Everything About AI
NextFin News - The S&P 500 Index sits within about 3% of its record high, yet 60% of its 500 components are languishing in bear market territory, down at least 20% from their own peaks. The divergence, flagged by Goldman Sachs macro trader Bobby Molavi, describes a market carried by a narrowing band of winners while the majority quietly deteriorates. The "strange thing," Molavi said, is that shares in both the technology megacaps whose capital spending fuels the AI rally and the companies receiving that spending are falling together - a signal that the AI capital-expenditure cycle may be rolling over rather than broadening.
"The strange thing is that shares in both the tech megacaps whose capex is at the heart of the AI-powered equities rally and the receivers of that spending are going down together," said Bobby Molavi, a macro trader at Goldman Sachs Group Inc. The remark cuts to the heart of what makes this breadth deterioration different from an ordinary rotation: in a healthy capital-expenditure-led rally, the spenders and the recipients of that spending should not decline in unison.
The Index Is Fine. The Market Underneath Is Not.
The index level tells one story; the breadth underneath tells another. As of the October 30 close, the S&P 500 stood at 7,560.68, roughly 3.05% below the record close of 7,818.93 set on October 6, 2026, and trading below its 50-day moving average of 7,696.69. A three-percent drawdown is noise in most markets. What is not noise is that three of every five constituents are down 20% or more from their record highs. Breadth of this kind does not usually accompany a healthy advance near all-time highs; it accompanies distribution.
The concentration is visible in the mega-cap technology group that has led this cycle. The Magnificent Seven has entered correction territory, with Amazon.com and Microsoft already in bear market territory, down more than 20% from recent highs, as investors punish heavy AI spending that has not yet produced commensurate cloud-computing growth. Alphabet has given back about 8% over the past month, Meta Platforms has surrendered the gains it recorded after emphasizing AI-driven earnings growth, Nvidia has traded range-bound for months, Apple has lost the advance it built on strong iPhone sales, and Tesla is down about 5% for the year.
The stakes are not merely technical. The AI rally of 2025 and 2026 was never just a valuation story; it was a capital-cycle story. Hyperscaler capital expenditure has swollen from a few hundred billion dollars into a cycle heading toward the trillion-dollar range. Consensus capital-spending estimates for the largest cloud-infrastructure companies reached $670 billion for 2026, equivalent to more than 90% of their expected cash flows this year, according to Goldman Sachs Research. AI investment is expected to drive roughly 40% of S&P 500 earnings growth this year. A Goldman Sachs basket of stocks tied to AI data-center construction returned nearly 60% this year, and the index rallied about 13% from late March, its sharpest advance since April 2020. When the companies placing the orders and the companies filling them decline together, the transmission channel of the entire rally is under stress.
The Capex Cycle Is the Market, and It Is Showing Its First Crack
The mechanism is straightforward and unforgiving. In a capital-expenditure-led bull market, the companies placing the orders and the companies filling the orders should not fall in lockstep unless the order book itself is peaking. The semiconductor makers and their equipment suppliers, the cloud builders, the power and cooling infrastructure names, and the software layer that monetizes the compute should rotate - they should not collapse together. Molavi's observation that both sides of the transaction are declining is the first hard evidence that the AI capital cycle may be moving from the acceleration phase, in which every incremental dollar of spending is rewarded, to the distribution phase, in which investors begin demanding proof of return on capital.
History supplies the template. Molavi's own historical sequence is consistent: stocks peak first, capital expenditure keeps rising for a while, then capex eventually rolls over and equities follow. The uncomfortable parallel is that concentration in today's index is already sitting in the same neighbourhood as the railroad mania of the 1840s, the Nifty Fifty of the early 1970s, and the dot-com era of 2000. In each episode, the index looked resilient at the top because the leaders kept making new highs even as the average stock broke down. In the dot-com cycle, the Nasdaq's final highs were built on a handful of networking and telecom-equipment leaders while the median stock had already peaked; in the Nifty Fifty era, a narrow group of "one-decision" growth stocks held the averages as market breadth narrowed; in the railroad mania, the capital spending on track and rolling stock kept climbing after share prices had turned. The breadth divergence we see now - 60% of the index in bear market territory with the index 3% from its record - is the same fingerprint.
This is a cyclical signal, not a structural verdict on AI. Breadth deteriorations of this magnitude are mean-reverting by construction: either the laggards recover, or the index falls to meet them. What is not mean-reverting is the capital stock already committed. Data centers under construction do not get un-built, and power contracts signed for AI loads do not simply vanish. That distinction matters: the cycle can revert through price without the underlying investment cycle reverting through quantity, which is why the distribution phase can last longer and hurt more than a simple valuation reset.
Why This Time Could Be Different: The Capex Tanker
The counter-thesis deserves its full weight, because it is not weak. Capital expenditure on this scale behaves, in Molavi's framing, like a massive oil tanker: once it starts moving, it takes a long time to stop or change course. Global research houses project worldwide AI-related capital spending could reach roughly $5.5 trillion by the end of the decade, and ratings analysts expect hyperscaler capex alone to approach $1.4 trillion by 2029. That is not a cycle that reverses on one soft quarter of cloud growth. The spending is underpinned by a genuine demand shock: AI training and inference require chips, data centers, energy, and networking that already exist and are in short supply. Unlike the dot-com buildout, where much of the fiber was laid for traffic that never arrived, today's AI demand is visible in enterprise adoption, sovereign AI programs, and the unit economics of replacing human labor with software.
