Samsung forecast $80 billion in profit. Its stock fell 43%.

I spent too much on a laptop last month.

Not on something fancy. A basic work machine. The same model I bought in 2024 for $899 now costs $1,140. The sales guy said it was the memory. RAM prices, he said. Something about AI.

He was right. A 32-gigabyte DDR5 memory kit that cost under $90 in early 2025 now sells for $409 to $529. J.P. Morgan estimates DRAM chip prices will have risen more than 400% from the start of 2024 to the end of 2026. Dell, Lenovo, HP, Acer, and Asus have all warned clients to expect 15% to 20% price increases on PCs this year.

The reason is simple. AI data centers need a special kind of memory called HBM—high-bandwidth memory. Samsung, SK Hynix, and Micron make it. They also make the regular DRAM that goes in your laptop. And they have been shifting factory capacity from one to the other.

So consumer memory got scarce. Prices went up. The companies got rich.

Samsung just forecast its best quarter in history—107.4 trillion won, roughly $80 billion, in operating profit. Micron posted fiscal-year revenue of $133 billion, up 256% from the year before, with $85 billion in net income. SK Hynix says its HBM chips are sold out through 2026.

But here is the part I keep coming back to.

Samsung stock is down 43% from its June high. SK Hynix is down 57%. Micron fell hard in October too. The Philadelphia Semiconductor Index dropped for eight straight sessions through October 16—its longest losing streak since October 2008, the month after Lehman fell.

Memory stocks as a group—Samsung, Micron, SK Hynix, SanDisk—have lost more than $700 billion in market value since October 6 alone. A broader basket of 61 semiconductor stocks has shed roughly $2.5 trillion.

Record profits. Record demand. Record losses in the stock.

I have no idea where the bottom is. But the divergence between what these companies earn and what the market pays for them is the widest I have seen in this cycle. SK Hynix trades at less than four times next year's expected earnings. Samsung fell 2.4% on the same day it announced that $80 billion profit estimate.

Something is off. The market sees a risk these companies do not talk about. And I think I know what it is.

When the world’s biggest memory-chip makers sell off, the damage doesn’t stay in the chip sector. It spreads to the country that depends on them.

The KOSPI, South Korea’s benchmark index, hit an all-time high of 9,385 on June 19. Samsung and SK Hynix together make up roughly half of the index’s market capitalization. By October 26, the KOSPI sat near 5,500. That is a 41% decline in four months.

Foreign investors have been leaving. In September alone, they pulled a net $14.16 billion out of Korean stocks—more than three times August’s $4.5 billion outflow. It was the eighth straight month of net selling across stocks and bonds.

The KOSPI fell more than 3% on October 12 alone, led by Samsung and SK Hynix. On October 13, Samsung fell 7.4% in a single session.

One country bet its economy on memory chips. The market just marked the Asian semiconductor sector down by $1.5 trillion.

Korean semiconductor exports, meanwhile, hit $27.5 billion in the first twenty days of October—an all-time record. The country’s total exports for that same period more than doubled from a year ago, to $61.6 billion.

So exports are at records. Profits are at records. And the stock market is down 41% from its peak. We are watching a country with the best chip earnings in its history get priced like the earnings are about to end.

And here is where the fear starts to make sense.

ChangXin Memory Technologies—CXMT—is a Chinese chipmaker based in Hefei. A year ago it held 4% of the global DRAM market by revenue. By the second quarter of 2026, Counterpoint put it at 10%. TrendForce reported its Q2 revenue share at 9.5%.

CXMT went public on Shanghai’s STAR Market on July 27. It raised $8.6 billion—Asia’s largest IPO this year. On its first day of trading, the stock rose 466%. It briefly became China’s most valuable listed company, worth about $488 billion, more than ICBC.

CXMT does not make HBM yet. It makes the older-generation DRAM that Samsung, SK Hynix, and Micron have been deprioritizing as they shift capacity to AI chips. In other words, CXMT moved into the gap the big three left behind.

That gap is the threat. If CXMT keeps taking share in commodity DRAM—and it has access to cheaper labor, state subsidies, and a massive domestic market—the big three lose pricing power in the half of the market that is not AI.

Analysts at TrendForce expect CXMT to push into HBM production over the next two years. It cannot access EUV lithography equipment due to export controls. But it is spending $4.1 billion from the IPO on technology upgrades.

Meanwhile, the memory shortage has a legal problem.

On June 25, seventeen plaintiffs filed a class-action lawsuit in the U.S. District Court for the Northern District of California against Samsung, SK Hynix, and Micron. The claim: the three companies coordinated a shift in production from commodity DRAM to HBM, restricting supply to the consumer market and driving prices up roughly 700% over four years.

