What We Told Hedge Funds

Happy Friday! I am writing it soaked through after cycling the kids around Copenhagen in what the Danish weather service insists on calling autumn and what the rest of us would call a horizontal rainstorm. The weather has turned very ugly over the past week, which in Denmark means the heating goes on, the gas storage numbers start to matter, and my children start asking why we do not own two cars like normal people (my wife had already left home in the SUV this morning). The honest answer is that a macro strategist who writes about energy scarcity cannot be seen in a Range Rover. So we cycle. And I write wet.

The investment week was kinder than the weather. Micron reported $54.2bn of revenue for the August quarter, up 379% year-on-year, beat consensus by $3bn, and guided the current quarter to $61.5bn against a street of $57bn. Gross margin 87%, free cash flow $62bn for the year, $73bn of cash.

Korea printed September exports up 83.5% year-on-year the night before. Everything we have said about memory since December- that it is the bottleneck, that the contracts change the cycle, that Korea tells you first- landed in one 48-hour window and confirmed our thesis.

We accordingly UP our conviction in a strong Q4 for AI investments. I will come back to it in the In Focus, because it is the biggest call in this letter. First, the macro, because the dollar story is where I think the market is most confused.

Theme 1: The US is softer than its peers… on inflation AND on the rate of change in growth

Start with the data this week. The ISM came in at 54.5 against 55.0 expected, which was well flagged by our nowcasts: the US rising-growth probability has fallen from 27% to 22% in a week, the sharpest drop on the board. The detail is a textbook energy shock: new orders fine, employment fine, and the prices-paid index at 77.9 against 72.3 expected, the highest since May. Headline hot, underlying soft.

And then the NFP, which landed as I was finishing this letter. Payrolls rose 29,000 against a consensus of 84-90,000, the unemployment rate ticked up to 4.2%, and average hourly earnings came in at 0.1% on the month against 0.3% expected, 3.0% on the year. Our withheld-taxes model had a base case just south of 50k, so we were almost spot on again. That is the third correct NFP call in a row, and on the Bloomberg ranking it makes us the best of all forecasters over that stretch. I mention it not to gloat (fine, a little) but because the model is doing exactly what it is built for: reading the tax receipts before the survey catches up.

The detail backs the soft read. Private payrolls rose 46,000, but government fell 17,000, education seasonals swung again, and the gains were concentrated in health care, construction and manufacturing while information, finance and energy shed jobs. Wage growth at 3.0% is not a number a central bank hikes on. The one soft spot in the soft read is that the unemployment rate rose partly because participation rose, which is a healthier reason than layoffs. The market took it the right way: October hike odds fell further, the front end rallied, the dollar gave back ground. Williams said one more and no rush. The data says maybe not even one.

Put it together, and the Fed is going to be dovishly surprised on the growth and labor side of the mandate. That is obviously not its main focus right now; inflation is. But core inflation and inflation BREADTH remain incredibly soft in our readings. The share of CPI components rising fast has dropped back to its long-run average; the PCE decomposition this week showed housing at 0.14% and durables at 0.07%, and the US rising-inflation probability sits at 34%, down from 47% a month ago, with Brent at 100. Headline, not so much. The Fed needs to see it in the official data first, of course, which is why Williams says one more hike and no rush. The market has more priced in. I think Williams is right.

Now compare that with the rest of the world, because this is the point. Tokyo CPI came in hot overnight, and it confirms what our nowcasting keeps repeating: inflation pressures are a LOT firmer outside the US, especially in countries with a terms-of-trade problem on energy products. The Eurozone rising-inflation probability sits at 86%, India at 90%, the UK at 71%, and, e.g., France printed 3.4% on Tuesday with services edging higher. Same energy shock, very different pass-through. The US refines its own diesel and exports the surplus; Europe and Japan import theirs and pay the war premium on every barrel.

And here is the part that makes the dollar tricky. Growth is holding up almost surprisingly well in Japan and Europe. The Eurozone growth probability is 78%, the UK 82%, against 22% in the US. The rate of change in growth is BETTER outside the US than inside it, even in the economies paying the energy bill.

So you have two forces pulling the dollar in opposite directions: inflation and rate spreads that favor the euro and the yen (firmer central banks, firmer growth), and a terms-of-trade and energy risk premium that favors the dollar for as long as the strait is shut. The first is the fundamental case. The second is the thing that has carried the dollar for a month. I am tempted to lean short USD still from here, because the fundamentals are on that side and the energy premium is the channel that fades first when Hormuz talks progress. But admittedly, this is not a crystal-clear case, and I size it that way.

