What We Told Hedge Funds This Week
Middle East wreaking havoc on the data...
Writing to you from a cabin next to the tigers, while Trump threatens HUGE attacks on Iran (again). Oil is bid, products are calmer, our July nowcast is still remarkably soft, and we are upping our conviction in hardware after Alphabet officially became a hardware store.
I am writing this week's letter from Knuthenborg Safari Park, where I am sleeping in a cabin right next to the tigers. They pace back and forth behind the glass all evening, looking hungry and vaguely threatening without ever actually attacking anything. I cannot think of a better metaphor for the current oil market.
All eyes are on Iran, the war, and the renewed pressure in energy. Interestingly, the pressure has migrated back into crude itself rather than the products, as crack spreads have come down while oil has come up in recent days. That is the opposite of the set-up we described just last week, and it matters for how you should trade this (more on that below). Meanwhile, it also feels like we are approaching some sort of climax in this round of the conflict. Trump is threatening HUGE attacks on Iran (and the Houthis) again, which is worth paying attention to, since the "1 bridge per ship" rhetoric was EXACTLY the rhetoric he deployed just before the crescendo in early April. This war moves in sinus waves, and we look to be near the top of one...
We recorded our monthly Macro Meets Micro show earlier this week, and I will lean on some of the points from it below, with the important caveat that the Iran/oil situation has gotten worse in the days since we recorded. Both myself and the tigers have been watching the tape deteriorate together.
Macro/Geopolitical update: All eyes on the TACO
The question I get most often right now is whether a TACO (Trump Always Chickens Out, for the uninitiated) is imminent. It does not appear to be. Dated oil prices are still 40-45 dollars below the peaks from the March/April panic, and on most scores the oil market is simply better balanced and better prepared than it was in round one. The world has had months to reroute, diversify and build buffers, and it shows up everywhere in the physical data.
The cleanest way to see it is through the Houston to WTI premium, which is the international "panic bid" for U.S. barrels. In March/April it spiked to 7 dollars as everyone screamed at the U.S. to export, export, export. Today it sits around 1.4 dollars. Nobody is begging for American oil this time. The prompt spreads tell the same story, as we have seen a re-acceleration in the Brent front spread, but nothing that even remotely resembles the backwardation panic of the spring. The market rarely discounts the same event twice, and so far that old rule is holding beautifully.
A source for concern is obviously that the Houthi militia in Yemen has now opened a new front in the war directly against Saudi oil exports through the Bab el-Mandeb — meaning the Southeast into the Indian Ocean and beyond. But the picture is more nuanced than a simple blockade.
For one, Saudi Arabia can still ship oil north through the Suez Canal. This isn't a full chokehold on exports - it just makes the eastward route far more complicated and expensive. That detail is telling: it suggests the campaign is partly aimed at the Chinese market. Iran would rather see Beijing buy up its own floating oil buffer than turn back to Saudi barrels once it resumes purchasing. Second, the Saudis carry outsized influence with Trump, yet remain acutely exposed to the standing threat of drone strikes on high-profile events like Formula 1 races, golf tournaments, and football matches involving Cristiano Ronaldo. Riyadh spent seven fruitless years trying to crush the Houthis outright, and MBS has since gone to considerable lengths to keep his country out of the current war.
Third, it would be a mistake to read the Houthis as merely an Iranian proxy. They operate with real independence and confidence, and this offensive may simply be their own opportunity to press leverage against the Saudis and extract further concessions.
None of this, in my view, necessarily pushes back the timeline for a U.S.-Iran peace process. If anything, it could accelerate it, particularly if China and/or Pakistan decide to get diplomacy moving.
The new wrinkle is on the deal-making front. The new agreement with Saudi Arabia, which includes Saudi acceptance of Israel and entry into the Abraham Accords, cuts both ways for the Iran deal prospects. On one hand, it isolates Tehran further and increases the economic gravity of eventually joining some sort of regional arrangement. On the other hand, a cornered Iran with fewer friends in the neighborhood is an Iran with less to lose from escalating, and the Saudi accession is precisely what triggered the Houthis to finally make themselves relevant in this conflict after sleeping through the spring.
And the Houthis genuinely make this round trickier than round one. They have declared a maritime blockade against Saudi Arabia, a Saudi tanker has already been hit in the Red Sea, and Kpler has now released a shipping tracker for the strait around Yemen on top of the one for the Strait of Hormuz. We officially have TWO straits to worry about. One strait used to be enough drama for a full market cycle; now we get a two-for-one. Bab el-Mandeb saw 7.4 million barrels a day transit in June, roughly 7% of global output, so this is not a rounding error if they actually start hitting tankers systematically.
