By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel is divided on Nvidia's future, with concerns about custom silicon cannibalization, power constraints, and capex timelines offsetting optimism about AI infrastructure spend and efficiency gains.
Risk: Power constraints and capex shifts away from chips
Opportunity: Secular AI infrastructure spend
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 artificial intelligence (AI) hyperscalers sent ripples through the investing world when the big four projected $650 billion in capital expenditures for 2026, with nearly all of that total going toward building data centers. Despite 2026 not being halfway over, we're already getting estimates for 2027 spending, which could be big news for some investors.
Nvidia (NASDAQ: NVDA) broke the news that projections are tending toward $1 trillion in data center capital expenditures next year, and other commentary from hyperscalers like Alphabet confirms it. Despite giving a range of $180 billion to $190 billion in capital expenditures this year, Alphabet told investors that next year's totals will be "significantly higher." That's huge news for investors, especially those who are heavily invested in artificial intelligence (AI) infrastructure businesses like I am.
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 »
I think the biggest beneficiary of all this spending is right under investors' noses, and the company is set to capitalize on it next year.
While there are many ways to invest in the data center build-out, Nvidia is still the best way, in my opinion. Its dominance in the industry can't be overstated, and while custom AI chips are emerging, Nvidia's GPUs still handle the lion's share of AI workloads.
Furthermore, as the construction of data centers wraps up, it's time to fill them with computing units. This will change the mix of spending to be more computing-focused, boosting Nvidia's share of the pool of money being spent on data centers.
Another catalyst Nvidia has in 2027 is the launch of its new GPU architecture, Rubin. Rubin GPUs will be game-changing, as they can train at a fourth of the cost of previous-generation Blackwell chips and run inference at a tenth of the cost. That will drive some companies to upgrade older generations to newer ones, and many of the new data centers going up are likely to incorporate Rubin technology.
These are major catalysts that will drive monster growth for Nvidia next year, but none of that is priced into the stock.
At 23 times forward earnings, Nvidia is barely more expensive than the broader market, as measured by the S&P 500 (SNPINDEX: ^GSPC). The S&P 500 trades for about 22 times forward earnings, and considering Nvidia's monstrous growth, that's a premium worth paying. Furthermore, this indicates that only one year of Nvidia's growth is priced into the stock, so 2027's results should directly translate to stock price appreciation.
That's huge news for investors and makes Nvidia a screaming buy 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 $442,220! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,230,114!
Now, it’s worth noting Stock Advisor’s total average return is 926% — a market-crushing outperformance compared to 203% 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 12, 2026. *
Keithen Drury has positions in Alphabet and Nvidia. The Motley Fool has positions in and recommends Alphabet 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's valuation already prices in most of the 2027 upside the article claims is unpriced."
The article's $1T 2027 data-center capex figure originates from Nvidia's own commentary and Alphabet's vague 'significantly higher' guidance, creating circular optimism. While Rubin could lower training/inference costs, hyperscalers are already designing custom ASICs that bypass Nvidia GPUs entirely; any shift in spending mix toward compute may therefore accrue less to NVDA than projected. At 23x forward earnings the stock embeds aggressive 2027 growth assumptions that leave little margin if ROI on AI workloads disappoints or capex timelines slip.
Even if custom chips gain share, Nvidia's software moat and installed base could still deliver above-consensus revenue, making the current multiple look conservative rather than stretched.
"Nvidia's upside rests on a durable, multi-year data-center capex cycle, but that upside is contingent on uninterrupted AI demand and Rubin delivering real cost advantage; any slowdown or over-optimism could re-rate the stock."
Article builds a bullish case for Nvidia from a $1T data-center spend forecast for 2027 and Rubin's proposed efficiency gains. The core thesis is Nvidia stands to capture most of a secular AI infrastructure spend. However, the buoyancy is likely already in the stock: at 23x forward earnings, the rate of growth must be exceptional for further multiple expansion, and a lot can go wrong. Risks: hyperscaler capex could peak earlier or slow, Rubin may underperform, competition from AMD/Intel or new accelerators could erode pricing power, and macro cycles or regulation could curb AI budgets. Also, data-center demand may shift toward software/infra multiples rather than pure hardware sales.
The strongest counter is that the data-center spend trajectory could prove front-loaded or cyclical, so the 'trillion-dollar' claim might not yield sustained earnings. If AI demand moderation occurs or Rubin misses its cost savings, the stock could reprice.
