In a nutshell
Nvidia (NVDA) posted $215.9 billion in FY2026 revenue, up 65% year over year.
The four major hyperscalers plan to spend roughly $660 billion on AI capex in 2026.
MIT research found 95% of generative AI enterprise pilots delivered no measurable profit impact.
OpenAI closed a $110 billion funding round at a $730 billion valuation in early 2026.
AI startups attracted $258.7 billion globally in 2025, representing 61% of all global VC.
The AI trade is the loudest thing in markets right now. Nvidia (NVDA) hit a $4.3 trillion market cap. Semiconductors surge 25% overnight on a single conference comment. Analysts are screaming about bubbles. Others are screaming that this time is different. Both camps are half right.
This is not a simple story. The AI stock frenzy gets some things exactly right and gets other things dangerously wrong. Separating the two is where the real edge is.
What the AI frenzy correctly identifies
The bull case on AI is not pure narrative. There is real revenue, real demand, and real infrastructure being built at a scale that has no historical precedent. The frenzy is picking up a genuine signal. It is just amplifying it past the point where valuation still connects to reality.
Nvidia's revenue is not speculation
Nvidia's Q4 2026 earnings reported over $68 billion in revenue, and for the full fiscal year the company posted $215.9 billion in FY2026 revenue, up 65% year over year. That is not vaporware. That is one of the fastest scaling revenue stories any company has ever produced.
The market is not wrong to price that growth. It is wrong about how long it can continue at that rate.
The hyperscaler backlog is contracted revenue
One key argument the bulls have right: this is not the dot-com era of pure speculation. Google Cloud reported a backlog near $460 billion in its Alphabet Q1 2026 earnings, almost double the prior year. That is contracted future revenue, not speculation. Microsoft is in a similar position. Azure grew 40% in the latest quarter while capex hit $34.9 billion.
Contracted revenue is not a promise. But it is a far sturdier foundation than eyeball counts and page views.
The infrastructure cycle is real
Big Tech's capex as a share of revenue has risen to its highest level in over a decade. That money is going into data centers, GPUs, and power contracts. Physical infrastructure does not disappear when sentiment shifts. It creates lasting competitive moats for the companies that own it.
Nvidia (NVDA), Microsoft (MSFT), and Alphabet (GOOG) are not pre-revenue startups. They are cash-generative incumbents funding expansion out of retained earnings. That distinction is critical.
Where the AI frenzy loses the plot
The frenzy is not wrong about the technology. It is wrong about timing, concentration risk, and what it costs to stay in the game.
The ROI gap between AI spending and enterprise returns
Here is the uncomfortable number that does not fit the bull narrative. MIT research found that 95% of generative AI pilots delivered no measurable profit-and-loss impact despite billions invested. Average enterprise AI spend is projected to jump roughly 65%, from about $7 million in 2025 to $11.6 million in 2026, even as most firms cannot prove a return.
Enterprises are spending more. They cannot show what they are getting for it. That is not a feature of a healthy adoption cycle. That is a warning sign.
The bulls point to future productivity gains. That may be correct. But "eventually" is not a valuation. It is a hope.
The capex numbers are becoming hard to defend
Big Tech's combined 2026 AI capex is $650 to $700 billion while most enterprises report zero returns on GenAI investments. Those two facts cannot coexist indefinitely. Either enterprise returns show up on the income statement, or the capex cycle slows. There is no third option.
Meta spent $72 billion on capex in 2025 and is expecting to double that in 2026 to between $125 billion and $145 billion. Meta's (META) AI bet may pay off. But doubling capex while ROI remains unproven is a risk that the stock price has not fully discounted.
Concentration risk hiding inside index funds
Nvidia, Microsoft, Alphabet, Amazon, Broadcom, and Meta Platforms account for almost 30% of the S&P 500, so an AI selloff would hit the index hard.
Most investors who think they are diversified are not. They are holding a leveraged AI bet wrapped in index fund packaging. That is a different kind of risk than people realise.
Circular funding inflating AI startup valuations
Nvidia invested $1 billion across 50-plus startups in 2024. Nvidia and other corporate investors fund AI startups who then purchase their products, creating circular financing loops that raise concerns about inflated valuations without genuine external demand.
When the supplier funds the customer who funds the supplier, the demand signal is not clean. Investors treating AI startup valuations as independent market signals are missing this dynamic entirely.
The stocks the frenzy has priced correctly and which ones it has not
Not every AI stock is equally caught up in the noise. The frenzy has done a better job valuing the infrastructure layer than the application layer.