Theme 2: The EUR is weak, the USD is not strong… and Le Pen is not the end of the world

Look at the dollar against everything rather than against the euro alone. Since July 1, the dollar is flat against the euro, down 3% against the yen, down 2% against gold and silver, and down 29% against Bitcoin. That is not a strong dollar. That is a weak euro that happens to be equally weak. The dollar-negative regime we have described since July has been right against every asset except the one everyone watches in Macro, and the euro has had two problems of its own that have nothing to do with America: gas and France.

Take France first, because that is where the market is most wrong, and let me be snarky about this, because the subject invites it. The setup: Le Pen polls at 34-36% in the first round and leads every second-round pairing by a widening margin, the vote is in April, French debt is at 119% of GDP, a post-war record, the 2027 budget presented yesterday promises 54 billion of savings with frozen wages and frozen pensions, the fiscal watchdog called it the minimum effort, two prime ministers have already been toppled over cost-cutting, the OAT-Bund spread is at 111 basis points, the widest since 2012, and Japanese investors have dumped French bonds at a pace we have not seen before. Every one of those facts is true. And every one of them was true, in its American version, in October 2016.

Remember what happened then. A right-wing candidate everyone in finance considered unthinkable was about to win. The commentariat explained, at length, that the world would end: trade wars, debt defaults, the Fed politicized, the dollar’s reserve status in doubt. On election night, S&P futures fell 5%. And then the market opened, went up, and did not stop for four months: plus 12% by March. Business as usual arrived with remarkable speed, because governments are constrained by bond markets, courts and bureaucracies far more than campaigns suggest, and because the fear was already in the price. Everyone fears the world ends when a right-wing candidate takes power. Honestly, much of it is business as usual, and that will be the case in France if Le Pen wins as well. She has spent a decade sanding the edges off her program precisely so the bond market will let her govern.

So this is a HUGE buy-the-rumor, sell-the-fact event… only in reverse: sell the rumor, buy the fact. The rumor is being sold right now, at 111 basis points, by Japanese pension funds and by every strategist who learned the word “Frexit” in 2017 and has been waiting nine years to use it again.

The fact is, if it arrives in May, it will be a president who needs the ECB more than the ECB needs her, a Senate she does not control, and a bond market that has already extracted the premium. I would BUY France if Le Pen wins, especially given how dire the current positioning is. The structural problems are real: a 5% deficit in a 0.7% growth economy, a pension system nobody can touch, and a parliament that cannot pass a budget without a confidence vote. But those are the reasons the spread is at 111, not reasons it goes to 200. The 2017 scare topped out around 80 and reversed the day after the first round.

The broader point: country risk in spreads, credit risk in credit markets, and the other stress signals are all tightening, and that is perfectly normal in a hiking cycle.

France is not the only spread that has widened; Italy, Greece, Belgium and Portugal have all moved with it. A central bank taking rates above neutral stresses the weakest balance sheets first, and in Europe the weakest sovereign balance sheet is French. The question is not whether the stress is real. It is how long the hiking cycle lasts, and I have my doubts about a prolonged one.

Two reasons for those doubts. First, the necessity impulse. Our basket of input costs for food, energy, and chemical producers is up around 25% on the year. It feels bad, because it sits on top of prices that doubled during Covid. But in 2022 and in 2008 the same basket was up 100% or more, and the impulse has already peaked and started to roll. Sequentially, which is what a central bank cares about, this is not a necessity crisis.

The chart says the curve flattener is short-lived, and so is the cycle behind it. Second, the peak of the Iran conflict is behind us. The Revolutionary Guard is running out of money on schedule, Tehran is pushing for a phased reopening, Qatari gas is flowing again at a toll, and the war moves in sine waves. A hiking cycle built on an energy shock ends when the energy shock does.