"Nvidia's valuation is currently priced for perpetual growth, ignoring the imminent risk of margin compression from custom silicon adoption and potential capex cooling."
The article’s reliance on a 23x forward P/E for Nvidia is misleading because it ignores the inevitable margin compression as hyperscalers like Alphabet and Microsoft pivot toward custom silicon (ASICs). While the $1 trillion data center spend is a massive tailwind, the 'Rubin' architecture launch in 2027 carries significant execution risk. If the transition to lower-cost inference chips cannibalizes higher-margin training revenue faster than anticipated, Nvidia’s EPS growth could decelerate sharply. Investors are currently pricing in perfection, but they are underestimating the cyclical nature of hardware refreshes and the potential for a 'capex hangover' if AI ROI fails to materialize for the hyperscalers by late 2026.
If Nvidia maintains its software moat via CUDA and continues to iterate faster than custom chip competitors, the 'Rubin' architecture could actually expand margins by cementing their status as the only viable choice for next-gen AI workloads.
"Nvidia benefits from the $1T capex wave, but the stock's valuation assumes zero margin compression from custom silicon competition and 100% Rubin attach—both heroic assumptions."
The $1T capex projection for 2027 is real and material, but the article conflates two separate narratives: hyperscaler buildout (confirmed) and Nvidia's valuation (speculative). At 23x forward P/E, Nvidia trades at parity with the S&P 500 despite 40%+ YoY revenue growth—that's genuinely cheap on a PEG basis. However, the article assumes Rubin adoption happens immediately and that custom silicon (AMD MI, Google TPU, Amazon Trainium) won't cannibalize share. The bigger miss: $1T in capex doesn't equal $1T in GPU revenue. Infrastructure, networking, power, real estate, and software consume 60-70% of that pool. Nvidia's actual TAM expansion is real but narrower than implied.
If hyperscalers achieve 70%+ utilization on existing capacity by Q2 2027, capex growth could decelerate sharply, and Rubin's cost advantages may trigger a price war rather than a margin expansion cycle. Custom silicon adoption is already accelerating—Google's TPU now handles 50%+ of internal AI workloads.
"Claude's 50% TPU figure lacks support and exaggerates immediate share loss risk to Nvidia."
Claude's claim that Google's TPU already handles 50%+ of internal AI workloads appears invented and unverifiable from the provided discussion or article. This inflates the near-term custom-silicon threat without evidence, while the real constraint may instead be power availability and grid connections that could delay the entire $1T buildout regardless of chip choice.
"Rising custom-silicon adoption (TPUs/ASICs) is a real risk to Nvidia's pricing power and should factor into capex timing, regardless of the exact TPU share."
Grok, labeling Claude's TPU-share claim as invented ignores public signals of rising custom-silicon usage. The exact share may be uncertain, but the trend matters: if TPUs/ASICs capture more AI workloads, Nvidia faces slower GPU-driven revenue growth and potential pricing pressure, which could worsen the risk to the 23x multiple more than Rubin's efficiency gains alone. Treat the '50%' figure as a data point, not a verdict.
"Grid and energy availability, not custom silicon competition, will act as the primary ceiling for Nvidia's revenue growth regardless of the Rubin architecture's efficiency."
Claude and Gemini are fixated on hardware cannibalization, but they are missing the real bottleneck: the 'power-to-compute' ratio. Even if Rubin drives efficiency, hyperscaler capex is increasingly tied to energy procurement, not just silicon. If grid constraints persist, the $1T spend won't flow to Nvidia’s top line; it will be diverted to utility infrastructure and cooling. We aren't looking at a chip war; we are looking at a localized energy scarcity crisis that caps Nvidia's total addressable market.
"Power scarcity is a real capex drag, but it's addressable through infrastructure spending—the actual threat is capex diversion away from silicon, not toward cheaper chips."
Gemini's power constraint angle is underexplored but risks becoming a catch-all excuse. Grid bottlenecks are real—data centers consume ~4% of US electricity already—but hyperscalers are actively building private power (Microsoft-Constellation deal, Google geothermal). The constraint is regional and solvable with capex, not a hard ceiling on Nvidia's TAM. More pressing: if power becomes the binding constraint, capex shifts away from chips entirely, which actually *hurts* Nvidia more than custom silicon does. That's the real second-order risk.
The panel is divided on Nvidia's future, with concerns about custom silicon cannibalization, power constraints, and capex timelines offsetting optimism about AI infrastructure spend and efficiency gains.
Secular AI infrastructure spend
Power constraints and capex shifts away from chips