By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel consensus is that while the $500B+ AI capex by GOOGL, AMZN, and META underscores Nvidia's (NVDA) near-term dominance, the long-term sustainability of this growth is uncertain due to potential shifts towards custom ASICs and the cyclical nature of semiconductor demand.
Risk: The shift towards custom ASICs by hyperscalers, which could significantly compress Nvidia's gross margins and growth rate.
Opportunity: The durable data-center expansion driven by AI capex, which suggests a strong near-term outlook for Nvidia.
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 →
Company updates this year put fresh numbers on the AI (artificial intelligence) build-out, and they are enormous. Alphabet raised its 2026 capital spending forecast to a range of $195 billion to $205 billion, up from $180 billion to $190 billion. Amazon has said it expects to invest about $200 billion this year. And Meta Platforms plans $125 billion to $145 billion, a range it lifted by $10 billion in April.
Add it up, and just three companies intend to spend more than half a trillion dollars in a single year, most of it on AI infrastructure. And that tally leaves out Microsoft, which has pointed to about $190 billion of its own.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »
No company collects more of that spending than Nvidia (NASDAQ: NVDA), the dominant supplier of the graphics processing units (GPUs) those data centers are built around. Yet Nvidia stock fell on Thursday alongside other big tech stocks, and it now sits about 12% below its 52-week high.
Customers committing record sums while the supplier's stock drifts lower? That's a disconnect worth examining, because one side of it is probably wrong.
The budgets are not all chips, but a large share of the money goes where Nvidia lives. Alphabet, for instance, said on itsearnings callthat about 60% of its technical infrastructure investment in the second quarter went to servers, with the rest going to data centers and networking equipment. The company also raised $49.6 billion in a June stock offering, with scaling AI infrastructure among the stated uses.
The flow shows up directly in Nvidia's results. In its first quarter of fiscal 2027 (the period ended April 26, 2026), revenue rose 85% year over year to a record $81.6 billion. Data center revenue climbed 92% to $75.2 billion, and total revenue rose 20% from the prior quarter as well. And the company guided for about $91 billion in revenue in its fiscal second quarter, all while holding its gross margin near 75%.
"The buildout of AI factories -- the largest infrastructure expansion in human history -- is accelerating at extraordinary speed," said CEO Jensen Huang in the company's fiscal first-quarter earnings release.
In other words, the customers' budgets and the supplier's income statement are telling the same story, and Alphabet's raise this past week extended it into the second half of 2026.
One possible explanation is that the market punished the spenders on Thursday. Alphabet's shares fell after its capital spending raise, and the sell-off spread across the megacaps, Nvidia included. When investors start doubting whether half a trillion dollars of AI spending will earn its keep, they also start discounting the revenue that spending creates -- and a large share of it lands on Nvidia's income statement.
That's the risk to hold in mind. Nvidia's growth is a direct function of a handful of customers' willingness to keep writing bigger checks. Budgets that accelerate for three years can also flatten, and the semiconductor industry has never escaped its cycles for long. A capital budget is a plan, not a contract, and plans built during a boom can get rewritten quickly.
The same customers are also working to need Nvidia a little less. Meta said on its first-quarterearnings callthat it is rolling out more than a gigawatt of its own custom silicon, developed with Broadcom, alongside chips from Advanced Micro Devices -- complementing, for now, the new Nvidia systems it keeps installing.
With all of that said, the valuation asks less than investors might assume. Nvidia trades at about 32 times trailing earnings -- a multiple many slower-growing consumer companies carry -- for a business that just grew 85%. The market, in effect, is already pricing in a meaningful slowdown. Against expected earnings for the next 12 months, the multiple drops to about 21.
So does Thursday's sell-off make Nvidia the way to own the build-out? I think it remains the most direct claim on those budgets, and at this valuation I'd keep owning it. But I'd size the position for what it is: a stock whose earnings depend on a handful of customers' capital budgets -- and capital budgets get revisited every year.
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 $377,990! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,269,518!
Now, it’s worth noting Stock Advisor’s total average return is 896% — a market-crushing outperformance compared to 206% 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 July 27, 2026. *
Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Broadcom, Meta Platforms, Microsoft, and Nvidia. 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
"Nvidia remains the clearest near-term beneficiary of hyperscaler AI spend, but its valuation already discounts a material deceleration that custom silicon and cyclical capex risk could accelerate."
The article's $500B+ 2026 AI capex from GOOGL ($195-205B), AMZN (~$200B), and META ($125-145B) underscores Nvidia's (NVDA) near-term dominance, with data-center revenue already at $75.2B (+92% YoY) and 75% gross margins. At 32x trailing/21x forward earnings amid 85% growth, the valuation already embeds slowdown fears. However, this ignores that much of the spend is multi-year infrastructure with long lead times, not pure GPU procurement, and hyperscalers are aggressively developing custom ASICs (Meta's Broadcom silicon, Amazon's Trainium/Inferentia).
Capex plans are not binding contracts; if ROI on AI models disappoints in 2026, these budgets can be slashed overnight, exposing Nvidia to a classic semiconductor downturn far sharper than the modest multiple compression already priced in.
"Nvidia's valuation is currently pricing in a soft landing for AI investment, yet it remains hyper-exposed to the binary outcome of hyperscaler ROI on their $500 billion spend."
