By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel is bearish on the long-term prospects of AI hardware providers due to power constraints, geopolitical risks, and overinflated valuations. They agree that the 'AI buildout spending' may not translate into durable earnings growth for suppliers.
Risk: Power constraints and geopolitical risks (Taiwan discount, export curbs)
Opportunity: None identified
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 →
The AI chip industry encompasses far more than Nvidia, the Wall Street darling.
Chip foundries, special production equipment, and even proprietary designs have all benefited from the AI boom.
These four companies could all grow earnings by more than 20% annually over the coming years.
The artificial intelligence (AI) build-out continues. AI hyperscalers will spend an estimated $725 billion this year, much of it on data centers and the chips and hardware that power them.
In discussions about AI stocks, Nvidia gets a lot of the credit, and for good reason. The graphics processing unit (GPU) juggernaut has been the runaway leader, selling the overwhelming majority of the AI GPU chips used in data centers over the past few years. But several other companies make AI possible and don't quite get the love they should.
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 »
These four semiconductor stocks are quietly capturing tons of value as the AI build-out continues. They are likely to continue delivering exceptional growth over the coming years, so consider taking a closer look.
Nvidia and most other chip companies don't actually manufacture their own products. Instead, they turn to foundries like Taiwan Semiconductor Manufacturing (NYSE: TSM). Taiwan Semi is the world's leading foundry, and by a wide margin. It holds approximately 72% of the global foundry market by revenue, and it has the most capacity and cutting-edge manufacturing techniques.
Taiwan Semi and the broader semiconductor industry have historically been cyclical, but the current AI boom is so immense that the company continues to grow. Remember, even outside of AI, almost every technology uses chips at some level. Taiwan Semi has reported stellar business results over the past few years, and analysts expect the company to continue growing earnings at an average annual rate of 22% over the next three to five years.
Nvidia will continue to face competition as the AI market swells. Broadcom (NASDAQ: AVGO) was already in data centers with its networking chips, but it has hit a home run with its XPUs: custom AI accelerator chips that it can design for customers' specific needs. That customization unlocks efficiency and gives AI hyperscalers an alternative chip supplier.
Broadcom's XPU customers include Alphabet, OpenAI, and Anthropic. CEO Hock Tan believes that the company's AI chip revenue could surge past $100 billion next year. It's a major growth catalyst for a business that earned $63.8 billion in revenue in 2025. The AI surge has juiced Broadcom's growth outlook. Analysts estimate the company will grow earnings at a blistering annualized rate of 49% over the next three to five years.
Look deeper into the AI chip supply chain, and you'll find ASML Holding (NASDAQ: ASML). Simply put, ASML builds and sells specialized equipment that Taiwan Semiconductor and other foundries need to manufacture chips. What's special about ASML is that the Dutch company is the world's only supplier of extreme ultraviolet lithography machines, which print remarkably tiny patterns on chips. ASML has a literal monopoly on this technology.
ASML has had an up-and-down few years; it got caught up in geopolitical tensions regarding export controls between the United States and China. These headwinds still linger a bit, but that hasn't stopped ASML's growth, as the need for AI investments continues to grow. The company reported strong first-quarter 2026 earnings on AI momentum, and analysts see ASML growing earnings by an average of nearly 30% annually over the next three to five years.
At the deepest level of the AI ecosystem is Arm Holdings (NASDAQ: ARM). Arm specializes in instruction set architecture, a chip's foundational language, for central processing units (CPUs) and other chips used in almost every type of electronic device. Arm's customers have shipped over 350 billion chips to date, creating a massive installed base of technology. Countless devices, from iPhones to personal computers, use chips with Arm's designs, and Arm collects royalties and fees on each chip its customers ship.
Recently, Arm announced it was expanding its business, unveiling its own AI CPU chip to capitalize on soaring demand for AI CPUs. Arm expects the chip to enter full production later this year, providing quite a growth catalyst over the next three to five years as it ramps up. Wall Street analysts estimate that Arm will grow earnings by an average of almost 25% annually over that period.
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 $477,813! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,320,088!
Now, it’s worth noting Stock Advisor’s total average return is 986% — a market-crushing outperformance compared to 208% 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 May 26, 2026. *
Justin Pope has positions in Alphabet. The Motley Fool has positions in and recommends ASML, Alphabet, Broadcom, 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 capacity risks are underpriced relative to the aggressive earnings forecasts embedded in these four names."
The piece correctly notes that TSM, AVGO, ASML and ARM sit deeper in the AI stack than Nvidia and carry 22-49% consensus EPS growth forecasts. Yet it glosses over TSM’s concentrated Taiwan risk, ASML’s ongoing China export curbs, and the fact that AVGO’s custom XPUs still depend on TSMC capacity that could be rationed. Arm’s new AI CPU also enters a market already dominated by AMD and Intel designs. With $725B in 2026 capex already priced into multiples above 30x forward earnings, any AI spending pause would compress these names faster than Nvidia itself.
Even if geopolitics or capex fatigue hit, the structural need for leading-edge silicon remains so acute that the 25-30% growth rates still materialize and multiples hold.
"Supplier stocks benefit from AI capex volume, but the article's 20-50% earnings CAGRs assume no margin compression, no geopolitical disruption, and full realization of nascent product launches—each a material assumption."
