Ken Griffin then wandered into an already open wound and poured in a little salt by saying AI is “almost certainly” in a bubble. Coming from someone sitting near the centre of the modern capital-markets money machine, that lands a little differently than another social-media hot take. It does not mean the technology is fake, the demand is imaginary or the buildout stops tomorrow. It means the market may finally be rediscovering an old distinction it tends to forget during booms: transformative technology and attractive investment returns are not the same thing.
Takeaways by Dark Side of the Boom™
• The Nasdaq fell more than 2% for its worst session since June, and the bigger warning was that a violent Treasury rally and softer Fed expectations still could not put a floor under tech.
• Meta forced free cash flow, funding costs and the return on AI capex back onto centre stage, while semis, momentum and hyperscaler credit all showed that the market is finally separating technological promise from investment economics.
• The long AI, short bonds trade was hit from both sides at once, while Brent above $110 left traders staring at the awkward combination of a growth scare and an energy shock.
• The yen and Swiss franc have moved to the front of the FX hedge queue, and the yen catching a bid despite the oil shock is a particularly telling sign of how defensive positioning has become.
Just a quick heads-up from Luang Prabang, where I’m preparing for tomorrow’s half marathon.
Today should see a pretty full stream of blog posts, but Saturday will likely be quiet as I’ll be out on the course rather than behind my screens.
Sunday is set aside for some sightseeing and the journey back to Bangkok, so I probably won’t get around to my usual run of Weekender notes until Monday.
Thanks, as always, for reading and supporting The Dark Side of the Boom™. Wishing everyone a great weekend, and I’ll be back at the screens soon enough!
The Pain Trade Comes Apart at the Seams
Wall Street finally saw what happens when the bond market offers a helping hand, and tech refuses to take it. The Nasdaq fell more than 2% for its worst session since June, semiconductors were mauled, Meta sat in the middle of the wreckage, and the S&P was dragged lower even as Treasury yields collapsed and Fed-hike odds were pared back. That should have been the silver lining. Instead, tech looked straight through it. When the implied discount rate falls, and the growth complex still cannot catch a bid, the market is telling you the problem has moved beyond valuation.
That's when the lightbulb over the equity desk stopped flickering amber, and someone kicked the fuse box in.
For months, the market had been happy to treat every larger AI capex number as proof of conviction, every data-center announcement as future revenue and every fresh financing need as somebody else’s problem. Meta changed the conversation because free cash flow suddenly walked back into the room wearing muddy boots, and once that happened the equity crowd could no longer pretend the credit desk had simply been worrying for sport.
The issue is not that AI demand has disappeared. Far from it. The issue is that the market has finally started asking whether demand and returns are still travelling in the same direction.
Meta can still be right about the size of the opportunity and wrong about what shareholders should be willing to pay for the journey. Lam Research can still sit in the middle of a secular buildout and disappoint. Tesla can still sell the dream of AI and robotics while cutting capex. Once those pieces land together, the old reflex of “more spending equals more upside” starts looking like one of those trades that worked beautifully right up until everyone owned it.
The price action told you that before the analysts did. The Nasdaq cracked, semis got hit harder, the Mag 7 offered no sanctuary, and high-beta momentum was taken apart with the sort of violence you usually only see when positioning has become part of the fundamental story. Neoclouds stayed under pressure, optical networking got dragged into the undertow, and the entire AI complex suddenly looked less like one secular winner and more like a crowded neighbourhood where everybody was trying to get through the same exit door at once.
Goldman’s prime-broker data made the point even more clearly. Fundamental Long/Short, Systematic and Multi-Strategy books were all down more than 1% on the same day for the first time since Covid, while the hedge-fund VIP long-versus-short basket suffered one of its ugliest sessions in years. Long-only accounts were heavy sellers, hedge funds leaned the same way, and there was essentially no sign of anyone stepping in to play hero around the AI theme.
That is when you stop calling it a pullback and start calling it pain.
Markets are very good at confusing the two when the money is easy.
They get much better at separating them when the financing starts to bite.
And that brings us to the other half of the trade, because the real story Thursday was not simply that tech fell. It was that long AI and short bonds, one of the most comfortable expressions of the year, got hit from both sides at once.
For months, that trade made everyone at the table look clever. Own the AI winners, fade duration, collect the momentum, and let strong nominal growth, heavy issuance, sticky inflation and endless capex do the rest. Thursday was the day the dealer pulled both cards at once. Tech rolled over, Treasuries ripped, and the same book that had been collecting carry and momentum suddenly started leaking from both ends.
Yields fell by as much as 20 basis points across the curve, money markets backed away from fully pricing another Fed hike before year-end, and Treasuries delivered their biggest rally in more than a year. Normally, that kind of move would be jet fuel for long-duration growth. Instead, tech kept sliding.
