Oil refining stocks have returned 73% so far this year. Semiconductors, the trade everyone has been talking about, have returned 45.2%.
The story of 2026 is the artificial-intelligence build-out. The best-performing trade of 2026, on the chart Jack Farley and Max Wiethe pulled up hours after the Fed hiked, is a commodity shock nobody was positioned for — and the one trade that has beaten refining is tankers.
"Yeah this maybe when we look at this year it will be defined by the commodity shock even more than the tremendous growth in the AI capex buildout."
Both men own the trade they are describing and said so on air: Farley holds Marathon Petroleum in the model portfolio he publishes, and Wiethe has owned Marathon Petroleum and Valero for years.
The full segment is covered here so you can skip it.
Here are the 10 takeaways that matter.
🎙️ Hosts: Jack Farley, who runs the Monetary Matters Network and publishes a model portfolio on his Substack, and Max Wiethe, his business partner at the network
🧩 Other segments: Joseph Wang, CIO of Monetary Macro, on the Fed's hawkish hike and Kevin Warsh's press conference
📰 Published: 16 September 2026 on YouTube (Monetary Matters with Jack Farley)
🔴 YouTube | ⏱️ length not available
Key Takeaways
The best trade of 2026 is refining, not semiconductors, and tankers have beaten both
Oil refining has returned 73% this year against 45.2% for semiconductors
Marathon Petroleum's refining segment earned $6.6B in a quarter, close to the June 2022 oil-shock peak
The crack spread blew out because governments hold strategic reserves of crude oil and none of refined products
Refining capacity shut in behind the Strait of Hormuz cannot be replaced quickly, because almost no new refineries get built
Farley and Wiethe disagree on peak earnings: Farley says the forecasts are too high, Wiethe expects a plateau rather than a spike and a crash
Diesel at $6 is 4% to 30% of a mining operation's all-in sustaining costs, so the shock spreads to metals
About half the world's sulfur passes through the Strait of Hormuz, and half of that goes into phosphate fertilizer
Farley says semiconductor equipment estimates are criminally low, and he would forecast $10.5B to $11B against the consensus $9.6B
The AI build-out is exposed at the short end of the curve, not the long end
Hyperscalers can absorb higher rates; the neoclouds and private-credit borrowers fund in the overnight market
1. The SEP Stole the Show
Farley opened the second half of the livestream by calling this the biggest and most consequential Fed meeting of Kevin Warsh's tenure so far, and said the market moves were larger than usual because there was a genuine surprise — nobody knew going in whether the hike would be dovish, hawkish, or not happen at all. Wiethe disagreed on the size of the moves.
Wiethe said the volatility was in line with what recent meetings have produced, given that the last meeting carried more uncertainty and more than 90% of a hike was priced in.
What actually surprised him was the projections, not the press conference: "I really do think that at the SEP and the raising of the target of rates, not just for next year, but looking out two years, was probably the biggest surprise because it shows that, rates are not just going up for a little bit and then coming right back down."
He noted that Warsh opted out of the projections himself, which makes them a reading of the rest of the committee
Farley agreed and said he did not know why the chair is choosing not to participate.
2. Refining Beat Semis
Farley put a total-return chart on screen and said the trade of the year is no longer semiconductors.
"Oil refining now up 73% total return. Semiconductors quote unquote only up 45.2%." He said Wiethe had pointed it out to him shortly before they recorded and he had not known it.
"Yeah this maybe when we look at this year it will be defined by the commodity shock even more than the tremendous growth in the AI capex buildout." Far more investors own semiconductor stocks than refining stocks, he added.
Wiethe drew the distinction between the narrative and the trade: "I think it's going to be hard to steal the thunder of the AI being the story of 2026." The story and the best-performing position are not the same thing.
A friend of theirs had told Wiethe that one trade has beaten refining this year: tankers. There is no tanker ETF to chart, he said, only individual stocks.
3. Marathon's $6.6B Quarter
Farley walked through Marathon Petroleum's segment reporting to show where the money is coming from.
"Really it's all refining. So they have a mid-stream business, but it really is the refining business. $6.6 billion in that refining segment."
"And going back that is almost as much as June of 2022 when obviously we had the oil shock then." He expects the current quarter to be higher still.
The host explained the crack spread for the audience: a refiner buys crude oil and turns it into kerosene, gasoline and the other distillates, and the margin on that swings between marginally profitable and extremely profitable. "It's a range and when we have oil shocks, this tends to blow out." The refiner is long that spread.
This is not a volume story. "There's not a lot of growth in refinery capacity because it's kind of an very mature asset class." Throughput is steady; the earnings move is entirely margin.
On that arithmetic Farley said the company will probably earn seven or eight billion dollars in the following quarter.
4. Why the Spread Blew Out
Three separate supply facts explain the margin, and the hosts covered each.
