By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panelists generally agreed that the article overstates the benefits and understates the risks of AI chip investments. While Nvidia, Broadcom, and Amazon stand to gain from AI demand, their valuations already reflect much of the potential upside, and there are significant risks related to stretched valuations, ROI, regulatory headwinds, and shifts in architecture.
Risk: The risk that hyperscaler ROI on AI spend falls short, triggering capex cuts, and the potential impact of escalating US export curbs on advanced nodes to China on Nvidia's data-center revenue.
Opportunity: The potential for hyperscalers to reduce inference costs via custom silicon, improving ROI and expanding cloud margins.
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 →
3 Chip Designers That Could Make Investors a Fortune
Keithen Drury, The Motley Fool
4 min read
I think the best-positioned companies in the artificial intelligence (AI) build-out are the chip designers. These are companies that design chips but have no hand in manufacturing. This allows them to scale up and down as necessary, and they only have to worry about making the best product possible and ensuring their suppliers produce what's necessary.
Should AI demand deteriorate, this protects these companies from having too much invested in production facilities. However, AI demand is still rising, and it could continue to do so for some time.
Will AI create the world's first trillionaire? Our team just released a report on a little-known company, called an "Indispensable Monopoly," providing the critical technology Nvidia and Intel both need.
I've got three companies that can take advantage of this trend, and even though they've all been great performers so far, they still have massive upside potential ahead.
1. Nvidia
Nvidia(NASDAQ: NVDA) is the chief chip designer, and its GPUs are the primary computing unit of choice for nearly every AI hyperscaler. GPUs are flexible, accelerated computing nodes that can handle a wide variety of workloads, giving them an advantage over custom AI chips designed around a specific workload. Nvidia GPUs hold the largest market share in the AI computing realm by far, and this has boosted Nvidia all the way to the top to become the world's most valuable company.
While Nvidia's run has been impressive, it's far from over. Management believes that global data center capital expenditures will rise to $3 trillion to $4 trillion annually. Considering the big four AI hyperscalers are planning on spending around $650 billion this year, that's a major jump. However, that's a global figure and doesn't include many of the upstarts like OpenAI and Anthropic.
As mentioned above, one of the rising alternatives to Nvidia GPUs is custom-designed AI chips. The idea behind these is to tailor them to a workload to maximize cost efficiencies. Still, many AI hyperscalers don't have the expertise necessary to produce custom AI chips, which is where Broadcom(NASDAQ: AVGO) comes in. By partnering with Broadcom, AI hyperscalers have a direct line for everything necessary to design and produce their own chips, and the market for these is expected to boom over the coming year.
Multiple clients have custom AI chips near launch, which should drive Broadcom's custom AI chip revenue to over $100 billion in 2027. For reference, Broadcom's company-wide revenue over the past 12 months was less than $70 billion, and AI chips were only a fraction of that total.
Broadcom is positioned to take advantage of this market shift, making it a smart investment.
3. Amazon
Many companies partner with Broadcom because they don't have the design and production expertise necessary to bring a custom AI chip to life. Amazon(NASDAQ: AMZN) didn't use Broadcom for its chips; it just went out and hired the necessary talent to do it. Amazon's custom AI chips are thriving and are growing at a triple-digit rate.
Furthermore, the demand is so high for its products that its third-generation chip, which started shipping at the start of 2026, has nearly all of its capacity spoken for. It's fourth generation, not due out for another 18 months, and has most of its computing capacity already spoken for.
Amazon's custom chip business dovetails nicely into its leading cloud computing platform, Amazon Web Services (AWS). By keeping all of these businesses within one unit, rather than partnering with another, it's maximizing its profits. Because cloud revenue is rising, this makes Amazon the most stable investment pick of the three, and if AWS revenue continues to climb, so will Amazon's stock price as a result.
Should you buy stock in Nvidia right now?
Before you buy stock in Nvidia, 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 Nvidia 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 $469,293!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,381,332!*
Now, it’s worth noting Stock Advisor’s total average return is 993% — 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.
Keithen Drury has positions in Amazon, Broadcom, and Nvidia. The Motley Fool has positions in and recommends Amazon, Broadcom, and Nvidia. The Motley Fool has a disclosure policy.
Four leading AI models discuss this article
"Nvidia faces material share erosion and capex disappointment risks that the bullish projections ignore."
The article pushes chip designers Nvidia, Broadcom, and Amazon as prime AI beneficiaries thanks to their asset-light models and custom ASIC tailwinds. Yet it glosses over stretched valuations—NVDA trades above 30x forward sales—and the risk that hyperscaler ROI on AI spend falls short, triggering capex cuts well before the cited $3-4 trillion annual run rate materializes. Broadcom's $100 billion custom-chip target by 2027 also assumes flawless execution across multiple clients, while Amazon's in-house chips may simply shift profits inside AWS rather than expand the total pie.
Nvidia's CUDA lock-in and software advantages have repeatedly defied forecasts of share loss, so even meaningful custom-ASIC adoption could still leave it with dominant economics in a much larger market.
"The article's $100B Broadcom custom chip forecast by 2027 lacks any verifiable sourcing and appears to be the linchpin holding up the entire bull case—without it, the thesis collapses to 'these companies are already winning.'"