If the market hands you a cheaper discount rate and you still cannot rally the growth complex, then the problem is no longer just multiples. It is whether the earnings path, the financing path and the capex path still line up well enough to justify the price.
That is where the credit market has been trying to drag equities for weeks.
Hyperscaler spreads have widened, AI-related CDS has pushed out, and the assumption that the industry can fund every new cluster, every new data center and every new power connection at friendly terms is starting to look less like analysis and more like hope. Demand can still be enormous, but supply of capital is not infinite, and once that distinction gets priced properly the market stops rewarding ambition and starts rewarding balance-sheet endurance.
What makes the whole setup more awkward is that the usual macro escape hatch is not particularly clean either.
Normally, a growth scare strong enough to produce a Treasury move like this would drag oil lower and give the market a cleaner disinflation story. Instead, traders are being asked to price softer growth on one screen and a harder inflation impulse on another. That is not a friendly probability tree. It is the sort of setup where nobody is looking for the perfect hedge anymore. They are looking for the position least likely to punch them in the face.
The Middle East risk premium is doing the punching.
Tankers have been attacked, there have been reports of mines near the Strait of Hormuz, Saudi facilities have been hit, embassy personnel have been moved and the market is still digesting the possibility of renewed US-Israeli strikes on Iran after the midterms. WTI has pushed back toward the old war highs, diesel has refused to behave politely, tanker rates remain extreme and the physical barrel is carrying a lot more friction than the headline supply numbers suggest.
The market is not simply pricing whether the barrels exist.
It is pricing what it costs to move them, insure them, refine them and replace them if one of these attacks finally turns from a geopolitical headline into a genuine logistics problem. Middle East exports can remain healthy and the oil market can still tighten through freight, insurance and refined products.
That is why Brent above $110 is more than an energy story.
It is a macro problem sitting on top of an equity problem.
That tells you the growth and financial-conditions impulse was powerful enough to dominate the immediate inflation signal, at least for one session. But if crude and diesel stay elevated, the bond refuge comes with a catch. The Fed can become less hawkish because equities and credit are tightening conditions for it, but an energy shock gives policymakers less room to ride to the rescue if growth starts slipping.
That is not the sort of backdrop where traders sleep particularly well.
The FX market is already voting accordingly.
The yen and Swiss franc have moved to the front of the hedge queue, and the yen is the more interesting tell because Japan’s energy-import dependence should, in theory, make an oil spike uncomfortable for JPY. Yet the currency is still catching a defensive bid. Traders are looking through the terms-of-trade hit and buying the hedge function anyway.
That tells you where the fear is sitting.
When the yen gets bought with Brent above $110, the market is not asking which currency has the prettiest macro story. It is asking where there is still somewhere quiet to stand while the rest of the room starts throwing furniture.
Gold did not provide much comfort. Bitcoin certainly did not. The dollar did not deliver the classic one-way risk-off surge either, which left the safe-haven map looking patchier than usual. Treasuries worked. The yen worked. The Swiss franc worked. Beyond that, there were not many places where you could hide without checking the exits first.
China adds another uncomfortable edge. The CSI 300 is now more than 20% below its June peak, putting the index into a bear market while the yuan sits near a four-year high and 10-year yields hover near record lows. Domestic demand remains weak, foreign investment inflows have softened and the local market is entering this global wobble without much internal momentum to lean on.
France is another loose thread. Sovereign spreads have widened sharply as political and fiscal stress deepen, and while that does not mean a euro-area crisis is around the corner, it does mean Europe is not exactly standing there with a safety net while US tech is being repriced and China is already in a bear market.
And that, more than any single percentage move, is what Thursday was really about.
This was not indiscriminate liquidation. The Dow and Russell held up far better, equal-weighted S&P performed much better than the headline index, and money rotated into healthcare, consumer and utilities. The market did not throw everything overboard. It took a hatchet to the most crowded part of the book and rediscovered the appeal of businesses that do not need a trillion-dollar capex cycle to justify the next leg of the multiple.
But the warning is in the combination. AI credit is widening. Equity leadership is cracking. Duration is rallying hard. Brent is above $110. Gold and bitcoin are not providing much cover. China is in a bear market. France is flashing sovereign stress. The yen is catching a bid despite an oil shock.
That is a market telling you the old map is not working quite as well as it used to.
The AI cycle itself may carry on perfectly well. Demand remains huge, the technology is real and the strongest companies may still produce exceptional returns. But markets do not trade destiny. They trade the path between today’s price and tomorrow’s cash flow, and that path suddenly looks more expensive, more crowded and a lot less forgiving.
For most of the year, long AI, short bonds and short fear was one of those trades that made everyone look smarter than they probably were.
Thursday was the reminder that crowded trades usually feel safest right before they become expensive.