By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel agrees that AI capex growth is real, but there's significant debate around its sustainability and the risks involved. Key concerns include execution risks, geopolitical frictions, and potential margin compression.
Risk: Geopolitical/regulatory frictions around Taiwan and AI exports could curb capacity utilization and compress margins for chipmakers and equipment makers.
Opportunity: AI capex growth presents a real, multi-year spending cycle with Alphabet, TSMC, and ASML sitting at plausible chokepoints.
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 →
Goldman Sachs recently predicted that artificial intelligence (AI) infrastructure spending could climb to between $920 billion and $1.4 trillion next year, up from the more than $700 billion expected to be spent this year. Those are some huge numbers, and there undoubtedly will be some nice winners in the space.
Let's look at three under-the-radar AI stock winners set to benefit from this surge in data center capital expenditures (capex).
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 »
Alphabet (NASDAQ: GOOGL) (NASDAQ: GOOG) is set to be both one of the big spenders and winners when it comes to AI infrastructure spending. The company plans to spend between $180 billion and $190 billion this year, with a significant increase in 2027. However, if there is any company that should be pushing up its capex spending, it's Alphabet.
The reason is that the company currently has a significant cost advantage with its tensor processing units (TPUs). By being less reliant on Nvidia's graphics processing units (GPUs) than its competitors, it is getting more bang for its buck with its AI infrastructure spending. This lets it train its Gemini model at significantly lower cost than peers and also save huge costs on inference. In many cases, this can also help provide it with a better return with Google Cloud, which is growing rapidly.
Alphabet's TPUs have become so well regarded that it is now allowing select customers, such as Anthropic, to place orders directly with co-developer partner Broadcom. This adds another high-margin revenue stream for Alphabet. Between this and its TPU cost advantage, this is a stock set to win from surging data center capex.
AI chip spending is now not only going up, but it is also widening. That's great news for Taiwan Semiconductor Manufacturing (NYSE: TSM). Whether the spending is going to GPUs, custom application-specific integrated circuits (ASICs) like Alphabet's TPUs, or high-performance central processing units (CPUs), this all benefits TSMC, which has a virtual monopoly in the manufacturing of advanced logic chips.
While chip designers will inevitably look to second source their manufacturing base if possible, right now they are beholden to TSMC, as it is the only foundry that has both the scale and expertise to produce advanced logic chips in mass quantities with high yields (few defects). This has made the company an integral partner with leading chip designers, who must turn to TSMC not only for help securing capacity but also for planning their entire chip roadmaps. As more chip companies fight to secure fab capacity, this benefits TSMC, which has already shown it has strong pricing power. Recent reports indicate the company will raise prices on its newer 3nm chips by 15% later this year.
This all makes TSMC one of the best stocks to own as spending on AI infrastructure continues to ramp.
While TSMC manufactures advanced logic chips, ASML (NASDAQ: ASML) provides the machines that make this possible. In fact, without its technology, there would be no AI infrastructure boom, as it is the only company in the world with extreme ultraviolet (EUV) lithography technology.
EUV machines are what make GPUs and other advanced chips possible, making ASML one of the most important companies, even though it is not a household name.
In addition to being needed in the manufacturing of advanced logic chips, these machines are also used to make high bandwidth memory (HBM), while its older DUV machines can also be used in the memory-making process. Its EUV machines cost upwards of $200 million, so these are pricey pieces of equipment, and the company is seeing robust demand from both foundries like TSMC and the big memory makers.
As AI capex continues to climb, ASML is a great under-the-radar stock to own.
Before you buy stock in Alphabet, 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 Alphabet 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 $433,268! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,259,391!
Now, it’s worth noting Stock Advisor’s total average return is 935% — a market-crushing outperformance compared to 207% 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 June 14, 2026. *
Geoffrey Seiler has positions in Alphabet. The Motley Fool has positions in and recommends ASML, Alphabet, Goldman Sachs Group, 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
"Geopolitical and cyclical risks around export controls and demand elasticity outweigh the article's monopoly narrative for TSMC and ASML."
Goldman's $1T AI capex projection by 2027 highlights clear volume tailwinds for TSMC's advanced nodes and ASML's EUV tools, while Alphabet's TPU strategy could improve Google Cloud margins. Yet the piece underplays execution risk: TSMC's 15% 3nm price hikes assume sustained demand elasticity, and ASML's $200M machines face multi-year lead times plus China export curbs that already cut off 20-30% of potential memory/foundry orders. Alphabet's capex ramp to $180-190B also risks diminishing returns if Gemini inference costs don't scale as modeled. Second-order effects include inventory digestion cycles in 2026 that could delay equipment bookings.
Even with geopolitical friction, hyperscaler AI budgets have historically exceeded forecasts, and TSMC/ASML's capacity moats could still deliver 25%+ revenue growth through 2027 if U.S. CHIPS subsidies accelerate domestic ramps.
"The AI capex wave can lift these names, but only if demand remains broad and supply tight; otherwise the rally may prove episodic."
