By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panel expresses concerns about execution risks, unit economics, and cyclical nature of AI capex, suggesting a potential correction in infrastructure spending. They also highlight the risk of margin compression due to custom silicon competition and energy constraints.
Risk: Lack of ROI for firms buying AI chips, leading to a correction in infrastructure spending
Opportunity: None explicitly stated
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 →
Investing opportunities like artificial intelligence (AI) don't come around all that often. With the major data center infrastructure build-out that's happening, I'm confident in calling this a generational investment opportunity, and I think five stocks offer the best ways to take advantage: Nvidia(NASDAQ: NVDA), Broadcom(NASDAQ: AVGO), Micron(NASDAQ: MU), Nebius (NASDAQ: NBIS), and CoreWeave(NASDAQ: CRWV).
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.
Investors may be growing tired of hearing about Nvidia, but it truly is one of the best ways to play the AI build-out. Its revenue growth has been consistently high, and it has showcased time and time again why its products are the best available. As for the future, Nvidia believes that global data center capital expenditures could rise from $600 billion in 2025 to between $3 trillion and $4 trillion annually by 2030.
Nvidia's leadership has more information about its sales outlook than the average investor does, so it's worth paying attention to bold projections like this. While the actual dollar figure for data center spending that year may not land in that range, the direction Nvidia sees the market headed in is probably correct. Given that it's still providing a large share of the high-end processors that this industry relies on, it's a must-own investment.
Broadcom
Broadcom is a similar investment to Nvidia, as it also provides AI computing hardware. However, it's coming at the data center market from a different angle. Instead of designing broad-purpose GPUs like Nvidia, it's partnering directly with hyperscalers to design AI chips that are customized to handle the specific workloads the buyers expect them to see.
This is a massive and growing business, and Broadcom's management team believes that segment could drive $100 billion in revenue in 2027. That would represent major growth for Broadcom, and if custom AI chips continue to be a hot commodity, the stock is primed to soar even further.
Micron
AI servers require massive amounts of high-bandwidth memory in close proximity to their processors to function properly, and currently, there aren't enough memory chips to go around. That has put makers of memory chips such as Micron in a strong position, allowing them to sharply boost their prices. Meanwhile, their input costs are staying relatively flat. The result: Micron's revenue and profits have skyrocketed.
While all the memory makers are building new foundries to boost output, those factories take time to bring online. Moreover, with the demand for memory chips expected to triple from 2025 to 2028, this shortage will persist for some time even as production grows. That bodes well for Micron stock, and with Wall Street expecting 193% revenue growth for the rest of this year and 57% growth next year, it has a lot of upside left.
Nebius
Nebius has the fastest-growing top line on this list. In Q1, its revenue skyrocketed a jaw-dropping 684%. It's growing that quickly because it's building out a neocloud platform that is tailored to AI. Nvidia is a major investor in Nebius and also a key component supplier.
As more companies scramble to bring AI computing capacity online, and as a shortage of data center capacity persists, Nebius will continue to grow in popularity, as its full-stack setup is quite popular among developers. Nebius has massive expansion plans, and its revenues will grow rapidly for some time as it builds out its computing footprint, which should lead to more upside for the stock.
CoreWeave
CoreWeave is similar to Nebius in that it's a neocloud operator and is focused on providing AI computing resources through the cloud. But instead of offering a full-stack setup, it's focused on providing GPU training banks for clients that need more computing capacity. This caters to a different need, but that need is still massive.
Like Nebius, it's rapidly growing -- revenue rose 112% year over year in Q1 to $2.1 billion. It has a nearly $100 billion backlog, underscoring how much demand there is for its product as it brings more data centers online. This bodes well for CoreWeave's future. If it can continue to build out its capacity and work its way toward profitability, it could be a top AI stock to own for the next decade.
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 $481,589!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,345,714!*
Now, it’s worth noting Stock Advisor’s total average return is 993% — 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.
Keithen Drury has positions in Broadcom, Nebius Group, and Nvidia. The Motley Fool has positions in and recommends Broadcom, Micron Technology, and Nvidia. The Motley Fool has a disclosure policy.
Four leading AI models discuss this article
"The $3-4T annual data center spend projection by 2030 is the linchpin assumption that, if revised lower, would simultaneously undermine all five highlighted stocks."
The article's bullish case rests on Nvidia's $3-4T data center capex forecast by 2030 and rapid revenue ramps at Broadcom, Micron, Nebius, and CoreWeave. Yet it downplays execution risk for the two neocloud names, which still lack sustained profitability, and ignores how quickly custom ASIC competition and memory supply ramps could compress margins. Hyperscaler ROI scrutiny on AI spend is rising; any pause in 2026-27 budgets would hit all five names simultaneously. Valuations already embed perfection, leaving little margin for the inevitable delays in fab builds and power constraints the piece never mentions.
Even if capex falls short of Nvidia's upper bound, the direction remains strongly upward and the five stocks could still compound at above-market rates for several years given structural demand.
"Demand for AI infrastructure is real, but the article mistakes top-line growth for sustainable returns—it ignores margin compression, competitive fragmentation, and cyclical commodity risk across the stack."
