Santoli Says AI's "Youthful Phase" Is Over as Nasdaq Slips
CNBC's Michael Santoli says AI's "youthful phase" has ended and the story has turned to risk, with long-dated order backlogs now suspect and the Nasdaq closing lower.
CNBC's Michael Santoli argued on September 15, 2026 that the "youthful phase" of artificial intelligence is over and the market narrative has flipped toward risk, as the Nasdaq 100 tracker QQQ closed at $709.18, down 0.80%, on September 14.
The argument that has carried artificial intelligence stocks for three years — that demand is effectively unlimited and the only constraint is how fast chips can be built — is losing its grip. CNBC's Michael Santoli wrote on Monday that the "youthful phase" of AI is over, and that the narrative has flipped from opportunity to risk. His case rests on two shifts: the technology is unpopular with the general public, and projections of order backlogs stretching beyond a few quarters are now suspect.
That is a meaningful change in framing from a commentator who tracks sentiment rather than makes price calls. For most of the cycle, the bullish case did not need to be argued. It was the default. Santoli's point is that the burden of proof has moved sides.
What "youthful phase" actually describes
The phrase is about market psychology as much as technology. In the early stage of a capital-spending boom, investors reward announcements. Any order, any partnership, any multi-year commitment gets capitalized into the share price immediately, because nobody can disprove it. The revenue is far enough out that scepticism has nothing to bite on.
The mature phase is different. Investors start asking when the announced spending converts into cash, who funds it, and what happens to the return on capital if it does not. Santoli's specific target — the credibility of order backlogs beyond a few quarters — is exactly the pressure point. A backlog is a disclosure, not a contract with a guaranteed cash flow attached. It can be reordered, renegotiated or quietly stretched. When the market is in an expansive mood, a long backlog is treated as revenue already banked. When the mood turns, the same number becomes a question about counterparty quality and financing.
The second half of his argument, that AI is unpopular with the public, matters for a reason that is easy to underrate. Public opinion is upstream of policy. It shapes what regulators feel free to do, how courts read novel disputes, how local authorities treat data center permits and power contracts, and how employers talk about automation. None of that shows up in a quarterly print. All of it affects the discount rate applied to earnings a decade out.
The tape is not confirming a boom
Price action does not settle an argument, but it is worth checking against the claim. In the most recent session before Santoli's piece, on September 14, the tech-heavy Nasdaq 100 tracker QQQ closed at $709.18, down 0.80% from the prior close of $714.88, with a day range of $702.74 to $712.95. The broader S&P 500 tracker SPY finished at $760.88, off 0.45% from $764.29. The Dow tracker DIA was the most resilient of the three, closing at $524.49, down 0.25% from $525.79.
The pattern inside those numbers is the part worth noting. The most tech-concentrated benchmark fell furthest, the most industrially diversified fell least, and the broad market sat between them. That is the signature of a market reducing exposure to a theme rather than one reacting to an economy-wide shock. On the illustrative arithmetic, QQQ's decline was roughly 0.35 percentage points steeper than SPY's and about 0.55 points steeper than DIA's on the day. A single session does not make a trend, and all three trackers were closed at the time of writing. But it is not the tape of a narrative still in its youthful phase.
Why long-dated backlogs are the fragile disclosure
Investors who want to test Santoli's thesis rather than accept it have a clear place to look. Backlog and remaining-performance-obligation figures are voluntary in their presentation, wide in their definition and inconsistent between companies. Two firms can report similar headline numbers while meaning very different things by them.
The questions that separate a solid backlog from a decorative one are straightforward:
- How much of it is contracted with take-or-pay terms, and how much is a non-binding letter of intent?
- How concentrated is it among a handful of customers, and how well capitalized are those customers?
- What portion converts within twelve months rather than over five or more years?
- Is the customer funding purchases from operating cash flow, or from debt and equity raised specifically for the purpose?
Investors who want to test Santoli's thesis rather than accept it have a clear place to look.
That last point is where circularity creeps into an infrastructure boom. When suppliers, customers and financiers overlap, a backlog can look like independent demand while actually reflecting the same pool of capital moving around. Nothing about that is fraudulent. It is simply more fragile than a diversified order book, and it revalues faster when funding conditions tighten.
How investors can respond without calling a top
Santoli's commentary, published by CNBC, is not a sell signal and does not read as one. A maturing theme is not a failing one. Electricity, railways and the internet all delivered enormous economic value while destroying capital for investors who paid for the first-phase narrative at first-phase multiples.
The practical implication is a change in what counts as evidence. In the youthful phase, an announcement was enough. In this phase, the things worth watching are depreciation schedules and their assumed useful lives, free cash flow after capital spending rather than before it, the gap between reported backlog and near-term recognized revenue, customer concentration disclosures, and how each company finances the next expansion. Firms converting the theme into cash will look increasingly different from firms converting it into press releases — and that gap gets priced.
Position sizing also matters more than direction. Because the same handful of names dominates the major indexes, an investor holding a broad market fund and a Nasdaq-weighted fund and a few individual AI winners may own far more of a single theme than they think. That concentration was an advantage on the way up. It works in both directions.
What to watch next: whether the divergence between tech-heavy and industrially diversified benchmarks persists beyond a session or two, whether capital-spending guidance from the largest buyers of AI infrastructure is reaffirmed or softened, and whether companies begin narrowing their backlog definitions. Any of the three would tell you more about Santoli's thesis than the next batch of announcements will.
Frequently asked questions
What did Michael Santoli actually argue?
CNBC's Michael Santoli wrote on September 15, 2026 that the "youthful phase" of artificial intelligence is over and the market narrative has flipped to a focus on risk. He cited two factors: the technology is unpopular with the general public, and projections of order backlogs stretching beyond a few quarters can no longer be taken at face value.
Does the market data support his view?
It is consistent with it, though one session proves nothing. On September 14, 2026, the Nasdaq 100 tracker QQQ closed at $709.18, down 0.80%, while the S&P 500 tracker SPY fell 0.45% to $760.88 and the Dow tracker DIA slipped 0.25% to $524.49. The most tech-concentrated benchmark fell furthest, which fits a theme-specific de-risking rather than a broad shock.
Why are order backlogs considered unreliable?
A backlog is a disclosure, not a guaranteed cash flow. Definitions vary widely between companies, some portions are non-binding letters of intent rather than take-or-pay contracts, and conversion can be renegotiated or stretched over years. When sentiment is positive, a long backlog reads as banked revenue. When it turns, the same figure raises questions about customer quality and financing.
Is this a signal to sell AI stocks?
Santoli's commentary is a change in framing, not a price call, and it should not be read as a sell signal. A maturing investment theme is not a failing one. The practical implication is that the standard of evidence rises: cash conversion, capital-spending discipline and backlog quality start to matter more than announcements and partnership headlines.
Why does public opinion about AI matter to investors?
Public sentiment is upstream of policy. It influences how aggressively regulators act, how courts approach novel disputes, how local authorities handle data center permits and power contracts, and how employers frame automation. None of that appears in a quarterly earnings print, but all of it affects the discount rate investors apply to profits expected many years out.
What specific disclosures should investors check now?
Look at depreciation schedules and assumed useful lives for AI hardware, free cash flow after capital expenditure rather than before it, the share of reported backlog that converts within twelve months, customer concentration figures, and how expansion is financed. Companies funding purchases from operating cash flow carry different risk than those relying on newly raised debt or equity.
Sources
Photo: https://kaboompics.com/ · Pexels Licence — source
