A Hundred Trillion Won and a Shrug

Prepared remarks to an investment committee, Thursday morning. Transcript lightly edited.

Good morning. Put the Samsung number on the screen, please, and leave it there while I talk.

107.4 trillion won. Roughly $80 billion of operating profit in a single quarter, up 782.5% from a year ago, on revenue of 195 trillion. The fourth straight record. The first time any technology company has guided to a quarterly operating profit above 100 trillion won. This one quarter is around two and a half times what Samsung earned in all of last year, about 44 trillion, and the nine-month total sits near 254 trillion.

Consensus at LSEG was 106.1 trillion. It beat. And the stock fell, by anywhere from a fraction of a percent to 1.6% depending on which hour you checked, dragging the KOSPI down roughly 1.5% on top of Wednesday's 1.98% drop to 6,803.90.

I want to spend our time on the shrug, not the number. The number is a fact. The shrug is information.

First, the beat was cosmetic. Estimates had been trimmed 7.7% since the end of August, so the bar came down to meet the print. Against Bloomberg's 108.67 trillion the result was a miss, and revenue of 195 trillion landed well under the roughly 207 trillion FactSet had penciled in. A beat achieved by lowering the bar has the nutritional value of a participation trophy.

Second, the price had already moved on without the earnings. Samsung changed hands near 268,000 won this morning, about 28% under its 52-week high of 374,500. Memory stocks have a long and ugly habit of peaking before the profits do. I remember a record Samsung quarter in 2018, something like 17.6 trillion won, that arrived many months after the shares had topped out. Nobody celebrated that one either. The market prices the second derivative, the rate of change of the rate of change, and operating profit growing 20% sequentially after a quarter that grew roughly 56% is a deceleration wearing a crown.

Third, and this is the one I would put in the minutes, a record quarter is a receipt. Every won of that profit is somebody else's capital spending. Samsung is the seller, and its income statement is the mirror image of a buyer's cash outflow. So the useful question is about the buyer. A Heatmap News report circulating this morning counts at least $260 billion of data-center investment canceled in 2026, with the pace of cancellations picking up through the September quarter amid funding and energy-supply disruptions. Set that next to yesterday's reports that SpaceX wants $40 billion to buy Nvidia chips, $10 billion in bank loans and $30 billion in investment-grade bonds, and you can see the shape of it. The orders are still being written. The question is who is signing them, and at what cost of money.

That cost is not a footnote. The ten-year Treasury touched 5.365% yesterday, the highest since April 2002. The thirty-year touched 5.732%. The Fed minutes say a majority of participants expect one more hike before year-end. A hyperscaler can fund a data center at those rates, but the project has to clear a hurdle that did not exist when the purchase orders for this quarter's memory were first drafted. Samsung's customers are signing at a price that Samsung's profit margin, north of 50% on these numbers, is quietly inviting them to renegotiate.

Fourth, the room. Foreign investors sold Korean shares across the board on Wednesday, and local reports described domestic liquidity as drier than in the first half of the year. The Korea Times noted that the won's recent strength is a margin worry for the exporter in chief. The Nikkei slid back below 70,000 this morning with the yen near 157.9 and Nomura flagging that AI and chip names could stay under pressure. When three or four independent pockets of the market decline the same good news, you are looking at a regime and should stop treating it as a mood.

Now, in fairness, because you pay me to be fair. The bulls have a case and it is not stupid. Samsung itself has said AI spending keeps memory supply tight through the second half, and HBM4 volume is ramping. Micron rose 4% on Wednesday after one broker argued the shortage could last two more years and the stock could triple. Shortages are real, and shortages are the thing that produces 55% operating margins in a business that historically earned a fraction of that. A tight market can stay tight longer than a skeptic can stay solvent.

But tight markets do not need a collapse in demand to turn. They need a slowdown in the growth of the marginal buyer, and the marginal buyer is exactly the one now paying 5.3% for the privilege. The existing capacity gets filled. It is the next fab and the next hyperscaler campus that carry the equity story, and those depend on funding that is getting dearer by the week.

So here is how I would read this morning. The print confirms the past tense of the cycle with great accuracy. It tells us nothing new about the future tense, and the market, which trades the future tense, has said so in the only language it speaks. Eighty billion dollars in a quarter and a down tape. If you want one data point that the discount rate has started to outweigh the earnings, take that one.

Over the next ten days I will be watching three things: whether further Asian memory names trade like Samsung, whether the AI debt pipeline clears without a spread concession, and what the first big American earnings reports say about the cancellations. We do not need to be early to the turn. We need to avoid being the last buyer of a receipt.

Thank you. Questions.