The weekender: You can’t keep a good market down

By late Friday, the tape felt like one of those old trading-floor parables: you push a good man underwater, he’ll stay there for a moment, but eventually he bursts back through the surface gasping, swinging, and ready for another round. After the week we just lived through—Japan’s bond seas lurching like a ruptured tanker, AI darlings convulsing with bubble-adjacent tremors, and retail momentum trades getting whipped like kites in a monsoon—you’d expect exhaustion. Instead, the market inhaled, steadied, and drove higher as if to say: you can bend me, but you can’t break me.

The S&P 500 staged a near-2% comeback with roughly 470 names advancing—an astonishing show of breadth for a market allegedly “saturated with risk.” Nvidia, the storm barometer of the AI age, erased a 4% intraday plunge and snapped back like a carbon-fibre spring. And right on cue, the rumour mill lit up: Trump’s team quietly floated the idea of selling Nvidia’s H200 chips to China. True or not, that whisper alone was enough to put a speculative sheen back onto the froth layer. Markets don’t need facts; they just need possibilities.

But this move wasn’t just rumour-grease. It was liquidity-physics. The real accelerant came from New York Fed President John Williams, who cracked the door on a near-term cut. In an environment where macro conviction is fragile and everyone’s staring at the same clouds gathering over the AI bubble, even a sliver of dovish daylight is enough to unleash the rate-cut animal spirits. December cut odds surged from 40% to 75% within hours—an almost cartoonish repricing that would have been unthinkable eight days ago. Treasury yields drifted lower, and outside of the Yen (which unloaded some JGB fear baggage), counterintuitively, the dollar stayed aloft at six-month highs, and Bitcoin—after briefly staring into the abyss—managed to claw back above $84k.

Still, the real story wasn’t the rebound; it was the reality check. November, usually a seasonally blessed month, now looks set to deliver the worst performance since March. And that’s the uncomfortable truth traders keep circling: the broader narrative hasn’t broken, but it’s definitely being stress-tested. When a market bends with no obvious smoking gun other than the unknowables swirling around AI valuations and the Japanese yield earthquake, you must assume positioning—not fundamentals—is driving the ship.

Friday also came with technical accelerants. Roughly $3.1 trillion in option notional rolled off the board, and the expiry flows alone made the intraday swings feel like trying to trade inside a jet engine. Systematic supply and discretionary de-risking have been washing through the market all week; Williams merely provided the narrative pretext for a relief rally.

But here’s the uncomfortable footnote: we still haven’t filled Thursday’s gap. That unfinished business keeps the weeds thick. My instinct is that we’ve seen the first real shiver of capitulation—the kind that transfers risk from weak to strong hands—but not the final one. The Thursday hangover hasn’t fully cleared. There’s still too much latent supply from CTAs, vol-target funds, and hedge-fund gross downshifts waiting to be expressed if we wobble again.

Which leaves the market in that familiar liminal zone: bruised but breathing, stressed but not broken. If the December cut repricing sticks, there is a good chance we will slingshot into a year-end melt-up if we can clear 6725 on the S&P next week. If not, the reset may drag a little deeper into month-end and beyond before we finally get the final rate cut push for the Santa Rally.

And here’s the part everyone tiptoes around, but every real trader feels in their bones: liquidity is garbage. The street is running with razor-thin top of book, so institutional desks can’t sell size without blowing through three levels on the order stack, and they can’t buy size without leaving a footprint the size of a yeti. That’s the unspoken amplifier in this tape—every flow gets exaggerated because the pipes are so narrow.

It’s not just that we’re short gamma; it’s that the market depth is a kiddie pool. Dealers are running balance-sheet anorexic, ETF hedging is automated and blunt, and the buyside is terrified of becoming the next one to mark their book 200 bps lower in one clip.

So every shove—whether it’s machines dumping, CTAs de-leveraging, or vol-control funds adjusting—reverberates like it’s going through a megaphone. But liquidity deserts work both ways: they turn downdrafts into air pockets, but they also turn rebounds into rocket launches.

Yet the quants aren’t panicking—and that’s usually the tell. A 3.5% intraday faceplant into the 100-day looks horrific on the chart, but QDS points out that this exact pattern—ugly reversal meets major moving average—historically resolves bullishly, not as the start of a new down-leg. The reason is simple: what looked like a sentiment collapse was mostly mechanical flow, not a philosophical one. Short-gamma dealers chasing price, levered ETFs puking exposure, and a cascade of forced hedging—not a rewrite of the macro book.

And yes, when the street is short gamma you always have the risk of one more violent overshoot. That’s part of the script. But these technical flushes tend to be the kind of turbulence traders buy when they have the stomach for it. Hedge-fund net exposures are now likely under 50%—and every time we’ve hit that depth this cycle, investors have treated it as a positioning low, not a structural warning light.

Trader lens: The AI Manhattan project

As I’ve hammered in my recent AI notes, if this really is an “AI Manhattan Project,” then the whole exercise of trying to size the ROI before the concrete is even dry becomes almost pointless. Manhattan Projects don’t wait for IRRs—they bulldoze capital toward strategic inevitability. In that kind of regime, investors will remain prone to periodic fits of exuberance in the more speculative corners of the market, just like October. Not because the fundamentals scream value, but because the spend won’t be constrained—and even misallocated capital itself accelerates the trajectory of the underlying technology.