The article conflates *demand* with *profitability*. Yes, AI capex is exploding—Nvidia's $3-4T projection by 2030 is directionally credible. But the piece cherry-picks five stocks without addressing brutal unit economics: Nebius grew 684% YoY in Q1 yet remains unprofitable; CoreWeave's $100B backlog means nothing if margins compress as capacity floods online. Micron's memory shortage is real but cyclical—fabs come online in 2026-27, and memory is notoriously commoditized. The article also buries that Broadcom's custom-chip strategy only works if hyperscalers don't vertically integrate further (they are). Nvidia's dominance is real, but at current valuations (~30x forward P/E), you're pricing in perfection.
The strongest case against: data center capex growth doesn't linearly translate to chip vendor revenue if (1) custom silicon accelerates, fragmenting Nvidia's TAM, and (2) memory oversupply arrives faster than expected, cratering Micron's pricing power by 2027.
"The current AI investment thesis relies on an unsustainable extrapolation of infrastructure spending that ignores the inevitable cyclicality and ROI pressures facing hyperscalers."
The article conflates infrastructure spending with sustainable enterprise profitability, ignoring the 'AI CapEx cliff.' While Nvidia (NVDA) and Broadcom (AVGO) are clear beneficiaries of current hyperscaler build-outs, the market is pricing in perpetual 50%+ growth that is historically unsustainable. Micron (MU) is cyclical; its current 'shortage' narrative ignores the inevitable supply glut once new capacity hits the market. Furthermore, highlighting private entities like CoreWeave alongside public equities creates a false sense of liquidity and risk parity. The real risk is not a lack of AI demand, but a lack of ROI for the firms buying these chips, which will eventually force a brutal correction in infrastructure spending.
If AI agentic workflows achieve widespread enterprise adoption in 2025, the current infrastructure spending may prove to be merely the 'foundational' phase of a multi-trillion dollar productivity cycle, making current valuations look cheap in hindsight.
"The AI data-center boom will likely reward durable, cash-flow-rich leaders (like NVDA and AVGO) rather than speculative near-term entrants (NBIS/CRWV), and valuations should reflect that risk."
AI capex is real and Nvidia remains the core lever, but this article cherry-picks five names, including two with dubious public-market visibility. The strongest risk is that data-center demand is highly cyclical and price competition could compress margins; memory cycles may soften, and Micron’s strength hinges on near-term pricing power. Nebius (NBIS) and CoreWeave (CRWV) are less transparent in fundamentals and may not deliver commensurate profitability or liquidity. A $3–4 trillion annual data-center spend by 2030 is a directional target, not a guaranteed outcome; sentiment could flip if AI software adoption or energy costs disappoint.
The strongest case against a neutral stance is that the AI hardware cycle could prove cyclical and volatile; if software adoption stalls or capex slows, Nvidia and Broadcom could see margin compression while Nebius and CoreWeave struggle with liquidity and profitability.
"Custom ASIC adoption timing could compress Nvidia before memory oversupply hits Micron."
Claude flags custom silicon as a margin threat but underplays the timeline: if hyperscalers shift even 15% of training workloads to in-house ASICs by late 2026, Nvidia's revenue ramp flattens before Micron's supply glut arrives. That sequencing matters more than isolated unit-economics concerns because it hits the highest-multiple name first and could trigger simultaneous de-rating across the entire basket.
"Custom ASIC risk is real but timeline-dependent; margin compression via supplier diversification is the more immediate threat than revenue displacement."
Grok's sequencing argument is sharp, but assumes hyperscalers can deploy custom ASICs at scale by late 2026—a 18-month window that historically favors incumbents with proven yield and software stacks. More likely: custom silicon captures 5-8% of incremental training by 2027, not 15%. The real threat is *optionality*—hyperscalers' willingness to diversify suppliers, which erodes Nvidia's pricing power before volume growth slows. That's a margin story, not a revenue cliff.
"The erosion of Nvidia's software-driven pricing power via custom silicon will compress valuation multiples well before revenue growth actually hits a wall."
Claude, you’re underestimating the 'software moat' argument. Hyperscalers aren't just building chips; they are building alternative software stacks to bypass CUDA. Even a 5% shift to custom silicon is a massive signal of pricing power erosion, not just a minor margin compression event. If Nvidia loses its 'must-have' status in the training layer, the valuation premium evaporates long before the revenue actually drops. The risk isn't just volume; it's the terminal value multiple.
"A 5-8% shift to custom silicon by 2027 underplays the software moat and could trigger margin compression from energy costs and capacity growth, threatening the valuation premium."
Claude, arguing 5-8% of incremental training goes custom silicon by 2027 understates the software moat and the momentum of integration around CUDA; even limited share shifts can suppress pricing given tooling lock-in and multi-vendor strategies. The bigger risk is margin compression from energy/cooling constraints and faster-than-expected capacity additions that outpace GPU demand growth, not just supplier diversification. This could tilt risk toward the valuation premium even if revenue stays robust.
The panel expresses concerns about execution risks, unit economics, and cyclical nature of AI capex, suggesting a potential correction in infrastructure spending. They also highlight the risk of margin compression due to custom silicon competition and energy constraints.
None explicitly stated
Lack of ROI for firms buying AI chips, leading to a correction in infrastructure spending