Still, our base case remains that markets will treat this LESS severely than round one. The panic premium is a depreciating asset, the rerouting infrastructure exists now, positioning is nowhere near as offside, and every physical indicator we track is screaming "known risk" rather than "regime shift". We are not short oil here (that would be terrible risk/reward with so little stress priced), but we are also not treating this as March 2.0.
Meanwhile, our nowcasts continue to show VERY soft inflation for July, and yes, that includes the re-acceleration in gasoline, diesel and jet fuel at the pump. It will be a surprise to MANY that inflation prints soft again in July. The problem, of course, is that every hawk on the committee has now been handed a beautiful excuse to not care about it, because "oil is rising". And I am genuinely saddened to watch how many of the wrong oil pundits are almost celebrating that the war has restarted, simply because it makes them slightly less out of tune with reality. Rooting for a war to rescue your forecast is quite the look...
Portfolio "In Focus": The Bloom Energy Lie - the concrete decoupling case...
Everything that relates to metals sits in our decoupling basket, i.e., everything that relates to the divorce between the U.S. and China. And these past weeks, the decoupling theme got its most concrete (and most entertaining) case study yet, courtesy of Hunterbrook's short-selling report on Bloom Energy, aptly titled "Bloom's Big Lie". I have read it, and credit where credit is due, it is really good.
The short version is this. Bloom Energy is a fuel cell company with incredible success and a tremendous pipeline looking 4-5 years ahead. Their fuel cells use a LOT of scandium, one of those rare earths that is basically a non-existing market. How non-existing? Global annual production is roughly 81 tonnes, of which China produces around 55. My household probably produces more garbage in a few years than the entire world produces scandium. And within that microscopic market, Bloom Energy is now the biggest buyer of scandium ON EARTH, buying around 22 tonnes a year. Bigger than the U.S. defense complex, which uses the stuff in fighters and missiles.
The credibility problem is that Bloom has been vocal about NOT being China-dependent in their supply chain, and even reiterated a "diversified supply mix" in an official exchange filing after the report dropped. I simply cannot get the math to support that. Can you source 22 tonnes outside of China? Add up the rest of the world, and you get to roughly 26 tonnes (8+7+4+4+3, give or take). So barely, in theory, IF Bloom had hoovered up essentially every gram of scandium produced by every country other than China. Not a snowball's chance in hell. So yes, they are lying, and Hunterbrook even found a sales guy at one of the Chinese rare earth companies happily confirming that they sell to Bloom.
But here is where I part ways with the short thesis, and it is an important nuance for the whole decoupling debate. Bloom has a credibility issue, not a supply issue. China is perfectly willing to provide the metal. The draconian part of the Chinese rare earth licensing regime applies to weapons, and Bloom is not producing weapons. So what exactly is the problem, as long as the scandium keeps flowing? Apparently the licensing regime has become such a hot potato that U.S. companies are now willing to lie about their exposure rather than just admit it, which is an interesting microcosm in itself.
Can Bloom "un-China" the supply chain over time? Actually yes, and this is the part most people get wrong about rare earths. Rare earths are not rare. Scandium is not rare at all. The reason these markets are difficult is that they are not SCALED, because historically nobody needed more than a few dozen tonnes (twenty years ago this was almost exclusively a jet fighter input). Bloom plans to roughly 5x production capacity over the coming years, implying scandium demand of around 100 tonnes annually versus 26 tonnes of ex-China supply today. That sounds impossible until you remember that a huge, creditworthy buyer can simply walk into Rio Tinto's office and say "we will pay for the mining set-up". Increasing scandium output 30-40% over five years is absolutely doable when someone finally provides the demand certainty.
So my conclusion is that this is a tempest in a teapot. They lied, they deserve the scolding, and they have to regain credibility. But the stock sold off roughly 25% on a credibility discount, not a supply crisis, and the next positive trigger is fairly easy to imagine, i.e., an announced partnership with a mining company guaranteeing scandium supply for the five-year plan. Bloom looks like a buy here. They have a PR problem, but the metals are flowing, and the pipeline is intact.
Portfolio Update: The scalers are upping their convictions, leading me to up mine in hardware
$7
And here is the pleasant surprise: we are actually UP on the week by the time of writing. That tells you the momentum rout has passed. The market has gotten more broadly worried (thanks to Iran), which paradoxically makes the stuff that performed all year until the rout suddenly look attractive again. Call it the "we know it works" logic. When macro uncertainty rises, capital crawls back to the trades with proven fundamentals, and nothing in this market has more proven fundamentals than the hardware complex.
