed S&P 500 earnings growth to more than 27%, which would be the highest level since 2021, FactSet senior earnings analyst John Butters wrote in a May 4 note.
Meanwhile, Morningstar suggests AI stocks are trading at their biggest discount in seven years, according to a CNBC report.
A look at the price of several AI giants in relation to earnings estimates shows that valuation has indeed come down significantly.
Meanwhile, we're entering a new phase of AI growth: Today and into the future, companies are using AI, resulting in strong revenue gains for infrastructure players. And infrastructure players' major investment in capacity, based on customer commitments, suggests this revenue opportunity will be big.
This growth, along with attractive valuations, may prompt investors to get in on AI stocks -- we've already seen some of that momentum in recent weeks -- and that's why my prediction is that these infrastructure players will crush the market in 2026.
Ever feel like you missed the boat in buying the most successful stocks? Then you’ll want to hear this.
On rare occasions, our expert team of analysts issues a “Double Down” stock recommendation for companies that they think are about to pop. If you’re worried you’ve already missed your chance to invest, now is the best time to buy before it’s too late. And the numbers speak for themselves:
Nvidia:if you invested $1,000 when we doubled down in 2009,you’d have $558,200!Apple:*if you invested $1,000 when we doubled down in 2008,you’d have $55,853!Netflix:if you invested $1,000 when we doubled down in 2004,you’d have $471,827!
Right now, we’re issuing “Double Down” alerts for three incredible companies, available when you join Stock Advisor, and there may not be another chance like this anytime soon.
**Stock Advisor returns as of May 10, 2026. *
Adria Cimino has positions in Amazon and Oracle. The Motley Fool has positions in and recommends Alphabet, Amazon, Broadcom, Meta Platforms, Microsoft, Nvidia, and Oracle. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Infrastructure stocks remain the only reliable play, but only if investors prioritize cash flow efficiency over the raw, speculative capital expenditure figures currently dominating the narrative."
The article conflates massive capital expenditure with guaranteed future profitability, a dangerous assumption in capital-intensive hardware cycles. While the $700 billion spend is real, the 'build it and they will come' strategy assumes a linear path to monetization that ignores potential latency in enterprise AI adoption. If GPU utilization rates plateau or if software-as-a-service (SaaS) companies fail to show meaningful margin expansion from AI integration, we are looking at a classic supply glut. I remain bullish on the infrastructure layer, specifically Broadcom (AVGO) and Nvidia (NVDA), but only if investors focus on free cash flow conversion rather than top-line revenue growth, which is currently being artificially inflated by internal cross-subsidization among the Magnificent Seven.
The primary risk is a 'capex cliff' where tech giants, facing investor pressure for higher dividends or buybacks, abruptly throttle spending if ROI on AI-driven revenue enhancements remains elusive through 2026.
"Committed $700B capex provides 2-3 year revenue tailwinds for NVDA and AVGO, positioning them to outperform S&P by 2x in 2026 if inference monetizes as projected."
The article's $700B capex projection from AMZN, MSFT, GOOG, META for 2026 (noting inconsistency with 'this year') signals robust AI demand, fueling NVDA's Blackwell ramp and AVGO's networking surge, with Q1 earnings already showing 27% S&P EPS boost from Big Tech. Inference workloads will extend the cycle beyond training, supporting 30-50% growth for infra leaders vs S&P's 10-15%. Valuations have reset—NVDA at ~40x fwd P/E on 50%+ EPS outlook looks reasonable post-pullback. However, this assumes seamless execution amid supply constraints.
Hyperscalers' massive capex could lead to oversupply if AI ROI disappoints or enterprise adoption slows, echoing telecom bust where infra overbuild crushed returns. Power grid bottlenecks (e.g., U.S. needs 100GW+ new capacity) may delay 2026 ramps, capping upside.
"The article mistakes sustained capex for accelerating capex and assumes current valuations don't already price in the $700B spend cycle."
The article conflates two distinct phenomena: (1) elevated capex spend by hyperscalers, which is real and documented, and (2) the assumption this translates to outsized stock returns. The $700B figure is accurate but misleading—it's not incremental to prior years; it's a continuation of a trend already priced in. Nvidia trades ~30x forward P/E; Broadcom ~25x. Morningstar's 'biggest discount in seven years' is cherry-picked—these multiples remain 40-60% above 2019 levels. The article also conflates infrastructure demand with infrastructure *stock* returns. If capex is committed but margins compress due to competition (AMD gaining share, or custom chips eroding Nvidia's moat), infrastructure stocks underperform despite strong revenue. Critically missing: what happens if AI ROI fails to materialize and hyperscalers pull back in 2027-28.
If capex truly reflects locked-in customer commitments through 2027, and if inference workloads scale faster than expected, infrastructure stocks could re-rate higher. But the article provides zero evidence of 'locked-in' commitments—that's speculation presented as fact.
"The AI capex wave is cyclical and may peak in 2026, risking revenue deceleration and multiple compression even if AI adoption stays intact."
Angling for AI infrastructure to beat the S&P 500 rests on a growth narrative: AWS and the hyperscalers will feed a multi-year capex cycle, with roughly $700 billion earmarked for 2026. The implication is compelling for Nvidia, Broadcom, and cloud peers as model training migrates to deployment. Yet the bear case lurks: AI capex is highly cyclical and lumpy, and revenue upside depends on sustained AI adoption and pricing power in a capital-intensive cycle. If cloud growth slows, margins compress as capacity, power, and cooling costs rise, and valuations don’t compensate for slower top-line gains. The upside hinges on margin capture, not just revenue.
Strongest counter: AI infrastructure demand is front-loaded and highly cyclical; a peak capex cycle could arrive sooner than expected and cloud growth may decelerate, causing valuation disinflation even if longer-term AI adoption remains intact.
"The AI infrastructure thesis ignores the catastrophic tail risk of extreme supply chain concentration in Taiwan."
Claude is right to flag the 'locked-in' fallacy, but everyone is ignoring the geopolitical risk premium. If TSMC faces even minor disruptions in Taiwan, the entire $700B capex thesis for Nvidia and Broadcom evaporates regardless of software ROI. We are betting the global economy on a hyper-concentrated supply chain. The 'capex cliff' isn't just about ROI; it's about physical supply chain fragility that no one here has fully priced into their valuation models.
"Interlinked supply chain and power constraints could force hyperscalers to reallocate $700B capex away from GPUs toward energy infrastructure."
Gemini rightly elevates Taiwan risks, but connects insufficiently to Grok's power grid point—TSMC fabs guzzle power too, and U.S. CHIPS Act subsidies won't fix transmission delays needing 5-7 years (FERC data). If data centers claim 8% of U.S. electricity by 2030 (EIA), hyperscalers ration capex between GPUs and generators, hitting NVDA's 2026 Blackwell volumes hardest amid 20-30% energy cost inflation.
"Supply chain concentration risk (TSMC) outweighs power grid risk as the binding constraint through 2026."
Grok conflates two separate constraints—power and supply chain—but misses the sequencing risk. TSMC capacity hits first (2025-26), power grid bottlenecks second (2027+). If Taiwan disruption occurs before U.S. fab ramps materialize, hyperscalers can't substitute domestic capacity fast enough. Energy cost inflation matters only if chips exist to power. The $700B capex becomes stranded if TSMC stumbles before Intel/Samsung fill the gap. Nobody's modeled this timing explicitly.
"ROI and timing risks, plus policy/geopolitics, matter more than the capex headline; a disruption would delay the ramp and compress returns for NVDA/AVGO."
Gemini highlights Taiwan risk, but the argument treats it as a binary barrier rather than a timing/ROI risk. Hyperscalers can diversify suppliers and stagger capex; a disruption would delay, not destroy, the cycle. The bigger missing piece is policy/geopolitics (export controls, subsidies, semiconductor restrictions) and ROI sensitivity; if ROI cools or controls bite, the 2026 ramp could slip, compressing risk-adjusted returns on NVDA/AVGO more than the raw capex line suggests.
The panel is largely bearish on the $700B AI capex thesis, citing potential latency in AI adoption, margin compression due to competition, cyclicality of AI capex, and geopolitical risks such as supply chain disruptions in Taiwan.
Strong demand for AI infrastructure, supporting growth for companies like Nvidia and Broadcom.
Geopolitical risks, particularly supply chain disruptions in Taiwan, and the cyclical nature of AI capex.