Refining capacity is stuck behind the Strait of Hormuz. Farley said he believed President Trump's claim two months ago that oil is still getting through, and that the US Navy is helping it get through — so the published charts of transit volumes are too low. Jeff Currie, the commodity analyst formerly of Goldman Sachs and Carlyle, made the point that matters: crude getting through is not the constraint, refined products are. "WTI is now above $100."
China has stopped exporting its refined products and has come back to buying crude. Wiethe said China had been hoarding products and not importing oil, which held prices down, and that the Shanghai benchmark is now rising. Farley put it bluntly: "Either they're back in the market or for some reason Shanghai oil is at $130."
The stockpile asymmetry is the piece most investors miss. "Another thing is yes, Max, there are commercial stockpiles of refined products, but really no strategic reserves of refined products."
"And that can help explain why the crack spread has blown out because the world's been short of oil, short of refined products, but the governments have been supplying the crude oil."
Farley's own reaction to the commodity complex is that it crept up on him: "But I think as a result, people got complacent, myself included." Everyone was warning in April and May that oil would ruin the economy, it did not happen, and the warnings stopped. "And really, it's amazing the stock market isn't down more."
5. Is This Peak Earnings?
This is where the two hosts disagree, and they worked it through in the open.
Wiethe framed it against the memory chip trade, where the argument was also about how long abnormal earnings can last. The difference, he said, is that refining has capacity physically shut in rather than merely slow to build.
Farley thinks the sell-side forecasts are too high, and read them off the screen: "So what we had as I said was 8.4. What's forecasted next is 9.4. So I think that makes sense to me. But here's the issue, Max, is that the forecasts are like 6.3 for the next 5.3, 6 billion, 6.2 billion. I'm like in September 2027, you really think they're going to be making 6.2 billion? I think that those estimates, Max, are a tad bit too high."
His argument is that refining is cyclical and mean-reverting. Earnings ramped into the June 2022 oil shock and then declined; the same should happen now. He put the trap in one line: "Like you know, like if the price of oil goes to $300, the bet shouldn't be that it goes to $600."
Wiethe's counter is duration, not direction. He is not arguing the cycle will not turn; he is arguing about when. "I think the refiners are going to be higher" in three or six months, because the assumption behind mean reversion is that the conflict ends this year, and Trump has said it would be over after the midterms more than once.
His scenario is a flat top rather than a spike: "I'm just saying instead of seeing that spike up and then a rapid decline, it's going to be more there's a potential for it to be more of a plateau."
He also named the risk that would break his own moderate case in the bullish direction: refining capacity in the Middle East being destroyed, the way Ukraine has hit Russian refineries, which he said would cause a genuine spike in the refiners.
6. The Hormuz Basket
Farley disclosed his positioning on air, and Wiethe disclosed his after being asked.
Farley's Hormuz basket, published in his Substack model portfolio, holds Marathon Petroleum, a shipping company, and a Canadian producer of roughly 10% of Canadian sulfur — sulfur being what he called the most extreme shortage, though he added, "Maybe diesel is actually more extreme than sulfur now."
Wiethe owns Marathon Petroleum and Valero, and has for years. Farley called it a monster trade and said his only regret was sizing: quoting Harley Bassman, "it's always about sizing."
On corporate governance Farley separated the two industries: "And I'll say this that Marathon Petroleum, MPC, I've looked at they have a very good history of treating their shareholders well of dividends and buybacks."
"And that is in stark contrast to the shipping world which I think their earnings are honestly probably even more on fire than refining if that's even possible." Some Greek ship owners, he said, run a private fleet and a publicly traded fleet out of the same office, which is not what a modern shareholder would expect.
7. Diesel at $6
Farley's longest stretch of the segment was an explainer on why a diesel price matters far beyond transport.
"That's why diesel's at $6 and this is now being, spread into like Wall Street Journal, Bloomberg."
"Diesel fuels rail cars. It fuels trucks." A mining operation uses it to run the mill and the trucks, and a gold mine in rural Africa uses a great deal of it.
"So I think that you know, for commodities, diesel's impact on like all-in sustaining costs is anywhere from like 4% to 30%." He expects that to hit the mining sector hard, though he said miners are extremely profitable on almost every metal except iron and nickel, with gold and copper particularly strong.
The fertilizer chain is the least-watched link. "Like 50% of the world's sulfur I mean comes from the Strait of Hormuz, maybe a little bit more." About half of that sulfur is used to process phosphate rock, much of which is mined in Morocco, into phosphate fertilizer.
Phosphate is the one of the three fertilizer types — nitrogen, potassium, phosphate — that does not have to be applied every year, so skipping a year creates a larger requirement the next one
"So I just think that there are severe imbalances in the commodity markets that have been imposed by the conflict in Iran." He expects $6 diesel itself to mean revert, but not the knock-on effects.
Wiethe pushed back on where that reasoning ends up. The implication of listing every industry that depends on diesel is demand destruction, and he said it is not visible at these prices. "So, this is by no means a call for $10 diesel or anything like that." His position: "And if we continue to stay in this range, they're going to continue to print. And I don't really see an end to the conflict at this point in time."