The infrastructure layer has a clearer case
Nvidia (NVDA) and Broadcom (AVGO) sell the picks and shovels. Their revenue is real and their customers are committed. Nvidia's price-to-earnings ratio stands at approximately 47 times as of early 2026, high but far below dot-com extremes. For a company growing revenue 65% per year, that multiple is defensible, though it leaves no room for any earnings miss.
Microsoft (MSFT) fits here too. Microsoft's AI revenue run rate exceeded $37 billion, Copilot paid seats surpassed 20 million, and Azure AI demand continues to outpace capacity. That is genuine monetisation of the AI theme, not a marketing slide.
The application layer is a different story
Enterprise software companies trading at premium multiples on AI narratives, without demonstrable revenue tied directly to AI, are the most exposed. The Stoxcraft news article on the explosive rise of AI in global markets covered the early momentum here, but the ROI question has sharpened considerably since then.
The chip boom and the giants shaping the next tech decade is a cleaner story than the SaaS layer trying to bolt AI onto existing products.
AMD sits in a contested middle ground
Advanced Micro Devices (AMD) is worth watching here. It competes directly with Nvidia in data center GPUs but holds a fraction of the market share. Its AI narrative is real but its execution has been inconsistent. You can check how its health and performance scores compare at the AMD Stoxcard.
What separates a real AI stock from a hype play
The frenzy treats all AI exposure as equal. It is not. Here are the filters that matter:
Revenue from AI, not revenue adjacent to AI. Azure AI growth is direct. A legacy software company adding a chatbot to its product is not.
Capex funded by cash flow, not debt. The companies making the capex bets in 2026 are profitable, cash-generative incumbents, not pre-revenue startups, and they are funding the buildout partly with retained earnings, not pure debt. That distinction matters when rates rise.
Customer concentration. If a company's AI revenue depends on one or two hyperscalers, its growth story depends on their continued spending discipline.
Backlog conversion rate. Backlog is not revenue. The speed at which it converts tells you whether demand is real or just reserved.
If you want to dig deeper into how to evaluate stock fundamentals, the Stoxcraft blog post on the Stoxcraft scoring system breaks down how health, performance, and risk scores are built, which is exactly the lens you need to apply here.
How the current frenzy compares to the dot-com bubble
The comparison is made constantly. It is useful but imprecise.
The current AI investment frenzy bears resemblances to 1999's dot-com bubble but also has crucial differences. The numbers that do not make sense today include sky-high valuations of AI startups with minimal profitability, OpenAI at $730 billion with no profit expected until 2030, and 61% of all global VC funneled into AI in 2025.
But the differences are meaningful too. This is not a clean dot-com analog. Some pieces look identical, including capex intensity, concentration risk, and hardware obsolescence exposure. Some pieces are genuinely different, including incumbent profitability, real revenue at the model layer, and productivity proof points at the high end of enterprise deployment.
The investor who says this is exactly 2000 is wrong. The investor who says it is completely different is also wrong.
What the AI race means for investors outside tech
The frenzy has a second-order effect that most analysis ignores. If a cash-rich behemoth like Alphabet needs to raise significant external capital just to keep up in the AI arms race, the infrastructure costs required are staggering.
That capital has to come from somewhere. It comes from sectors being de-prioritised. Energy, utilities, and industrials building AI infrastructure are direct beneficiaries. ExxonMobil (XOM) and power producers are getting pricing power from data center electricity demand. Investors sitting out the AI frenzy entirely may be missing these adjacent plays.
The five biggest forces shaping the stock market in 2026 covers this macro context in more depth.
Where the AI stock trade goes from here
Communication services and information technology are trading at price-to-sales ratios near or above tech bubble peaks. Because each sector earns more profit from every dollar of sales today compared with the tech bubble, the price-to-earnings ratios are elevated but have not reached those tech bubble levels.
That is the honest position. Elevated. Not absurd. But with zero margin for error.
The trigger for a correction is not hard to identify. Any meaningful earnings miss from a major hyperscaler, any evidence that enterprise ROI on AI is structurally lower than expected, or any surprise in interest rates could reprice the sector quickly. The Stoxcraft news piece on why this selloff feels different addressed the fragility of momentum-driven rallies in this environment.
Cutting through the noise on AI stocks
The AI frenzy is not delusional. Revenue is real. Infrastructure is being built. The productivity case is plausible. But the frenzy is priced for perfection across a sector where 95% of enterprise pilots are not delivering measurable returns yet.
The investors who win here will not be the ones who picked the right narrative. They will be the ones who separated infrastructure from hype, checked cash flow against capex, and held a position size that matched the uncertainty. The frenzy is loud. The edge is in being quieter than it.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Always conduct your own research before making investment decisions.
Comments
Log in or sign up to join the conversation.