Photo by Random Thinking on Unsplash
The Common Belief
Ask most investors what would pop an AI bubble, and they will describe a moment: a bad earnings call, a chip glut, a headline. As of August 1, 2026, the more useful question is quieter. Not when does it break, but what number has to show up for it not to. According to refresh, the debate over whether AI valuations have outrun AI economics remains unsettled — and the honest answer is that the deciding figure has been sitting in plain sight since 2024, largely unpaid.
The thesis of this piece: the AI bubble question is not a sentiment problem, it is an arithmetic problem, and the arithmetic has a specific, checkable number attached to it. Sequoia Capital warned in 2024 that AI companies would need to generate $600 billion in revenue to justify current infrastructure spending — what the firm called the "AI value gap." That is the scoreboard. Everything else is commentary.
Where It Breaks Down
Start with the spending, because it is the least disputed part. AI infrastructure spending by major tech companies reached approximately $200 billion in 2024, with Microsoft, Google, Amazon, and Meta all sharply increasing capital expenditure. Nvidia's market capitalization exceeded $3 trillion in 2024 on the back of that chip demand. Venture capital piled in behind it: over $50 billion into AI startups across 2023-2024, roughly 25-30% of all VC funding.
Now do the division that the surface reporting tends to skip. Sequoia's $600 billion revenue bar sits against roughly $200 billion of 2024 infrastructure spending. That is a required revenue-to-capex ratio of about 3 to 1 — for every dollar of data center and chip spending, the industry needs roughly three dollars of annual AI revenue to make the math work. In plain terms, that is not a rounding error to grow into. It is the entire question.
Here is the second calculation, the one that reframes the whole debate. Venture capital committed over $50 billion to AI in 2023-2024. Big Tech committed roughly $200 billion in 2024 alone. So corporate capex is running at roughly four times the pace of the venture money — which means the AI trade is far less a startup-speculation story than the dot-com comparison implies, and far more a bet on four balance sheets that can absorb losses for years. That is a meaningfully different risk shape. Startups die loudly. Capex programs get quietly trimmed.
The demand side is genuinely improving, and any fair version of this argument has to say so. Enterprise AI adoption climbed from 35% in 2023 to approximately 55% in 2024, according to McKinsey surveys. That is a 20-percentage-point jump in a single year — a roughly 57% relative increase in the share of enterprises using the technology. Adoption is not the weak link.
Chart: The money going in versus the revenue Sequoia Capital said in 2024 would be required to justify it. Bars are not to a single shared scale of meaning — capex is annual, VC is a two-year total, and the $600B is a revenue threshold, not a spend.
What the adoption number does not tell you is price. An enterprise running a pilot counts as an adopter. So does one paying nine figures a year. The gap between those two is exactly where the $600 billion has to come from, and adoption surveys are structurally incapable of measuring it. This is the point the bulls and bears talk past each other on: bulls cite the adoption curve, bears cite the revenue bar, and both are looking at real data that does not settle the argument.
Valuation is where the strain shows. Many AI-focused firms have traded at price-to-sales ratios above 20x (the stock price divided by annual revenue — at 20x, you are paying twenty dollars for each dollar of yearly sales). Revenue growth has lagged the stock moves. OpenAI's valuation reached $157 billion in late 2024 on limited current revenue. Meanwhile, several AI startups took down rounds or shut down in late 2024 as investors started asking for a path to profit rather than a demo. Both of those things happened in the same quarter, which is the most honest snapshot of this market anyone has produced.
Jim Covello, Goldman Sachs' head of stock research, put the skeptic's case in 2024 by questioning whether AI is solving problems complex enough to justify its enormous costs. Goldman Sachs research that year raised the same doubt about whether generative AI would deliver returns sufficient to warrant the infrastructure bill. That is not a fringe view from a permabear — it is the research desk of a firm underwriting much of the trade.