The $500 billion aggregate CAPEX figure is staggering, but investors must distinguish between 'infrastructure spend' and 'AI monetization.' While Nvidia’s 75% gross margins are currently protected by an unmatched moat in CUDA software and H100/Blackwell demand, we are witnessing a classic 'arms dealer' scenario. The hyperscalers are essentially subsidizing the AI ecosystem, hoping for a killer app to emerge. If these companies don't see a material inflection in cloud revenue or ad-targeting efficiency by late 2026, the 'build-out' will hit a wall. Nvidia’s 21x forward P/E looks attractive, but it assumes the current CAPEX trajectory is sustainable, which ignores the cyclical nature of semiconductor demand.
The 'cyclical' argument ignores the structural shift where compute has become the primary unit of production for the global economy, rendering historical semiconductor cycles irrelevant.
"Nvidia's valuation assumes capex growth persists, but the real risk is not capex slowdown—it's capex reallocation away from Nvidia toward AMD and custom silicon, which would compress growth rates and multiples simultaneously."
The article conflates two separate questions: whether the capex is real (it is) and whether it will generate returns (unclear). Nvidia's 21x forward P/E on 85% growth looks reasonable in isolation, but masks a critical dependency: these budgets are concentrated among four customers who collectively control Nvidia's fate. The article notes this risk but undersells it. Meta is already deploying custom silicon at scale—not as a future threat but as current fact. If even one mega-cap customer shifts 20-30% of GPU volume to AMD or custom chips over 18-24 months, Nvidia's growth rate compresses from 85% to mid-30s, and that 21x multiple becomes 35-40x on normalized earnings. The sell-off Thursday may reflect this realization, not irrational panic.
Nvidia's gross margin at 75% and pricing power in a supply-constrained market suggest customer lock-in is real; even if custom silicon gains share, it complements rather than replaces Nvidia chips, and the total TAM is expanding fast enough to absorb both.
"The AI capex wave is likely to slow and reallocate, and Nvidia’s high valuation already prices in a never-ending growth tail that may never arrive."
AI capex by Alphabet, Amazon, and Meta suggests a durable data-center expansion, but the hype may overstate persistence. Capex cycles are notoriously front-loaded and price-sensitive; macro surprises or a shift to in-house silicon (as Meta and others pursue) could curb Nvidia’s addressable GPU demand. Much of the spend is on servers, networking, and data-center capacity, not exclusively GPUs, and fierce competition from AMD, Google TPUs, and in-house accelerators could compress pricing and margins. Valuations imply strong, ongoing growth; a cooling cycle or worse-than-expected AI ROI could re-rate NVDA quickly.
Even with a slowdown, hyperscalers tend to default to top GPUs due to ecosystem lock-in and software advantages, so Nvidia could retain a durable revenue base even in a softer capex cycle.
"Custom silicon is already substituting inference workloads, pressuring NVDA margins and multiples faster than consensus expects."
Claude undersells how custom ASICs already erode Nvidia's pricing power today. Meta's MTIA and Broadcom silicon deployments aren't complements; they're direct substitutes for inference workloads that comprise 60%+ of future GPU demand. If one hyperscaler diverts even 25% of 2026 capex, NVDA's 75% gross margins compress to low-60s, turning 21x forward into 28-30x normalized.
"The primary risk to the $500B capex thesis is power infrastructure constraints, which will throttle compute growth regardless of whether hyperscalers use Nvidia GPUs or custom ASICs."
Grok and Claude are fixated on the GPU-vs-ASIC substitution, but both ignore the physical constraint: power. Even if hyperscalers pivot to custom silicon, the bottleneck isn't just chip architecture—it's the massive, multi-year energy infrastructure required to run these clusters. If power availability lags, the $500B capex plan becomes a stranded asset problem, not a margin compression problem. Nvidia’s risk isn't just competition; it’s the physical ceiling on data center capacity expansion.
"Power scarcity is a shared constraint, not a Nvidia-specific hedge against ASIC competition."
Gemini's power constraint is real but incomplete. The energy bottleneck affects all players equally—hyperscalers face it whether deploying Nvidia GPUs or custom ASICs. What matters: can they solve it faster than competitors? If TSMC/Samsung can't scale fab capacity to feed custom silicon demand, Nvidia's supply advantage persists despite margin pressure. The constraint isn't a moat; it's a shared ceiling that favors incumbents with existing allocation.
"Rapid in-house silicon adoption by mega-clients could hollow out Nvidia's pricing power faster than expected, risking margin downgrades and meaningful multiple compression."
Grok's substitution worry is valid but underweights speed and scale. If Meta/AWS/others front-load in-house silicon, the share shift could be faster than anticipated, not just a margin squeeze. The immediate risk isn't only price erosion but a rapid chunk of GPU demand vanishing from the external market. If a 20–25% reallocation occurs by 2026, NVDA's gross margin could slip from ~75% toward the low-60s, and the forward multiple may compress meaningfully.
The panel consensus is that while the $500B+ AI capex by GOOGL, AMZN, and META underscores Nvidia's (NVDA) near-term dominance, the long-term sustainability of this growth is uncertain due to potential shifts towards custom ASICs and the cyclical nature of semiconductor demand.
The durable data-center expansion driven by AI capex, which suggests a strong near-term outlook for Nvidia.
The shift towards custom ASICs by hyperscalers, which could significantly compress Nvidia's gross margins and growth rate.