Goldman’s $0.92–1.4 trillion AI capex forecast signals a real, multi-year spending cycle, and Alphabet, TSM, and ASML sit at plausible chokepoints in that cycle. Yet the piece glosses over key risks: the cycle is highly concentrated among a few customers and applications, so a pullback could hit badly; Nvidia still dominates acceleration, potentially narrowing Alphabet’s TPU edge; geopolitical/regulatory frictions around Taiwan and AI exports could curb capacity utilization; and price/volume dynamics (e.g., 3nm pricing) may compress margins for chipmakers and equipment makers more than they boost profits. The bullish setup is real, but durable upside depends on a smooth demand backdrop and tight supply.
Bearish view: AI capex could be episodic rather than structural; a macro downturn or inventory corrections could stall or delay the cycle, eroding the case for durable upside in Alphabet, TSM, and ASML.
"The valuation of AI infrastructure providers is currently disconnected from the uncertain timeline of enterprise-level AI revenue realization."
The article's premise relies on a linear extrapolation of Goldman's $1 trillion capex forecast, yet it ignores the law of diminishing marginal returns in AI infrastructure. While Alphabet, TSMC, and ASML are foundational, the market is currently pricing these names for perfection. Alphabet’s TPU advantage is real, but it faces significant execution risk in migrating enterprise workloads from CUDA-optimized environments. Furthermore, ASML is susceptible to geopolitical export controls that could severely truncate its addressable market in China. Investors should be wary of the 'pick and shovel' narrative; if hyperscalers fail to monetize AI at the application layer, this massive capex cycle will inevitably face a brutal correction.
The thesis assumes that AI infrastructure spending will follow traditional cyclical patterns, ignoring the possibility that AI compute is becoming a permanent, utility-like strategic necessity rather than a discretionary capital expenditure.
"The article mistakes cyclical equipment demand visibility (already priced in) for structural growth, while ignoring geopolitical fragmentation and competitive capacity additions that will compress margins by 2027."
The article conflates three different value propositions and misses critical timing/saturation risks. Yes, TSMC and ASML benefit from AI capex growth, but the article treats this as linear when semiconductor equipment cycles are notoriously lumpy—ASML's order book already reflects 2-3 years of demand visibility, so 2027 upside is priced in. Alphabet's TPU advantage is real but overstated: the article doesn't mention that hyperscalers are already diversifying (AWS Trainium/Inferentia, Meta's MTIA), eroding Alphabet's cost moat. The $1T spend forecast is also top-line noise—what matters is *margin* expansion, which compresses as competition intensifies. TSMC's 15% price increase on 3nm is a near-term tailwind, but geopolitical risk (Taiwan, China export controls) and capacity additions by Samsung/Intel are structural headwinds the article ignores entirely.
If AI capex truly accelerates to $1T+ by 2027 and TSMC remains the sole viable foundry, then current valuations for TSM (~25x forward P/E) and ASML (~35x) may actually be cheap relative to multi-year revenue CAGR and pricing power.
"ASML's current backlog does not fully capture potential 2026-2027 upside from U.S. capacity ramps."
Claude overstates how much of the 2027 cycle is already locked into ASML's order book. Multi-year lead times on $200M EUV tools mean current visibility largely reflects 2024-2025 demand; any acceleration from CHIPS-funded U.S. ramps or new AI nodes could still expand the backlog materially beyond what is priced today. The same capacity constraints that create lumpiness also limit how quickly supply can catch up if spending exceeds forecasts.
"Backlog visibility is not a guarantee of 2027 upside; potential acceleration could push orders beyond today’s backlog and export controls could shift spend, risking a mispricing of the cycle."
Responding to Claude: Even with 2–3 years of backlog visibility, the 2027 upside isn’t locked in. CHIPS subsidies or new AI node rollouts could push orders beyond today’s backlog, meaning earnings visibility could extend, not be priced in. The bigger risk is not just volume but margin compression if buyers push timing or switch to cheaper tech. Export controls could redirect EUV spend, undermining the premise that the cycle sits safely priced.
"AI compute is becoming a strategic utility, shifting the focus from application-layer ROI to absolute infrastructure availability and foundry pricing power."
Gemini’s 'utility' argument is the missing link. If AI compute is a strategic utility, hyperscalers will prioritize availability over margin efficiency, rendering the 'diminishing returns' critique moot for the next 36 months. Claude is wrong to focus on margin compression; in a supply-constrained environment, TSMC and ASML hold all the pricing leverage. The real risk isn't the application layer failing to monetize, but a systemic failure in power grid infrastructure to support this density.
"Infrastructure constraints (power, not just semiconductors) may be the binding constraint on AI capex acceleration, not application-layer monetization or margin dynamics."
Gemini's power grid constraint is real but underspecified. Data centers already consume ~4% of U.S. electricity; scaling to support $1T capex requires grid upgrades that take 5-7 years. This timing mismatch could throttle capex acceleration in 2026-2027 far more than margin compression or geopolitical friction. Neither TSMC nor ASML can overcome a power bottleneck. This deserves quantification before dismissing 'diminishing returns' as moot.
The panel agrees that AI capex growth is real, but there's significant debate around its sustainability and the risks involved. Key concerns include execution risks, geopolitical frictions, and potential margin compression.
AI capex growth presents a real, multi-year spending cycle with Alphabet, TSMC, and ASML sitting at plausible chokepoints.
Geopolitical/regulatory frictions around Taiwan and AI exports could curb capacity utilization and compress margins for chipmakers and equipment makers.