A Better Frame
The dot-com comparison gets reached for constantly, and it is half right in a way that misleads. The similarities are real: rapid capital deployment, speculative valuations, uncertain paths to profitability. The difference is who is holding the bag. In 1999, the spending was largely done by companies that had no other business. In 2026, the bulk of it is being done by four firms with enormous existing cash flows, which is why the H100 GPU shortage persisted through 2024 with lead times of 6 to 12 months — buyers with real money kept ordering through the doubt.
So the more likely failure mode is not a crash. It is a grind: capex guidance quietly revised down, one bad quarter of cloud growth, multiple compression across the AI complex over several quarters while the underlying technology keeps working fine. Slow disappointment is harder to trade than a crash, and much harder to notice while it is happening to your investment portfolio.
Concentration is the risk that actually reaches a normal person's account. A single company crossing $3 trillion in market capitalization means index funds are no longer the diversification instrument most people think they bought. For a 30-year-old earning $60K with $40,000 in a total-market index fund, the question is not whether to have AI exposure — it is whether they know how much they already have and would be comfortable if it halved. That is a five-minute check, not a trading decision. The same underlying question — what am I actually exposed to versus what am I told I am exposed to — runs through what Investor found when parsing Powell's inflation warning.
Three moves worth making this week:
Open your largest fund's top-ten holdings page and add up the AI-linked names. Most broad index investors are considerably more concentrated than they expect. You are not looking for a number to act on — you are looking for a number that would not surprise you if it fell by half.
Write down, in one sentence, what you would do if AI names dropped 30%. Buy more, hold, or trim. Financial planning is mostly the practice of making decisions while calm and executing them while not. A rule written today costs nothing; a decision made mid-drawdown usually costs plenty.
The single most informative number in this story is what Microsoft, Google, Amazon, and Meta say about next year's capital spending. If that guidance starts coming down, the market has answered Sequoia's question before any analyst does. Quarterly earnings calls are free and public.
On the AI-tools front, the irony is not lost: the same wave of AI investing tools now marketed for portfolio screening and personal finance is itself part of the asset class in question. Useful for pulling holdings data and flagging concentration quickly — but a screener built on AI has no special insight into whether AI is overvalued.
Bottom Line
Our read: the AI value gap is real, and it is more likely to close through years of multiple compression and trimmed capex than through a single dramatic pop. Adoption at 55% of enterprises says the technology arrived; price-to-sales ratios above 20x and a $600 billion revenue bar say the price arrived first. Both can be true, and for a long-term investor the practical response is boring — know your concentration, write your rule, and watch what the spenders guide to rather than what the stock market today happens to do.
Frequently Asked Questions
Is the AI bubble going to burst in 2026?
No one can date it, and anyone who claims to is guessing. What is checkable: Sequoia Capital's 2024 estimate that roughly $600 billion in AI revenue is needed to justify infrastructure spending of approximately $200 billion in 2024. Whether that gap closes through revenue growth or through falling valuations is the actual open question.
How much AI exposure do I already have in an index fund?
More than most people assume. With Nvidia's market capitalization exceeding $3 trillion in 2024, market-cap-weighted index funds carry meaningful concentration in AI-linked names. Check your fund's top-ten holdings page directly rather than estimating.
What does the "AI value gap" actually mean?
It is Sequoia Capital's 2024 framing for the distance between what the industry is spending on AI infrastructure and the revenue that spending would need to generate to be economically justified — a figure the firm placed at $600 billion.
Is AI investing today the same as the dot-com bubble?
Partially. The research notes similar patterns: rapid capital deployment, speculative valuations, unclear profitability timelines. The key structural difference is that most 2024 AI capex came from large, cash-generating firms rather than companies with no other revenue — which changes how a downturn would likely unfold.
Explore Our Network
Disclaimer: This article is editorial commentary for informational purposes only and does not constitute financial advice. It reflects analysis of publicly reported information, not independent product testing or personalized recommendations. Consult a qualified financial professional before making investment decisions. Research based on publicly available sources current as of August 1, 2026.
No comments:
Post a Comment