NASDAQ at New Highs, Yields Near 5%: The Most Crowded Traps and the Fine Line of Being Contrarian

The thin line between being contrarian and just being stupid anchors the whole conversation. Jason Shapiro — founder of Crowded Market Report, featured in Jack Schwager's Unknown Market Wizards, and a trader who has read the weekly Commitment of Traders (COT — the CFTC's who-owns-what census for futures) for 26 years — tells David Lin how going against the crowd is misread. On Kalshi, the prediction market, the NASDAQ 100 finishing the year above 31,800 is priced at 59% while the index hovers near 30,600; the mood is modestly bullish, not euphoric. For Shapiro the question is not price itself but how crowded positioning is: when everyone faces the same way, the payoff symmetry breaks.

Why the bond selloff is global, not just American

Headlines chased Washington until the US 10-year yield nudged toward 5%, but the picture Shapiro paints is wider. The selloff reported around 11-15 September by Bloomberg, Kitco/Reuters and Morningstar hit not only US Treasuries but Japanese, German and British government bonds together, with China the lone exception. Two facts matter at once: the move cannot be explained by what the US Treasury secretary wants, and buying only US bonds will not stop Japanese or German yields from rising. It is like trying to stop a building's water leak by closing a single tap — the system is connected.

That creates a tight loop. 1) Foreign yields rise as bonds sell off. 2) If Washington tries to keep domestic rates artificially low, the dollar weakens. 3) A weak dollar lifts imported inflation. 4) Inflation circles back to a rates problem. Shapiro nods to the textbook cure — tighten your belt, spend less, tax more — and then dismisses it: no one wins an election promising less money and higher taxes, so belt-tightening has been forgotten, especially since 2009. Policymakers feel drunk on the old habit, the market tries to stay sober, and that explains the drift toward funding long-term debt with short-term bills.

The most controversial bridge for that short-term funding is demand from stablecoins. Shapiro says upfront he is no expert, but the thesis he distills from what he has read is clear: a bill expected in Congress could create fresh demand for short-dated Treasury bills (T-bills) via stablecoin issuance, opening room to fund and even buy back some long-end debt with short-term money. The bill has not passed, success is uncertain, yet that novelty is exactly what he likes. No economic history book lists such a tool, so older macro economists — himself included — have no template for it. When the traditional toolbox is empty, a new tool is being tried.

No buying until capitulation

What would make Shapiro a buyer of bonds? Capitulation. Everyone gives up, says the US is broke and will never make payments — yields spike higher, and only there would he get interested. What he sees today is the opposite: many are trying to be contrarian by buying bonds. He triangulates the crowd via three channels. 1) A YouTuber he has long faded — someone he has watched be wrong often — started a month ago calling bonds a great contrarian buy. 2) Friends on his Discord report the same chatter spreading locally. 3) A large hedge-fund fixed-income manager told him in an hour-long call, unprompted, that almost every bond trader he speaks to is trying the same contrarian long and getting run over. Price action confirmed it that week — the crowd was punished.

That is where rules and human judgment diverge. A COT and rules-based system might have flashed long bonds three to four weeks ago; Shapiro avoided it because of that crowd intelligence, and taking a position without measuring crowdedness is like navigating fog without a map — the direction can look right and the timing can still be disastrous. The recent story for stocks is resilience: crude rose from the mid-70s in early August to above 100 dollars, 10-year yields pushed higher, daily political noise from Donald Trump kept coming, and on top of that the headline that AI will end humanity circulated — yet the S&P and NASDAQ walked to new highs; even if oil eases from the highs it remains far higher than before, and Monday's gap down was never retested.

COT in energy and the seasonality myth

David Lin pulls up the COT sheet for energy futures and options combined: open interest near 30,000 contracts, longs about 13,000, producer-merchants near 25,000, swap dealers long 1,800 short 400, managed money near zero, other reportables 10,000 long 400 short. At first glance it looks long-heavy, but Shapiro reads it differently. The surprise is not crowding but quiet: in crude, Brent, WTI, heating oil and unleaded gasoline, speculators chased far less than he would expect given the rally. That suggests longs could still grow, so the run may not be done. He adds an honest caveat: over 26 years, energy has been the worst sector for his COT-based trading — so he treats that signal with extra caution.