This isn’t a normal capex cycle where CFOs sharpen pencils and justify cap tables. This is geopolitical, epochal, arms-race spending. And arms races don’t slow down for discounted cash flow models.

To borrow a line from top Goldman Sachs trader Mark Wilson:

“Most of the time macro drives equities, but there are rare periods when equities drive macro.”

We’re living through one of those periods—and AI is the catalyst. When a technology becomes a national-priority variable, markets start dictating the macro narrative rather than simply reacting to it.

But here’s the rub: AI cannot always be traded like a macro factor. And that’s exactly where the market has misfired. The microarchitecture of the AI ecosystem—specifically, who’s building what, who’s leading, and who’s being disrupted—has shifted faster than the macro consensus can digest. The street has been laser-focused on the OpenAI orbit and the handful of names hard-wired to that ecosystem. The trade became overly concentrated, over-owned, and dangerously blind to competitive asymmetry.

And then this week landed like a thunderclap. Google’s Gemini suddenly stopped being the punchline and started looking like a genuine step-change competitor—a model that, at least in certain domains, is now pressing beyond ChatGPT. The market wasn’t priced for that. Positioning wasn’t aligned for that. And the AI-as-macro trade was never built to handle a micro shock of this magnitude.

So while the big picture still screams “AI Manhattan Project”—unconstrained spending, strategic urgency, and multi-cycle tailwinds—the leadership within that project is no longer static. The ecosystem is mutating faster than investors expected, and any break in that narrow leadership has immediate consequences for broader risk sentiment.

Stale data, fresh problems

The Fed is staring at old data and new problems, and the timing couldn’t be worse. September’s delayed Employment Situation report—48 days late thanks to the shutdown—showed payrolls up 119k, but with 33k worth of downward revisions that leave August now at –4k. Two negative prints in four months isn’t noise; historically, that’s recession territory. The three-month trend sits at just +62k, the weakest since the pandemic era and before that the post-GFC drag.

Unemployment ticked up again to 4.4%, and the unrounded figure (4.440%) tells you it’s not a borderline move. That’s three straight increases as job growth fails to keep up with new entrants. Part of this is employers turning cautious. Part of it is the quiet spread of AI tools across the workplace—incrementally lifting productivity while trimming the marginal hiring need.

The bigger issue is what the Fed can’t see. The BLS just pushed the November jobs report to December 16—six days after the next FOMC meeting—and October’s release will be partial, with no unemployment rate. So the Fed will have to stitch together two months of labor-market conditions using private-sector proxies, high-frequency trackers, and sentiment surveys. They managed this patchwork for one missing month earlier—but doing it for two raises the odds of misreading the true state of the economy.

The minutes already showed the Committee’s conviction wobbling. “Many” participants were leaning toward keeping rates unchanged for the rest of the year—a minority, but a meaningful one. And the September dot plot revealed how fragile the two-cut consensus really was: ten members wanted two or more cuts, nine wanted one or none. That’s not conviction; that’s a coin toss.

Now inflation data are also going dark. September CPI made it into the October meeting, but 40% of the index was imputed because BLS staffing was thin—hardly ideal for assessing price stability. And the inflation trend inside that imperfect data set was uncomfortable: three- and six-month changes accelerated across almost every primary core measure.

The BEA hasn’t provided dates for PCEPI updates. October’s CPI has been canceled outright, and November’s CPI—scheduled for December 10, the morning of the Fed meeting—has been delayed to December 18. So on decision day, the Fed will be flying even blinder than last month: no official jobs data, no official inflation data, and a split committee trying to make a call off stale information and surrogate indicators.

Rates remain slightly restrictive, and the Fed’s risk-management tilt still slightly favors the labor-market weakness over the stickier inflation read-throughs. That keeps the door cracked for one more cut in December. Markets are pricing about 70% odds—after swinging from below 35% only days ago.

Bottom line:

The Fed is making a close call with incomplete instruments, stale readings, and a labor market that looks softer beneath the surface. In this environment, uncertainty isn’t a risk—it’s the baseline.

Charts of the week

Prediction markets on the Fed’s December decision have swung sharply over the past 24 hours. Volatility spiked after the cancellation of the jobs report, while remarks from influential Fed member Williams reignited rate-cut bets.

Germany now imports more capital goods from China than it exports

Germany crossing this line is a big tell. For the first time, it now imports more capital goods from China than it exports—a clean break from decades when German machinery dominated global industry. That reversal is forcing Berlin to rethink its entire industrial strategy: cut dependence on China, rebuild domestic capability, and reignite innovation.

But this isn’t just Germany’s problem. It’s happening everywhere. Countries are reshoring advanced manufacturing, hardening supply chains, and pouring capital into energy, defence, and infrastructure.

This isn’t simple deglobalization.
It’s the early stage of a global industrial renaissance.

Liquidity thin a function of the VIX bounce?