The article conflates three very different theses into one. Nvidia's dominance is real but already priced in at $3.4T market cap—the $3-4T capex claim doesn't automatically flow to proportional GPU revenue given custom chips' rising share. Broadcom's $100B custom AI chip revenue by 2027 is speculative; the article provides zero evidence (no client names, no signed contracts, no timeline detail). Amazon's chip business is real but represents a rounding error in AWS—claiming it 'dovetails nicely' glosses over the fact that AWS revenue is $90B+ annually while custom chips remain nascent. The piece reads like three separate bull cases duct-taped together rather than a coherent investment thesis.
If custom AI chips genuinely capture 40-50% of hyperscaler spend by 2027 (plausible given cost pressures), Broadcom and Amazon could see explosive growth that justifies the optimism. Nvidia's moat may be weaker than the article assumes.
"The 'fabless' model shifts capital risk away from fixed assets but concentrates it into a single point of failure: the dependency on limited foundry capacity and the unproven long-term ROI of AI-driven data center spending."
The article's 'fabless' thesis—that chip designers avoid manufacturing risk—is dangerously reductive. While designers like Nvidia (NVDA) and Broadcom (AVGO) avoid the massive capex of owning fabs, they are entirely captive to the supply chain bottlenecks and yield rates of TSMC. The assumption that AI demand is a permanent, linear 'build-out' ignores the looming risk of diminishing marginal utility; as hyperscalers spend hundreds of billions, they must eventually prove that these chips generate commensurate revenue growth. If AI model performance plateaus or enterprise adoption stalls, we risk a massive inventory glut that will crush the margins of these high-flying designers far faster than the article suggests.
If the hyperscalers are essentially building their own proprietary infrastructure, the 'moat' for these designers may be wider than I fear, as switching costs for custom silicon and software stacks like CUDA create extreme vendor lock-in.
"The 'indispensable monopoly' narrative is unproven and AI compute demand is likely cyclical, with capex, competition, and margin pressures enough to cap upside."
While the article highlights Nvidia as a GPU king, Broadcom as an enabler of custom AI chips, and Amazon as an in-house challenger, it treats AI demand as a near-inevitable tailwind. Nvidia remains dominant, but its valuation already prices in continued exponential growth. The Broadcom angle hinges on a surge of hyperscalers outsourcing chip design—a bet that could disappoint if customers pursue in-house development or if margins compress. Amazon’s in-house chips add optionality but aren’t a cash-flow jackpot yet. A major omission is the risk of a data-center capex slowdown, regulatory headwinds, or shifts to alternative architectures that could blunt earnings trajectories.
Nvidia’s moat and the multi-year AI adoption cycle could sustain strong growth even if near-term demand cools; pricing power and ecosystem leverage may keep margins expanding and the stock outperforming regardless of some deceleration.
"Nvidia's TSMC capacity reservations create a supply advantage that offsets some custom-chip risks."
Gemini flags TSMC dependency for fabless players but misses how Nvidia's multi-year wafer reservations secure priority access during shortages. That allocation edge could preserve its margins and market share even if custom ASICs from Broadcom and Amazon scale. The bigger unmentioned risk is escalating US export curbs on advanced nodes to China, which would hit Nvidia's data-center revenue harder than the others given its current exposure.
"Hyperscaler ROI on AI capex, not chip supply dynamics, is the real pivot point the article and panel have underweighted."
Grok's wafer-reservation edge is real but time-bound. TSMC's capacity expansion and competing demand from non-AI sectors erode that advantage by 2026-27. More critical: nobody's addressed whether hyperscalers' ROI math actually works at scale. If a $10B capex generates only $2-3B incremental revenue annually, the entire thesis collapses regardless of who supplies the chips. That's the load-bearing assumption nobody's pressure-tested with actual unit economics.
"Hyperscaler capex is driven by the necessity of lowering inference costs to protect long-term cloud margins, not just speculative revenue growth."
Claude is right to fixate on the ROI, but the panel is missing the 'utility' shift. Hyperscalers aren't just buying chips; they are building sovereign AI clouds. If Amazon or Microsoft can reduce inference costs by 40% via custom silicon (Trainium/Inferentia), the ROI math changes from 'revenue generation' to 'operating margin expansion.' The capex isn't just for speculative growth; it’s a defensive move to commoditize Nvidia’s stack and protect long-term cloud margins.
"ROI and unit economics are the true gatekeepers; without margin expansion or scalable incremental revenue, the capex-driven AI-infra boom collapses."
Claude’s ROI critique nails a core risk, but it understates two levers. First, margin uplift from cloud-infrastructure efficiency (lower energy, better utilization) can unlock sustained ROI even if realized revenue is lumpy. Second, a 40-50% share of hyperscaler AI spend by 2027 seems aggressive risk-adjusted; if it lands closer to 15-25%, capex returns compress quickly. The thesis rests on margin expansion and unit economics—not just top-line share gains.
The panelists generally agreed that the article overstates the benefits and understates the risks of AI chip investments. While Nvidia, Broadcom, and Amazon stand to gain from AI demand, their valuations already reflect much of the potential upside, and there are significant risks related to stretched valuations, ROI, regulatory headwinds, and shifts in architecture.
The potential for hyperscalers to reduce inference costs via custom silicon, improving ROI and expanding cloud margins.
The risk that hyperscaler ROI on AI spend falls short, triggering capex cuts, and the potential impact of escalating US export curbs on advanced nodes to China on Nvidia's data-center revenue.