Which leaves gas, the euro’s other problem, and the one I take more seriously than France. European storage is around 72% full against a seasonal norm near 80%, well below the five-year band. The question is what that means for the winter. We ran three withdrawal paths: a normal winter takes storage to around 17% by April, uncomfortable but survivable; a mild winter leaves it near 40%; a harsh winter like 2021-22 empties it by March. That is the distribution. The base case is fine, the tail is not, and the tail is what the TTF curve is pricing. If the Hormuz talks deliver even a phased reopening of products, the tail shrinks and the euro’s energy discount with it, and remember that we still have a few weeks left to fill up storages here (and 5-6 months to secure better flows, as they are already improving out of Qatar, for example). The TTF stress / Nat Gas fears are, by now, blown out of proportion.

Now the part that turns the analysis into a trade. Positioning is already sour almost everywhere you look. Among individual investors, the AAII bull-bear spread is worse than it was at the peak of the Iran war in March. Among advisors, the bulls have crept up only hesitantly from the lows.

The NAAIM exposure index, the professional allocators, is clearly more bearish than one or two quarters ago; still decently invested, not stretched, but a long way from the levered territory of the spring. The dollar positioning is LONG, though not as long as a week or two ago; the specs have started to trim. And oil markets remain very long, at the same 44% of open interest we saw at the March peak.

Theme 3: But positioning is already VERY sour

Add the market breadth chart, the narrowest since 2000, and you have a market in which everyone owns the same few things and is short or underweight everything else. The setup is mostly ripe for a correction in the direction of a weaker USD, higher risk assets, and lower energy, if you get just a slight improvement in the energy situation.

Maybe it is not the base case. But in a game of weighted probabilities, it remains a decent play: the downside if nothing improves is a market that is already positioned for nothing improving, and the upside if something does is a crowd on the wrong side of three trades at once. That is the asymmetry I want to own into Q4.

In focus: Memory and the (missing) cyclicality… and why Korea tells you to stay upbeat

Let me start with the critics, because they have been loud and they have been wrong. The argument goes that memory is the most cyclical business in semiconductors, that every supercycle felt permanent at the top, and that 2027 or 2028 will look like 2019 or 2023 when the capacity arrives. It is the kind of argument you make when you learned the memory cycle from a textbook in 2015 and have not looked at a contract since.

Here is what has changed, in plain numbers. Micron, SK Hynix, and Samsung have sold out their ENTIRE 2027 DRAM and HBM capacity. Customers are receiving 60-70% of what they asked for, and the allocations are already finalized. Micron has around $100bn of take-or-pay contracts and $22bn of customer deposits, and can fill half to two-thirds of its biggest customers’ demand. Samsung has roughly 70% of its memory capacity under long-term agreements that run to 2031, with Microsoft, Nvidia and Google as the counterparties. SanDisk has $94bn of contract value with average terms over four years and price floors.

Think about what that means for the shape of the cycle. In every previous memory boom, the customer base was thousands of PC and phone makers buying spot, and the moment demand wobbled, pricing collapsed because nobody had committed to anything.

Today, three or four companies, Nvidia first, then Google, Microsoft, and the rest of the hyperscalers, have bought all the HBM for 2027 on multi-year, take-or-pay terms with deposits attached. The cash flows for 2027 and most of 2028 are known today, to within the allocation percentage. That is not a cycle. It is a queue. Yes, capacity arrives in 2028-29, from the M15X fab, from Indiana, from Samsung’s expansion, and yes, CXMT is the overhang the bears point to. But new capacity that is already committed under long-term agreements does not become spot supply. It becomes delivery against contracts that are already signed. Calling that cyclical in the old sense is absolute nonsense, from people who do not understand what the contracts do.

Micron and SanDisk are the US expressions, and they are priced at 7x and 10x run-rate earnings, less than half the market multiple, for businesses with two to three years of contracted visibility and 85%-plus gross margins. The Koreans are the other half of the story, and here the license question matters. SK Hynix and Samsung run roughly one-eighth of the world’s memory output out of their China fabs, Wuxi, Dalian and Xi’an, on a US permit that was converted to a one-year authorization and expires on December 31. Non-renewal would take a chunk of global supply offline overnight, and the incentive structure in Washington leans towards exactly that: Micron and SanDisk can walk into the White House and ask why their Korean competitors get to keep a Chinese supply chain when they do not. Renewal keeps the Koreans whole and the shortage as it is; non-renewal tightens the shortage further, and hands share to the US names. Both outcomes are bullish for memory, but only one of them is bullish for Hynix and Samsung, which is why we prefer the US over Korea going into the decision.