The article conflates 'AI buildout spending' with durable earnings growth for suppliers. Yes, $725B capex flows through the chain—but foundry margins are compressing (TSMC gross margin ~54% vs. historical 60%+), and custom chips (Broadcom XPUs) cannibalize higher-margin GPU sales. The 49% earnings CAGR for Broadcom assumes $100B+ AI chip revenue materializes; that's speculative. ASML's EUV monopoly is real, but geopolitical export controls remain a live tail risk. Arm's CPU play is nascent—royalty models don't scale like product revenue. The article treats supplier optionality as certainty.
If hyperscalers truly face GPU supply constraints and Broadcom/custom silicon genuinely reduces Nvidia's TAM, then these suppliers capture real margin expansion—not just volume. The article may be underweighting how much of this $725B flows to non-Nvidia players over 3-5 years.
"The current growth projections for these semiconductor leaders are highly sensitive to hyperscaler CapEx sustainability, which remains unproven beyond the short term."
The article correctly identifies the 'picks and shovels' of the AI gold rush, but it ignores the massive valuation premiums already baked into these names. For instance, ARM is trading at forward P/E multiples often exceeding 80x, pricing in perfection that leaves zero margin for error in royalty growth. While TSM remains the indispensable foundry, investors are downplaying the geopolitical risk premium—specifically the 'Taiwan discount'—which could trigger a valuation contraction regardless of earnings growth. Broadcom and ASML are strong, but their success is contingent on hyperscalers maintaining current CapEx intensity. If AI ROI remains elusive, these hardware providers will face a brutal de-rating as growth expectations reset to reality.
If the AI buildout represents a paradigm shift in computing infrastructure rather than a cyclical bubble, these companies are effectively the new 'utilities' of the digital age, justifying premium valuations as long-term essential infrastructure.
"The next 12–24 months will prove whether these four beneficiaries can sustain double-digit earnings growth, contingent on continued, visible AI capex from hyperscalers."
Good take on the broader AI supply chain, but the article leans into inevitability. The strongest counter: AI capex is likely front-loaded, so 2026 earnings visibility hinges on continued hyperscaler demand and capex persistence. If growth slows—through a slower AI compute cycle, memory constraints, or policy/geopolitics limiting fab upgrades—TSM, ASML, Broadcom, and Arm face margin and growth risks. Broadcom’s XPUs could face competition and pricing pressure; ASML’s EUV capacity remains a bottleneck; Arm’s royalty model hinges on a widening install base amid IP competition. Valuations may already reflect optimistic growth.
Counterpoint: AI capex durability could surprise to the upside; hyperscalers have a multi-year need for capacity, and these players have entrenched, long-duration contracts that cushion margins. If capex stays robust, the article's thesis about 20%+ growth across four names could actually play out.
"Power grid constraints represent a binding limit on AI capex deployment that could derail supplier earnings growth before valuations or geopolitics do."
Gemini's valuation warning overlooks a structural limit: hyperscalers' AI rollouts are hitting power availability walls, with data center electricity demand projected to double by 2026. This caps utilization for TSM and ASML capacity even if capex holds, pressuring the 25%+ growth assumptions across the board. The Taiwan and export risks are secondary to whether the chips can actually be powered at scale.
"Power constraints delay, not prevent, near-term supplier growth—but create a hard cliff in 2027-2028 that valuations aren't pricing."
Grok's power constraint is real but overstated as a *cap* on growth. Hyperscalers are actively solving this—renewable energy PPAs, on-site nuclear, efficiency gains. The bottleneck isn't 2026 capex; it's 2027-2028 when power becomes the binding constraint. TSM and ASML still see 20%+ growth through 2026 because capex *frontloads* before power walls hit. The article's 2026 thesis survives; the real de-rating risk is 2028 when growth rates normalize sharply.
"Power grid constraints are currently forcing hyperscalers to throttle data center deployment, creating an immediate ceiling for hardware growth that the market is ignoring."
Claude is missing the immediate impact of the power wall. It isn't a 2028 problem; it is already causing hyperscalers to delay data center commissioning. If power availability forces a shift from high-density GPU clusters to more efficient, specialized silicon, the 'picks and shovels' thesis shifts. Broadcom’s custom silicon gains, but TSM’s foundry utilization risk increases if the total compute volume is throttled by grid capacity. We are pricing in peak utilization that the grid cannot sustain.
"Near-term grid/power constraints could throttle hyperscaler utilization faster than Claude expects, causing earlier margin pressure in 2026."
Claude’s 2027–28 power constraint timing assumes steady utilization; I’d push a more urgent reading: grid and power bottlenecks are already delaying data-center commissioning, not just pushing it out. If hyperscalers pivot to more power-efficient AI silicon sooner, the near-term margin dynamics for TSM, ASML, and Broadcom could deteriorate faster than your thesis implies. That makes 2026 a real risk year for multiple expansion, not just a back-end adjustment.
The panel is bearish on the long-term prospects of AI hardware providers due to power constraints, geopolitical risks, and overinflated valuations. They agree that the 'AI buildout spending' may not translate into durable earnings growth for suppliers.
None identified
Power constraints and geopolitical risks (Taiwan discount, export curbs)