US market wrap: When momentum turned into gravity

There was no panic, just exhaustion. Traders rotated out of high beta into anything that didn’t move — like cash. It was an unwind of valuation faith, not earnings fundamentals. ( see below)

Gravity

By the end of the New York session, the Nasdaq 100 sell-off had deepened, closing down about 2% as AI valuation worries finally broke through the surface. Palantir’s post-earnings drop dragged the broader complex lower, and the S&P followed in sympathy. There wasn’t a single smoking gun — just a confluence of bearish factors weighing on a market that had been looking stranger by the day. Hyper-momentum names were suddenly out of favour, crypto was unravelling, and the Fed’s latest bout of hawkish chatter added to the unease.

The day started as a routine pullback and ended as something more symbolic — a collective sobering-up across everything that had flown too high. Palantir, a poster child of the AI narrative, had beaten, raised, and still got sold hard. A stock that had soared more than 170% this year couldn’t find another buyer at the top. The same tone echoed through the MAG7 and the semi names, which all traded 1%–3% lower, while defensives quietly inched green. It was one of those days when momentum’s invisible hand stopped supporting the tape, and price became its own gravity.

Across markets, the mood turned risk-off. Bonds and the dollar were bid, commodities were sold, and cryptos tumbled — Bitcoin slipping below $104,000 in a clean break that erased the last month’s gains. There was no panic, just exhaustion. Traders rotated out of high beta into anything that didn’t move — like cash. It was an unwind of valuation faith, not earnings fundamentals.

Top CEOs were out in force, flagging the risks that valuations had climbed too far, too fast, just as traders were taking a harder look at names like Palantir. They didn’t sound panicked, but there was an unmistakable caution in their tone — a recognition that equity multiples had become stretched and that a 10–15% correction wouldn’t just be tolerable, it might even be healthy. Markets, they implied, had priced perfection into a world still full of policy risk and geopolitical noise. Those comments didn’t trigger the selloff, but they added a new layer of sobriety to a market already wrestling with gravity.

It’s one of the strangest earnings seasons in years — maybe ever. AI has already conquered the stock market, but traders are starting to wonder if the next battlefield might be the bond market. The numbers are spectacular, the headlines are euphoric, and yet… the market just shrugs.

Roughly two-thirds of the S&P 500 have reported so far, and the results are blistering: 64% of companies have beaten EPS forecasts by more than one standard deviation — a feat matched only by the pandemic-era stimulus sugar high. On paper, it’s a blockbuster quarter. But here’s the twist — the market doesn’t seem to care. The median stock that beats its earnings is only outperforming the index by 0.32%. That’s a third of the usual reward. In market terms, it’s like hitting a home run and getting booed rounding the bases.

That tells you everything about where we are in this cycle: we’ve priced in perfection, gift-wrapped it, and laminated it for good measure. There’s no upside left to celebrate. The tape’s reaction function has become allergic to good news — when everything’s already perfect, even “better” feels redundant. 

Indeed, multiples had already stretched into the danger zone. The S&P 500 was trading near 23 times forward earnings, above its five-year average, and the Nasdaq 100 around 28 times — a far cry from the sub-20 multiples of 2022. It wasn’t just about price-to-earnings; it was about the emotional overreach of a market that had been rewarding narrative over numbers.

By the afternoon, liquidity thinned and selling became self-reinforcing. There was no obvious catalyst — just too much air in too many trades. Tech names that had become macro assets unto themselves were being marked down to something closer to reality. Crypto followed suit, and even AI-adjacent industrials were pulled into the downdraft. The rally that had been all conviction and no caution finally hit a wall of introspection.

And that’s where I ended up, too — staring into the mirror after taking a deeper dive into Palantir’s print and asking the tricky question: how much of a bubble were we in?

The answer, when laid out clinically, isn’t simple. The market already displayed most of the classic preconditions. The buy-the-dip reflex had become doctrine after a decade of equity outperformance. The “this time is different” belief was alive and well, built around generational optimism for AI. It had been roughly 25 years since the last great mania — long enough for a new generation to believe the old rules no longer applied. Profits outside the top ten companies were stagnating, while concentration had reached record levels in both market cap and earnings. Retail money was visibly engaged, and financial conditions, despite all the talk of tightening, remained loose.

Overvaluation was apparent in spots. The leading tech cohort traded at 35x trailing earnings — high, but not yet at the 45x–70x extremes of past bubbles. The equity risk premium hadn’t collapsed to 1%, but valuation models were already stretching TAMs to implausible levels. The path to justify current prices increasingly depended on near-perfect adoption curves and permanent margin expansion — assumptions that history rarely validates.

As for the traditional warning lights, not all were flashing yet. ICT investment as a share of GDP was elevated but not excessive. Leverage across hyperscalers was manageable; most could still expand capex 40% without tapping debt. Breadth was narrow, but not yet broken. The Nasdaq hadn’t seen the serial 10% corrections typical of a late-stage bubble, and credit spreads hadn’t yet troughed the way they did before past peaks.

Still, the setup carried that unmistakable late-cycle scent — overconfidence meeting valuation gravity. It echoed the early phase of every major market turn: the narrative still strong, the math quietly fraying. The danger wasn’t collapse; it was complacency.

History’s lesson was clear. In the TMT era, the ideas were right but the prices were wrong. The same risk hung over the AI trade — the technology might transform the world, but that didn’t mean every stock deserved a divine multiple. The future could still belong to the builders, but the road there would extract its toll.

So maybe this was the market’s first healthy exhale in months — the point where belief met arithmetic. If this really was the early stage of a bubble, it was one that hadn’t yet burst — only begun to breathe.