8. Lam's Criminal Estimates
Farley said he holds a low-conviction view that Marathon Petroleum's estimates are too high and a high-conviction view that one semiconductor equipment maker's are too low.
The company most recently reported what he called a bombshell quarter and guided to a large step up for the quarter now in progress. What he objects to is the growth rate forecast after that, which he called anemic.
"And then 8.1 billion September 2026 into September 27, it's at 9.6." That is roughly 17% growth, which he said is the company's average growth rate rather than a boom rate.
His own number is higher: "I would be at 10.5 billion, not 9.6. I might even go to 11." He said he would be the highest forecast on the street.
The reason is who the buyers are. Logic fabs such as TSMC and Intel buy the machines, but 30% to 40% of the demand — and rising — is memory: SK Hynix, Samsung and Micron, which he described as the biggest bull market in history. Those producers will do everything they can to add capacity.
"I think that these estimates are criminally too low."
9. Rates and the AI Buildout
Wiethe tied the market discussion back to the Fed segment by way of an argument he had heard on the long bond.
The Atlanta Fed's nowcast came out the same day: "It's 5.1% GDP," and Wiethe confirmed to Farley that the figure is real, not nominal, which implies a much higher nominal rate once inflation is added.
Farley's read is that a 30-year yield near 5% is not obviously cheap against that growth. His test for an obvious buy: "Like if they were at 7% and growth was at 4%, they would be an obvious buy." On the wider version: "If the overnight rate was at 3% and the 30-year was at 7% and that's like a 4 percent spread, that would be historically very wide and like a clear pound the table buy, even though all the articles would be how it's a crisis." He shares Joseph Wang's bull case on bonds but is less fervent about it.
Wiethe relayed the convexity argument: the 30-year is fairly priced, so the cheapest way to bet on long rates falling is not the bond. "AI stocks are one of the best ways to play it because the capex is being funded by borrowing at this point."
Farley disagreed and located the rate exposure at the other end of the curve: "I don't think that there's that much of a connection because actually I think most of the long-term financing for AI capex on the long end is the hyperscalers who actually I don't think really are that sensitive to the rates." A few hundred million or a few billion of interest expense does not change what Meta or Microsoft does.
"I think actually that the sensitivity is greater on the short end" — the speculative borrowers, CoreWeave and the private-credit neoclouds, fund in the overnight market. That gave him pause about whether today's hike could crimp the build-out.
Wiethe agreed and generalized it. Saying the long end is all that matters has become a fashionable line, and it is true for a homeowner, but business finance is almost entirely short-end: bank borrowing, commercial real estate at about five years, and private credit at five to seven years but floating rate.
10. Small Caps Take the Hit
Farley's closing thoughts moved off commodities entirely.
"Yeah, I think this is going to put pressure on small caps. People have been super bullish on small caps, but those are the most indebted firms."
He does not think the hike changes the fundamentals of the AI trade much. What he thinks is more consequential is regulation and the labs themselves: "I think far more consequential is we haven't talked about it, but the potential regulation and the fact that these labs are saying we've got to slow down."
He read that both ways. A company saying its own technology is too powerful to release quickly is a risk, and it is also an argument that the technology works — against the bear case that the algorithms do not work and the spending has no return.
Bonus Insights
Wiethe closed by warning that the reaction to a Fed meeting does not finish inside the first hour or two: "We the volatility generally does not end itself just in the one or two hours post Fed meeting." The next session could continue the move or reverse it, and neither of them would guess which.
Farley discounted the headline growth figure he had just cited. Adding roughly 3% inflation to a 5.1% real nowcast gives a nominal rate around 8%, which he said will not actually be the outcome and is a quirk of the particular quarter now ending.
A viewer named John Coch wrote in to say good morning, which produced an exchange about whether the correct abbreviation after noon is GM or GPM. Farley's answer was that GM is good morning and GPM is good afternoon.
The two hosts end in different places on the same trade: Farley thinks refining earnings are cyclical and the forecasts have not priced the decline, while Wiethe thinks the shut-in capacity behind the Strait of Hormuz holds the margin up long enough to make a plateau the base case — and both of them own it.
Products, Companies & Tools Mentioned
Marathon Petroleum (The refiner both hosts own: $6.6B from the refining segment in a quarter, close to the June 2022 peak, and the centerpiece of Farley's Hormuz basket)
Valero (The second refiner Wiethe has owned for years)
Lam Research (Farley's high-conviction call: he would forecast $10.5B to $11B against a consensus $9.6B, because memory buyers are a growing share of demand)
Micron, Samsung and SK Hynix (The memory producers Farley says are 30% to 40% of the equipment demand and rising)
CoreWeave (Farley's example of the speculative AI borrowers funded at the short end, where he thinks a rate rise actually bites)
Meta and Microsoft (The hyperscalers he says will build regardless of whether the 30-year is at 4.5% or 5.2%)
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: