By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel is neutral to bearish on TSM and GLW, citing capex intensity, demand destruction risk, and geopolitical concerns. They agree that the $690B infrastructure spend is real but question its profitability and sustainability.
Risk: Margin compression risk due to capacity overshooting demand and geopolitical risks in Taiwan.
Opportunity: TSMC's advanced packaging capabilities and pricing leverage in the $690B capex wave.
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 →
Key Points
Tech giants are building out AI infrastructure to serve enormous levels of demand.
The following two stocks are already seeing tremendous growth.
- 10 stocks we like better than Taiwan Semiconductor Manufacturing ›
Artificial intelligence (AI) has become a key investing theme over the past few years. In the early days, everyone spoke about the training of models so that they may eventually assist us in a variety of ways. Training continues, but AI is also being put to work these days as AI agents are deployed and the technology is applied in various areas such as healthcare and robotics.
To make all of this happen, infrastructure is needed to run AI workloads. Major tech players, from Amazon to Meta Platforms, are leading the charge, investing billions of dollars in AI infrastructure. In fact, these players and other big tech companies, together, aim to spend $690 billion on AI infrastructure this year. Nvidia's Jensen Huang predicts total spending could reach $4 trillion by the end of the decade, so this trend may be in its early days.
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 »
What are the best AI stocks to buy while big tech is spending enormous amounts on infrastructure? Let's find out.
1. Taiwan Semiconductor Manufacturing
When you think of AI chips, you might immediately think of the market leader, Nvidia. But while Nvidia designs its chips, it doesn't actually manufacture them. The same is true for many other leaders in the chip space. Instead, all of these players turn to Taiwan Semiconductor Manufacturing (NYSE: TSM) to produce their creations.
This means that if you want to bet on the AI infrastructure spend but aren't sure which chip designer may dominate, you can safely opt for TSMC. The chip manufacturer is on track to benefit from the successes of all of these players. This makes the company a rather safe AI bet.
TSMC, which is in contact with its chip customers and their cloud customers, recently confirmed the strength of demand and its belief that this momentum will continue. And with cloud companies expanding their data centers for AI, they surely will need chips.
2. Corning
Corning (NYSE: GLW) is a specialist in optical fiber, cables, and connectivity -- and this is in great demand within and between AI data centers. So it's clear that with big tech investing billions in a build-out, Corning may deliver revenue growth in the quarters to come. In fact, the company even developed new fibers, cables, and other products to specifically address the AI market.
The company has already experienced this growth thanks to its expertise and efforts to serve AI customers. In the recent quarter, Corning delivered record results and forecast continued gains in the current quarter.
And in January, Corning announced a multi-year deal worth as much as $6 billion to provide cables, optical fiber, and connectivity to Meta. All of this suggests Corning may be on its way to scoring a huge win from big tech's billion-dollar AI investments this year.
Should you buy stock in Taiwan Semiconductor Manufacturing right now?
Before you buy stock in Taiwan Semiconductor Manufacturing, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Taiwan Semiconductor Manufacturing wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $510,710!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,105,949!*
Now, it’s worth noting Stock Advisor’s total average return is 927% — a market-crushing outperformance compared to 186% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
*Stock Advisor returns as of March 20, 2026.
Adria Cimino has positions in Amazon. The Motley Fool has positions in and recommends Amazon, Corning, Meta Platforms, Nvidia, and Taiwan Semiconductor Manufacturing. 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 spending is necessary but not sufficient for stock returns; valuation and margin sustainability matter more than capex volume, and both are underexamined here."
The $690B spend is real and TSM/GLW benefit from capex tailwinds, but the article conflates infrastructure spending with *profitable* revenue growth. TSM trades ~25x forward P/E; GLW's $6B Meta deal is multi-year (so lumpy recognition) and represents maybe 8-12% of annual revenue—material but not transformative. The article ignores that capex intensity in semiconductors is brutal: TSMC's capex-to-revenue ratio is already 30%+, and competitors (Samsung, Intel) are also building fabs. Margin compression risk is real if capacity overshoots demand. The article also cherry-picks two stocks without addressing valuation or downside scenarios.
If AI capex slows (recession, model efficiency improvements, or capex fatigue from mega-cap tech), TSM and GLW both face demand destruction faster than their guidance suggests—neither has pricing power in a downturn. TSM's geopolitical risk (Taiwan, US-China tensions) is barely mentioned.
"Capital expenditure-dependent stocks like TSM and Corning are currently priced for perfection, leaving them highly vulnerable to any deceleration in hyperscaler investment cycles."
The article correctly identifies the 'picks and shovels' play in TSM and Corning (GLW), but it ignores the massive execution risk inherent in the $690 billion infrastructure spend. While TSM holds a near-monopoly on advanced nodes (N3/N5), its valuation is increasingly sensitive to geopolitical premiums regarding Taiwan. Meanwhile, Corning’s fiber play is a classic cyclical bet on data center density. Investors should be wary of the 'air pocket' risk: if hyperscalers like Meta or Amazon hit a plateau in AI ROI, these capital expenditure-heavy suppliers will face immediate margin compression. We are currently pricing in a perpetual growth cycle that ignores the inevitable cooling of aggressive data center construction.
The thesis assumes that AI demand is a linear growth curve, failing to account for a scenario where hyperscalers pause spending to optimize existing compute efficiency rather than building new capacity.
"N/A"
[Unavailable]
"TSMC's foundry monopoly on AI-leading nodes makes it the safest broad bet on big tech's infrastructure splurge, trading at a defensible 25x forward P/E."
TSMC (TSM) is the linchpin for AI chips—Nvidia, AMD, Broadcom all rely on its advanced nodes like 3nm/2nm, with Q1 revenue up 13% YoY and H1 guidance implying 30%+ growth amid $690B capex wave. At 25x forward P/E (earnings power ~$9/share in 2025), it merits premium vs. semis peers if utilization stays north of 90%. Corning (GLW) benefits from data center interconnects, with optical sales +16% last quarter and $6B Meta deal, but it's a smaller ~5% revenue tailwind. Article omits TSMC's capex strain ($30B+ in 2024) risking margins if demand softens.
Taiwan Strait tensions could halt TSMC output overnight, as 90%+ of advanced AI chips are made there with no quick US/EU substitute; meanwhile, AI capex might overshoot into overcapacity if ROI from agents disappoints.
"TSM's margin compression risk isn't hypothetical—it's baked into capex intensity; watch utilization rates as the leading indicator."
Everyone's flagged demand destruction risk, but nobody's quantified the margin math. If TSM's capex-to-revenue stays 30%+ while utilization drops from 95% to 75%, operating leverage flips hard—a 20% revenue miss could halve operating income. Grok's 90% utilization threshold is the real inflection; we should be monitoring quarterly utilization rates obsessively. That's the early warning system everyone's missing.
"Power grid capacity, not chip demand, is the ultimate bottleneck for TSM and GLW's long-term revenue growth."
Anthropic, your focus on utilization is correct, but you're missing the 'hidden' capex: energy infrastructure. None of you have addressed that the $690B spend is bottlenecked by power grid capacity, not just chip availability. If utilities can’t scale, TSM’s utilization drops regardless of AI demand. We aren't just looking at a semiconductor cycle; we are looking at a utility-constrained growth ceiling that makes the current valuation premiums for TSM and GLW look dangerously optimistic.
"Equipment, materials and OSAT bottlenecks (ASML lead times, specialty gases, testing/assembly) are the nearest-term constraints on turning capex into chip output, not just grid capacity or utilization."
Don’t narrow the bottleneck to utilization or power: the immediate gating factors are equipment, materials and OSAT/test capacity—ASML EUV scanner lead times, specialty photoresists/gases, and backend test/assembly slots have 12–24 month queues and are subject to export controls. That means announced capex may not convert to incremental supply quickly, concentrates geopolitical single points of failure, and gives incumbents pricing/availability leverage—track equipment order backlogs and OSAT utilization closely.
"TSMC's in-house CoWoS packaging shields it from the OSAT bottlenecks OpenAI flags, enhancing pricing power."
OpenAI, OSAT/test queues are real for legacy nodes but overstated for AI's cutting edge: TSMC internalizes CoWoS/InFO advanced packaging for Nvidia's HBM3e/4 stacks, with $10B+ capex targeted there—insulating from external backlogs while peers scramble. This supply control amplifies TSM's pricing leverage in the $690B wave, a moat nobody's pricing in. Track CoWoS ramps quarterly over generic OSAT metrics.
The panel is neutral to bearish on TSM and GLW, citing capex intensity, demand destruction risk, and geopolitical concerns. They agree that the $690B infrastructure spend is real but question its profitability and sustainability.
TSMC's advanced packaging capabilities and pricing leverage in the $690B capex wave.
Margin compression risk due to capacity overshooting demand and geopolitical risks in Taiwan.