On the rout itself, remember what actually happened. Q2 was the best quarter for semiconductors ever, forcing institutional rebalancing at exactly the moment the Korean leverage casino blew up (roughly 3% of the Korean market cap sat in levered plays, and 400,000 margin accounts were wiped out in a couple of weeks). The Goldman prime book data shows the biggest tech book readjustment in the entire history of the data set, roughly 10% of gross market value sold in eight weeks. Those flows tend to revert. And for the sake of transparency, I miscalculated how stretched this trade was going into the month.
I gave my wife a "1 cent update" on the holding company, and I am not looking forward to the follow-up meeting. But positioning is no longer stretched, the technical damage is done, and the backtest says momentum at sub-30 RSI has historically delivered about 9.5% over the following three months versus 6% in any random period.
Which brings me to memory, where I am upping my conviction. Through this entire bizarre episode, everything FUNDAMENTAL in memory went up while the stocks went down. Spot prices up, contract prices up, order books up. Nothing happened to the businesses, which is exactly why I consider the rout technical in nature. A lot of people still call memory a commodity. This does not look like a commodity on my screens.
The most telling chart in our whole current deck is the simplest one. Micron and Hynix are today exactly where Nvidia was in 2023. Earnings growth will peak in percentage terms this year, without a doubt, because you cannot compound 1000% growth rates (if you extrapolate those, go touch grass).
But look at what Nvidia's stock did in 2023, 2024, and 2025 DESPITE the peak in percentage earnings growth. If Micron merely doubles earnings from this base in 2027, that is an outrageous number in dollar terms. Consensus projections put Micron's free cash flow around 400bn dollars over the next 24 months against a roughly 1 trillion market cap. If those projections are anywhere near true, they could buy back 40-50% of the company with two years of cash flow. Who cares about skeptical stockholders, you can just buy them.
And why am I so certain memory IS a Nvidia story? Because the HBM content per GPU just keeps climbing with every chip generation. If you are a data center designer, this is a package deal by now. You need the Nvidias, and you need the high-bandwidth memory to feed them. You cannot use one without the other. That is why it was so bizarre to watch Nvidia trade up while Micron and Hynix traded down on the same days during the rout. It is the same trade, a bundle trade, and the divergence was pure cognitive dissonance (with a healthy assist from the Korean margin calls).
People said the same about Nvidia in 2023, that it was "just desktop graphics processors, a commodity with no margin to be earned". We know how that aged.
Zooming out, the roadmap remains the sequential bottleneck framework: GPUs peaked as the binding constraint in 2023-24 (but remains a thing), memory is mid-cycle now (and remember, the final third of the compute/GPU cycle still delivered great returns, which is roughly where memory sits), and power is next, where we are still very early.
That is why I keep a few lottery tickets in the power semi space (Wolfspeed being the obvious one, a 2027 story that I am being as patient with as possible), while the core of the book stays in memory and hardware.
USD shorts and the inflation divergence
So where does new cash go this week? The same answer I gave on the show earlier this week: memory, Jordan. Sweet.
As for the USD shorts, the renewed escalation of the war makes it difficult to lean aggressively into that trade, even though our own data suggests the ECB should ultimately be more hawkish than the Fed. Inflation is ticking up slightly in the Eurozone while falling outright in the U.S., creating one of the widest nowcasting divergences I can remember.
That also has implications for the broader short USD thesis. As long as natural gas in Europe remains a concern, it is hard to build conviction against the dollar, which is why I am also holding back on adding to metals.
Even though our nowcasts continue to point to very soft U.S. inflation, I doubt we will see a meaningfully weaker USD until inflation cools further without the war giving central bankers an excuse to stay hawkish. Despite having access to thousands of research reports, sophisticated nowcasts, and endless data, central bankers are often remarkably simple in their reaction function:
Oil up = hawkish. They rarely have the bandwidth to look much beyond that.
On Monday, I'll be back with a full update on our nowcasts for liquidity, growth, and inflation across regions, including what's driving the changes, and how it impacts our cross-asset thinking.
No changes this week, as we are in wait-and-see mode, assessing whether we are close to reaching a climax in Iran again. We prefer not to be chopped up in Oil (so have no explicit bias in the book) by Truth Social Posts as we haven't bought early access to the feed, lol.
