NextFin

Lopsided Tech Bets Get Hit in AI Stock Selloff

Summarized by NextFin AI
  • AI-driven rally faced its first real reckoning as the Nasdaq 100 fell 2.2% to 25,196 and the S&P 500 dropped 0.7% to 6,827, triggered by AI safety warnings, Broadcom's softer chip outlook, and the 10-year Treasury yield rising above 5%.
  • Five hyperscalers committed $660B–$690B in capex for 2026, nearly double 2025 levels, with roughly 75% ($450B) directed to AI infrastructure, raising market concerns about whether returns will justify the cost of capital.
  • Broadcom shares plunged ~15% after reiterating rather than raising 2026 guidance, dragging Nvidia down ~6% and triggering a crowded-trade unwind across semiconductors and mega-cap software names.
  • Tech sector corrected ~11% from its early-June peak, the fifth double-digit pullback of the bull market and the shallowest of the previous four, with forward P/E resetting 30% from 32x to ~22x.

NextFin News - The artificial-intelligence trade that carried U.S. equities to record highs ran into its first real reckoning on Monday, as a selloff in chipmakers and mega-cap software names dragged the Nasdaq 100 down 2.2% to 25,196 and the S&P 500 0.7% lower to 6,827. The trigger was a collision of three forces that had been building for weeks: a public call by leading AI executives to deliberately slow frontier-model development, a weaker-than-expected AI chip outlook from Broadcom that sent its shares plunging roughly 15%, and a bond market that pushed the 10-year Treasury yield back above 5% for the first time since 2007. What looked like a routine profit-taking day quickly became something more instructive: a reminder that the most crowded trade in the market is also the most exposed when the narrative shifts even slightly.

This is not 2000. The companies at the center of the AI build-out are profitable, cash-generative, and spending against visible demand. But the speed of the unwind exposes a structural tension that no amount of earnings growth has yet resolved: five hyperscalers have committed to spending between $660 billion and $690 billion on capital expenditure in 2026, nearly double 2025 levels, and the market is starting to ask whether those dollars will earn their cost of capital. The technology sector has now corrected about 11% from its early-June peak - the fifth double-digit pullback of the current bull market, and the shallowest of the previous four, according to Truist chief investment officer Keith Lerner.

The Trigger: Safety Warnings Meet a Guidance Miss

The immediate catalyst came from inside the industry itself. On September 12, Dario Amodei, chief executive of Anthropic, published a 3,800-word essay calling for a deliberate global slowdown in the pace of AI capability development. "We must slow the pace at which we improve the capabilities of AI models," Amodei wrote. "Progress will still seem fast, and we must make wise use of the time we gain." The essay followed the resignation of an Anthropic researcher, Jacob Coxon, who said the people building AI "earnestly believe that it could kill us all by the end of the decade." Hours later, OpenAI chief executive Sam Altman told Fortune that taking the company public in 2026 would be "ill-advised," pushing the most anticipated initial public offering in market history to 2027 or later.

"We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain." - Dario Amodei, Anthropic CEO

For investors who had priced the AI trade as a straight line higher, the symbolism was jarring: the very executives whose companies have been the beneficiaries of the rally were publicly questioning its tempo. The message did not need to be a call to halt development to move markets. It only needed to introduce doubt about the one assumption the trade cannot survive without - that the pace of progress, and the spending that follows it, is self-justifying.

Then came the numbers. Broadcom, one of the purest listed proxies for AI infrastructure demand, reiterated rather than raised its 2026 guidance and offered a softer chip outlook. Shares fell roughly 15%, and the selling rippled through the semiconductor complex. Nvidia dropped about 6%, while Microsoft, Salesforce, and Snowflake also declined. The pattern was textbook crowded-trade unwinding: hedge funds that had entered the correction with hedges in place still saw losses mount as volatility accelerated and brokers forced position reductions.

The bond market added its own pressure. With Brent crude climbing toward $108 a barrel, the 10-year Treasury yield forged above 5%, reaching as high as 5.033% in early trading - the highest level since 2007. Growth stocks, whose valuations are derived from earnings far in the future, are mathematically sensitive to the discount rate. When the risk-free rate resets to a level not seen in nearly two decades, the denominator of every AI valuation model gets heavier at once.