Seasonality gets a similar skeptical read. September is supposed to be the worst month, October the selling month — the memes were everywhere on X last year — yet they did not pay. Shapiro does not believe in seasonality because the market is a discounting mechanism: once everyone sees the calendar, the calendar stops working. He links that straight to Bitcoin and its most cherished calendar: the four-year cycle. Declaring a law of nature from just two examples is nonsense, and that belief has kept bulls who tried to time the October-November low on the sidelines, now forced to chase.

Bitcoin's action fits that chase. A two-day jump on 19 August after Treasury's bond-market intervention chatter, two flat weeks, then another 13-14% surge in the last 48 hours that looks random. Media yells new bull cycle at each spike and goes silent in the chop; in 2021 the narrative was bullish every day regardless of price, now the narrative follows price. Shapiro's read is that as the expected four-year low approaches in October-November, those who missed it must buy in, some even shorting to make up for the missed long — classic bad behavior — and that squeeze lifts price. The real test arrives in December when the calendar excuse expires: if you believe, you must be in, no excuse left. Then what?

Does Shapiro trade Bitcoin? No. He still holds the coins he bought when the ETF traded at a 40% discount after the FTX (SB) blow-up, but he does not actively trade it. Reason is methodological: COT history for Bitcoin is not long enough to form a decent conclusion, and COT is central to his process. He follows it, discusses it, but does not force a trade where his edge does not apply. That is part of contrarian discipline: you do not have to play every moving thing; waiting for the easy meal is also a strategy.

Positioning in indices aligns with sentiment: modestly bullish. NASDAQ was flat on the day after a strong prior session, S&P near the highs. Kalshi prices an 87-89% chance that NASDAQ 100 stays above 31,000 into year-end — not a bet on 40,000, just on holding current levels or a bit higher. Shapiro says COT shows the same: positioning is bullish but not out-of-control bullish. Sentiment and positioning, both modestly long, confirm each other rather than diverge.

Not a bubble, but a hate indicator

Is NASDAQ and AI a bubble? Shapiro answers flatly no. He recalls a psychology book that lingered on the South Sea bubble: a real bubble is when everyone loves the same story for every reason on every corner. What he sees today is the opposite. His teacher wife and his twentysomething hippie daughter hate AI, town-hall meetings chant we do not want a data center here, headlines warn AI will end humanity. That is not bubble talk; bubble talk is love, not hate. Meanwhile at home the kitchen and bathroom remodelling was planned with AI, saving time and money efficiently — the hated tool quietly does the work. As with the internet — which did change the world yet still saw stocks fall 80-90% before the follow-through — a market can fall without being a bubble.

Valuation talk is therefore slippery. Ed Yardeni's note Tech is top dog again argues forward earnings (analysts' next-year profit forecasts) are rising, so multiples look cheaper than last year. Shapiro is distant: forward estimates are almost never accurate; five years ago no one forecast Nvidia's and Micron's earnings of today correctly. If no one, not even AI company heads, knows what the AI future looks like, cheap versus expensive is an exercise in futility. Rumors swirling in Silicon Valley of Anthropic at a 2 trillion valuation and 21-31 times sales keep recruits hesitating, fearing option packages could be worth far less a year later. More people betting against it, he notes, often raises the odds they will be wrong. The outcomes are too spread to call cheap or dear with confidence.

What a real mania looks like? Go back to 2008. The Big Short scene where the stripper owns five condos and does not know her rate. Shapiro asks: the last dancer you spoke to, was she pitching leveraged Micron and Nvidia ten times? No. If the chatter is leverage, it would be mania; if the chatter is AI ending humanity, it is fear, not mania. In the internet mania the talk was the internet will change the world — and it did — yet many names still fell 80-90%, some 100% first. The frenzy then was a thousand times louder than now. Could it become a bubble? Yes. If it does, you want to be long, not short, because the inflating phase pays the long.