So here is the call, and I will make it big because the evidence is. Korea says the semiconductor complex is pricing less than half of the earnings it will deliver. Micron just delivered a quarter and a guide that nobody on the sell side had. The HBM for 2027 is sold, the DRAM for 2027 is sold, the NAND is on four-year contracts, and the equities trade at single-digit multiples because the market still believes in a cycle that the contracts have abolished.

We see semiconductors up 100% in Q4 as our base case. Not a scenario, the base case. The gap between the Korean export data and the semiconductor equities has never been this wide, and every previous time it opened, it closed upwards. This time the fundamentals are contracted. The only question is how fast the market reads the contracts.

Portfolio Watch: Close to 100% conviction in the AI themes… this is April all over again

The book is up 13.5% year-to-date after a week where several of the Semiconductor names did the heavy lifting and a few of the small positions did the damage. Micron is now +297% on entry at 3.8% of the book, Bloom +197%, CrowdStrike +120%, Dell +128%, SanDisk back to within 7% of entry, and Marvell, bought two weeks ago, is +11%. The losers are the ones we have discussed: Definium -3% on the week as the catalyst vacuum bites, Beam, IREN, Energy Fuels.

For a weekly update, I will only write about what is actually happening, so three things:

First, AI Bottlenecks is now 28.5% of the book with a short-term conviction of 9.1, the highest of any theme, and End-consumer Scarcity, the Dell and HP trade on memory costs hitting the consumer, sits at 12.5% with a 9.0. Between them, that is more than 40% of the portfolio riding on the memory thesis in one form or another, and after this week I am comfortable with every basis point of it. We do not add to Micron after a 297% run, but we do not trim it either. SanDisk is the cheaper way in and the license-cliff option, and we hold it at 6.1%.

Second, we have upped the conviction on two of the Australian counter-drone names, DroneShield and Electro Optic Systems, from 6 and 7 to 8 each. Both have shown recent momentum after a rough quarter; the counter-drone demand from Europe is arriving in the last months of the year as we flagged, and EOS is back above our entry. The Military Drones theme sits at 6.9% of the book. We are not adding yet; conviction goes up before weight does… DroneShield has put in a sequence of higher lows since the August washout, reclaimed its 50-day average, and is pressing the downtrend line from the spring high, which is the first constructive structure the chart has shown in six months. EOS has already done the work: it is back above our entry and above the summer range. In a market that is punishing promises and rewarding delivery, two names with revenue momentum, order momentum, and now price momentum deserve a higher conviction than the 6 and 7 we carried through the summer. Hence the 8s. Weight follows if the breakout holds.

Third, Definium: The conviction has come down to 7 from 10, not because the science changed but because the trade is crowded and the next catalyst is a year away. It remains the largest position at 12.9%. We said last week that we are considering trimming, and we still are; nothing has happened this week to change either the thesis or the sizing question. Beam stays at 9, held through the safety noise.

横向滑动查看完整图表

Trades and themes across asset classes

In FX: short USD as a lean. The dollar is flat against the euro and down against everything else; the fundamental case (soft US core, soft US growth, firmer central banks abroad) favors the euro and the yen; the energy premium carries the dollar. We lean short, not crystal clear, sized accordingly, and we would add on a Hormuz products deal rather than before it.

In Rates: one more Fed hike, no rush, and the market has two priced. The curve flattener is short-lived because the necessity impulse has already peaked. French spreads at 111bp are the rumor being sold; we would own OATs into a Le Pen win, not sell them.

In Commodities: oil longs at the March peak into a war that is past peak escalation. Short the cracks, short the barrel from this zone, gold and silver over crude, Bitcoin the best-performing dollar short in the book.

In Equities: semiconductors up 100% in Q4 as the base case. Memory first (Micron, SanDisk, US over Korea into the license decision), Marvell as the inference expression, the end-consumer scarcity trade in Dell and HP, security through CrowdStrike and SentinelOne. Counter-drones upgraded. Definium held while we weigh the trim.

横向滑动查看完整图表

Have a great weekend, and good luck out there!

Andreas Steno

Andreas is the core macro consultant in Steno Research. Building on years of experience as Global Chief Strategist at Nordea Bank, he is one of the most quoted and sought-after macro analysts out there. Andreas’ expertise is the FX/rates, energy, real estate and equity spaces, but doesn’t shy away from hot takes on other topics if the underlying analysis is strong enough. Andreas anchors the weekly editorial ‘Steno Signals’.