Why It Broke Now: Concentration, Capex, and the ROI Question

The deeper story is not about one day's tape. It is about how concentrated the AI trade had become, and how much of the market's forward return now depends on a single thesis holding true. The five largest cloud and AI infrastructure providers - Microsoft, Alphabet, Amazon, Meta, and Oracle - have collectively committed to spending between $660 billion and $690 billion on capital expenditure in 2026. At the top end, that is more than four times what the entire publicly traded U.S. energy sector spends drilling wells, refining oil, and delivering gasoline. Roughly 75% of that tab, about $450 billion, goes directly to AI infrastructure: GPUs, servers, networking equipment, and data centers.

The bull case for that spending is not theoretical. Cloud growth is showing up in reported revenue: AWS grew 24%, Google Cloud grew 48%, and Azure grew 39% in recent quarters. Microsoft disclosed an $80 billion backlog of Azure orders it cannot fulfill because of power constraints - demand outpacing even an aggressive build-out. Unlike the dot-com era, today's hyperscalers are profitable and generating massive cash flows. Their customers are paying.

But the counter-argument is equally concrete, and it is the one the market is now pricing. Capital expenditure is a claim on future cash flow that arrives years before the revenue it is meant to generate. When five companies commit to doubling their spending in a single year, the market is no longer buying a growth story - it is underwriting a timing bet. The question shifts from "Will AI be valuable?" to "Will the returns arrive before investors lose patience?" That is a harder question to answer, and it does not have a clean consensus.

Valuation had also done much of the work ahead of the news. The technology sector's forward price-to-earnings ratio has declined to about 22 times, down from an October peak of 32 times - a reset of roughly 30%, in Lerner's reading. The sector's valuation premium versus the S&P 500 has narrowed to 13%, down materially from the nearly 50% peak seen earlier in the bull market. By that measure, much of the air has already left the balloon. Monday's move was not the start of a valuation collapse; it was a repricing of the marginal buyer who had assumed the AI trade could only go up.

The Positioning Layer: When Everyone Is Long the Same Thing

There is a mechanical reason the selloff felt disproportionate to the news. The AI trade had become the most crowded position in the market, and crowded trades do not unwind gradually - they gap. Market participants reported that several funds were forced to reduce positions as volatility accelerated, with crowded trades unwinding rapidly across technology sectors. Even funds that entered the week with hedges in place saw losses mount as selling intensified and liquidity thinned.

This is the hidden leverage in a momentum-driven market. When a large share of active capital is positioned the same way, the marginal seller does not need a fundamental reason to move price - they only need another seller. The Broadcom guidance miss did not change the AI thesis; it changed the conviction of the investor who was longest and most levered, and that was enough.

The concentration risk extends beyond hedge funds. The top 10 stocks in the S&P 500 - the primary beneficiaries of the AI build-out - account for roughly 40% of the index's market capitalization. A selloff in those names does not stay in the technology sector. It transmits to index funds, target-date retirement accounts, and volatility-control strategies that rebalance mechanically. That transmission channel is why a 15% drop in one chip stock can pull 2.2% out of the Nasdaq 100 and 0.7% out of the S&P 500.

Cyclical or Structural: Separating the Two Trades

Getting the diagnosis right matters, because the two possible answers point in opposite directions. Is this a cyclical pullback within a continuing bull market, or a structural break in the AI thesis?

The evidence favors the cyclical read for the price action itself. This is the fifth double-digit percentage pullback of the current bull market, and by Lerner's measure it has been the shallowest of the previous four. The sector's forward P/E has already reset 30% from its peak, and the premium to the broader market has compressed from nearly 50% to 13%. The drivers of the selloff - positioning, a single guidance miss, and a rate spike - are all mean-reverting by nature. Positioning resets when levered hands are washed out. Guidance gets revised. Yields fluctuate with oil prices and Fed expectations.

But layered underneath the cyclical move is a structural question that will not mean-revert on its own: can $660 billion to $690 billion of annual AI infrastructure spending earn an adequate return? That question is not answered by a bounce in Nvidia shares. It is answered over quarters, as cloud revenue growth is weighed against depreciation schedules, and as hyperscalers either justify their capex or are forced to restrain it. The cyclical trade is about where stocks find support this month. The structural trade is about whether the AI build-out is an investment boom or a capital misallocation.

These two layers can point in opposite directions at the same time. The stocks can recover from oversold positioning while the ROI question remains unresolved - and that divergence is exactly what makes this market difficult. A tactical bounce does not validate the thesis, and a structural doubt does not preclude a rally.

The Second-Order Read: The Bond Market Is Now Part of the AI Trade

The first-order reading of Monday is straightforward: AI safety warnings hurt chip stocks. The second-order implication is more consequential, and the market has not fully priced it. The AI capital cycle now runs through the bond market, and the bond market is sending a warning.

When the 10-year Treasury yield sits above 5%, every dollar of future AI revenue is worth less today. That is simple discounting. But the deeper channel is the cost of capital for the build-out itself. Hyperscalers fund data centers with a mix of cash flow and debt. A sustained move of 100 basis points in long-term rates changes the economics of a multi-year infrastructure program by tens of billions of dollars. At some level, higher rates do not just compress multiples - they change which projects clear the hurdle rate.

This is the second-order risk that the consensus has underweighted. The market has focused on whether AI demand is real. The more binding question is whether the cost of supplying that demand stays low enough to make the investment worthwhile. If the bond market keeps the 10-year yield above 5% while oil remains elevated, the AI trade faces a squeeze from both sides: slower discounted revenue and more expensive capital.

The Counter-Thesis, and What Would Break It

The strongest argument against the bearish read is also the simplest: the spending is demand-driven, not speculative. Hyperscalers are not building data centers on a hope - they have order backlogs they cannot fill. Cloud growth rates in the 20% to 50% range are real, reported, and accelerating. The CEOs calling for caution are managing risk, not signaling an end to the cycle; slowing the pace of model development does not reduce the need for inference capacity, and may extend the infrastructure runway. On this view, Monday was a healthy shakeout of excess leverage, not the start of a thesis break.

That counter-thesis is credible, and it is why a blanket "AI bubble" call would be as lazy as the bullish one. But it rests on one fragile link: the timing between spending and returns. If cloud revenue growth holds above 20% while capex guidance stabilizes, the bull case survives intact. If, however, hyperscalers keep raising capex while cloud growth decelerates below 20%, the ROI question moves from theoretical to urgent - and the multiple will not hold.

The falsifying signal is specific and observable. Watch two metrics over the next two earnings seasons: first, whether Broadcom and Nvidia raise guidance above current consensus after this reset; second, whether the 10-year Treasury yield falls back below 4.5%. If both print, the selloff was a cyclical positioning wash and the AI trade remains intact. If instead capex guidance keeps climbing while the 10-year yield holds above 5%, the structural ROI doubt is confirmed, and this is not a dip to buy - it is the first chapter of a longer repricing.

What Comes Next

In the short term, the market's direction depends on positioning and liquidity rather than fundamentals. The oversold bounce is a real possibility once forced selling exhausts itself, and the shallowest-of-five-pullbacks pattern argues for a contained drawdown rather than a bear market. Traders will watch whether the Nasdaq 100 holds above its 50-day moving average and whether semiconductor names stabilize after Broadcom's guidance reset.

Over the medium term, the focus shifts to earnings. The next wave of hyperscaler reports will show whether cloud growth is keeping pace with the capital being deployed. This is where the thesis is actually decided - not in CEO essays, and not in one day's tape.

Over the long term, the structural question dominates: does the AI build-out generate returns above its cost of capital? If yes, the current volatility will look like a footnote in a decade-long investment cycle. If no, the capex wave will crest, guidance will roll over, and the market will relearn a lesson it has learned before - that spending is not the same thing as value creation.

Monday's selloff did not answer that question. It only made it impossible to ignore. The AI trade was never just about the technology. It was about the price paid for it - and for the first time in this cycle, the market is negotiating.

Explore more exclusive insights at nextfin.ai.

Search, ask or analyze anything...