A Pew Research survey on AI and jobs mirrors that fear split. In wealthier countries (Australia, South Korea, US, Greece) more respondents say AI will mean fewer jobs (blue) than not make a difference (green); in developing countries optimism is higher that AI will add jobs. Shapiro tells three anecdotes: a young dentist, great at her work, who insists AI cannot do what she does with gums and lasers; his preschool-teacher wife who believes parents will always want human care for toddlers; his lawyer brother who says AI gives only wrong answers so he is not worried. Everyone worries about everyone else's job but not their own. Shapiro's punchline is blunt: if AI is close to what is advertised, it will probably take all our jobs and be a lot smarter than all of us — himself included.

Where remains human edge? Shapiro gives a case where rules said buy bonds but human context said fade. His favorite fade YouTuber, his Discord crowd, and a corporate fixed-income desk all pointed to the same crowded contrarian long; three whispers became a signal. Can AI suss that out? If it were a thousand times smarter than him, it should, but will it? He leaves the question open. The signal lives not in a chart alone but in the room's whisper — a model would need to hear chatter, not just candles.

Where the crowded trades sit and what it means

The most crowded trades today sit where few retail eyes look: agriculture. A super El Niño impulse plus the worsening war in Ukraine, a key wheat and grain source, lifted grain markets and pulled a huge speculative chase into sugar, corn, soybeans and soymeal. That crowded long could be a mild optimism for food inflation plateauing: if those longs unwind, food-at-home prices flatten. Shapiro adds two qualifications. First, crowded does not mean must turn; it only gives asymmetry. In his words 70% of his trades lose, so the flip is still 50/50 heads-tails, but the payoff shifts: win 70 cents, lose 30 cents. Second, even if it works, the move may last three to four weeks, perhaps 15-20% down, then the fundamental resumes straight back up. Crowdedness gives a tactical fade, not a strategic reversal.

Copper is the cleanest illustration of that asymmetry. The demand story is 100% true: AI infrastructure, data centers, power grids, EVs, defense and electrification all need copper; the UN warned on 9 May 2025 that a looming shortage could stall the energy transition, and IEA and Bloomberg peg 2025-2026 data-center demand alone at an extra 475,000 tonnes a year with overall copper demand up 30% into 2040. What happened? Two months ago copper got very crowded and then just sat sideways, and a fall a week ago washed out about 15% of longs. Not because the thesis changed but because positioning did: a break of the 20-day average stopped out trend followers, the crowd thinned. Shapiro's lesson is crisp: fundamentals can be right and still not lift price if everyone is already long — it takes fresh money to push higher, and fresh money is scarce when crowded. Now he is neither long nor short; the crowded long was thinned and the edge is gone.

That decision depends on process. A fundamental investor who believes that thesis must be long copper somehow — via the metal or a copper stock — but should own within that universe what shows the best relative strength, the names trading best. That is sound trading craft: follow strength inside the thesis. Shapiro does the opposite because his mandate is to be negatively correlated to trend followers. Trend followers are and should be long copper here, near a new high; his clients already get that exposure in their trend portfolios. His job is not to pile on but to hunt crowded longs for asymmetric shorts. Like hearing the double bass while everyone plays violin — quiet but essential for balance.

Gold closes the same loop. Price is still near highs versus a year ago, but sideways action killed chatter and sentiment faded. After a semi-frenzy into early this year, interest waned as price waned. Shapiro does not see a crowded setup now, does not trade it well, and will not force it. He uses a hunting metaphor: wait in the weeds for the injured deer to limp by, then pounce for the easy meal. If not easy, wait. For a long-term holder who believes fiat is poor and debt addiction ends badly, owning some gold for diversification makes sense — that is insurance, not a trade. The question remains the same as elsewhere: is there crowdedness, is there asymmetry, is timing right?