The lawsuit calls it a deliberate choice, not market forces. The three companies control roughly 87.6% of global DRAM revenue—Samsung at 39.4%, SK Hynix at 24.9%, Micron at 23.3%, per TrendForce Q2 data. The plaintiffs argue that concentration makes coordination possible.

I do not know whether the lawsuit will succeed. These cases take years. But I know this: Samsung, Hynix, and Infineon paid a combined $645 million in criminal fines to the U.S. Department of Justice between 2004 and 2006 for doing something similar. Samsung alone paid $300 million. Fifteen executives pleaded guilty. Several went to prison.

That earlier cartel ran from 1998 to 2002. The companies fixed prices on DRAM sold to Dell, HP, and other PC makers. Micron received immunity for reporting the conspiracy to the DOJ.

The new lawsuit does not allege secret meetings. It alleges something more subtle—a coordinated production shift that achieved the same result. Whether that holds up in court is a separate question from whether the market believes the earnings are sustainable.

That is what matters here. Not whether the suit wins. But whether the 400% price increase was a market outcome or a market structure. And whether it lasts.

Three more things worth keeping track of this week.

Subprime auto loans keep getting worse. Fitch reported that 60-plus-day delinquencies on subprime auto loans hit 6.90% in January 2026, the highest rate in 32 years. Total auto loan balances reached a record $1.71 trillion in the second quarter, per the New York Fed. The average new car loan now carries a monthly payment north of $730 with terms stretching to nearly six years. The borrowers most likely to fall behind are the same ones paying 22% credit-card rates. This is not a systemic risk yet. But it is a growing tab in an economy that already has a consumer-credit problem.

The 10-year Treasury term premium reached 0.98% on October 9, a 12-year high. The term premium is the extra yield investors demand for holding long-dated bonds instead of rolling short-term bills. When it rises, it means the bond market wants more compensation for uncertainty—about deficits, about inflation, about who will buy all the debt. At 0.98%, it is higher than any point during the 2022–2023 rate-hiking cycle. It helps explain why stocks fell hard in early October even though corporate earnings were strong.

Global container freight rates rose to a 15-week high. The Drewry World Container Index climbed 2% to $4,585 per 40-foot container as of October 22, led by the Asia-to-Europe route. Rates have been rising since mid-October after falling for thirteen straight weeks. European buyers are competing more aggressively for U.S. LNG shipments ahead of winter, and the Red Sea disruptions continue to reroute traffic around the Cape. Higher freight adds to goods inflation with a three-to-six month lag. We’ll see.

I want to end this week with a story about memory.

Not the kind in your computer. The kind in a courtroom.

In 1998, executives at Samsung, Hynix, Infineon, and Micron began meeting—sometimes at trade conferences, sometimes at private dinners—to agree on the prices of DRAM chips sold to Dell, HP, and other PC makers. The conspiracy ran for four years. It worked. Prices stayed high. Consumers paid more for their computers and never knew why.

In June 2002, Micron went to the Department of Justice and reported the cartel. Micron received full immunity. The DOJ spent the next five years dismantling it.

Samsung paid $300 million. Hynix paid $185 million. Infineon paid $160 million. Fifteen executives pleaded guilty. Several went to prison.

The European Commission fined the group another €331 million in 2010. Thirty-three U.S. state attorneys general won a $173 million settlement. Young Hwan Park, the president of Samsung’s U.S. semiconductor subsidiary, pleaded guilty and served time in a federal prison. It was the second-largest criminal antitrust fine in American history at the time.

Samsung, Hynix, and Micron paid the fines, served the sentences, and went back to work. Together they still control 87.6% of the global DRAM market in 2026.

On June 25, a new class-action lawsuit was filed in the same Northern District of California. Seventeen plaintiffs allege the same three companies coordinated again—not with secret dinners this time, but by simultaneously shifting production from commodity DRAM to AI memory, restricting supply to the consumer market, and driving prices up roughly 700% in four years.

The lawsuit may not succeed. Proving coordination without evidence of direct communication is much harder. But the structure is familiar. Three companies, one product, a restricted supply, and a 700% price increase.

The fines in the original case came to more than a billion dollars. The executives did their time. And twenty years later, the same market, the same players, the same accusation.

Maybe the market learned something from that story. Maybe that is why the stocks keep falling even as profits keep rising. A 700% price increase built on a three-company oligopoly is exactly the kind of profit the market does not trust to last.

We’ll see.