Goldman Sachs Research still forecasts US stocks rising 6% in 2026, powered by earnings growth rather than multiple expansion, with a year-end target of 7,600 set in April. The bull case is that breadth deterioration is a normal mid-cycle consolidation, that the laggards are rotating rather than breaking, and that once the current guidance cycle clears, leadership will broaden again. Molavi has argued publicly that pessimists sound smart but optimists make money, and that the pillars supporting the bull run - above all the AI capex boom - remain clear. On this read, a 60% bearish breadth reading is not a top signal but a buying opportunity in the 300 stocks that have already done the falling.
There is also a valuation discipline that the bears must respect. A capex cycle does not end because spending is large; it ends because the marginal return on that spending turns negative. So far, the hyperscalers are funding the bulk of their investment from operating cash flow rather than from debt or equity issuance, which is materially different from the dot-com era, when capital was raised in markets that were happy to fund anything with a ".com" suffix. As long as free cash flow covers capex, the cycle has an internal shock absorber that 2000 did not.
The Second-Order Question: What Happens When the Recipients Stop Rising
The market has not fully priced the second-order consequence, and this is where the bearish reading earns its keep. The first-order effect of slowing AI spending is obvious: chip and data-center stocks fall. The second-order effect is what happens to the megacaps themselves. For years, investors tolerated massive capital spending from Amazon, Microsoft, Alphabet, and Meta because every dollar spent was assumed to buy future cloud and AI revenue growth. If the recipients of that spending - the semiconductor and infrastructure chain - are falling because order growth is peaking, then the megacaps' own spending is about to face the same scrutiny from the other direction: their capex guidance becomes the thing investors question, not the thing they celebrate.
That is the asymmetry. In the acceleration phase, more capex is bullish for everyone in the chain. In the distribution phase, more capex becomes evidence of overbuilding, and less capex becomes evidence of weakening demand. Either way, the multiple compresses. This is why breadth matters more than the index level: a narrow index can be held up by momentum and passive flows for months, but it cannot be held up once the leaders' earnings revisions turn negative. The 2026 earnings season will therefore be less about whether the megacaps beat on the top line and more about whether their capital budgets for 2027 accelerate, hold, or roll over.
The expectation gap is the real story. The consensus baseline - a 6% index gain for 2026, capex of $670 billion this year, and a multi-year buildout measured in the trillions - already assumes the acceleration phase continues. If the next round of guidance merely holds flat rather than accelerates, that is a negative surprise against a consensus that is still modeling growth. This is the classic late-cycle trap: good news that is less good than expected gets treated as bad news, because the multiple was priced for acceleration, not maintenance.
What Would Prove the Bearish Read Wrong
The falsifying signal is specific, and it should be watched rather than assumed. First, if hyperscaler capital-expenditure guidance for 2027 continues to rise - not merely hold flat - and cloud revenue growth re-accelerates in the same period, then the "capex peak" thesis is wrong and the breadth divergence is just a rotation. Second, if the share of S&P 500 constituents trading above their 200-day moving average climbs back above 50% while the index makes a new record high, the divergence resolves bullishly and the market broadens rather than breaks. Until one of those two conditions is met, the burden of proof sits with the bulls.
Who Benefits, Who Is Exposed, and the Road Ahead
The near-term implication is that volatility will remain elevated and leadership will stay narrow. The beneficiaries of a slowing capex cycle are the traditional defensive sectors - health care, consumer staples, and utilities - and the quality names with free cash flow that do not depend on AI spending to justify their valuations. The most exposed are the semiconductor-equipment suppliers, the data-center real-estate and power names, and the megacaps whose capital budgets are largest relative to their cash flow.
Split by time horizon, the picture is not uniform. In the short term, sentiment and positioning dominate, and a three-percent index drawdown can reverse on a single benign inflation print or a dovish signal from the Federal Reserve. Over the medium term, the next two quarters of hyperscaler capex guidance and cloud revenue growth will decide whether this is a consolidation or a cycle top. Over the long term, the structural question is whether AI delivers the productivity gains that justify a trillion-dollar capital cycle - and on that question, the evidence is still being written.
Three scenarios frame the path. The base case is a prolonged period of narrow leadership and elevated volatility, with the index range-bound while breadth either gradually repairs or deteriorates further. The upside case is that capex guidance holds, cloud growth re-accelerates, and leadership broadens to the 60% of the index now in bear market territory - which would be a powerful rally, because those stocks have the most ground to recover. The downside case is that capex guidance rolls over in early 2027, the megacaps trim spending, and the breadth deterioration spreads into the index level, taking the S&P 500 into a proper double-digit correction.
The data to watch is unambiguous: 2027 capex guidance from the hyperscalers, cloud revenue growth rates, the share of the index above its 200-day average, and the earnings-revision trend for the semiconductor-equipment group. If guidance accelerates and breadth improves, stand down. If guidance flattens while revisions roll over, the 60% figure is the beginning of the story, not the end.
This is not a market telling you the AI story is dead. It is a market telling you the easy money from the capex acceleration phase is over, and that the next phase will reward proof of earnings, not promises of spending.
Explore more exclusive insights at nextfin.ai.
