The U.S. stock market in 2026 makes me a little nervous about my earnings.

“Making money in A-shares proves you have strength, good luck, courage, skill, vision, awareness, and patience.”

“Making money in U.S. stocks only proves you have money parked in U.S. stocks.”

This is the situation for most people trading U.S. stocks in 2026.

Those who quietly bought U.S. storage stocks, casually uninstalled the app, and decided to lie flat—one day they log back in and find their account has risen by several multiples again.

With support from U.S. stocks, storage stocks in A-shares also started to soar.

And people in the crypto circle, who used to chat about memes and copycats, have shifted to talking about U.S. stocks: “I live every day in fear—U.S. stocks are rising, while BTC is falling.”

Newcomers to the market post in group chats, asking their soul-searching question: Why is it so easy to make money in U.S. stocks?

I. Who, exactly, is driving the rise in U.S. stocks?

In 2026, the main capital storyline in global markets is clearly storage.

Sun Yuchen was the first to call out the storage sector at the end of 2025.

Netizens calculated that assuming you bought U.S. storage-related stocks at the moment Sun Yuchen issued his buy call:

If you bought Micron, you made +222%; if you bought Seagate, +256%; if you bought Western Digital, +280%; if you bought SanDisk, +515%.

If you bought SanDisk stock for 500,000 yuan today last year, you would now have 15 million yuan.

So what exactly is storage?

Storage chips are the parts inside computers and phones that are responsible for remembering things. They come in two types: DRAM handles short-term memory and is used to temporarily store data when programs are running; NAND handles long-term memory, where photos and files on your phone are stored. When you choose a phone with 128G or 256G, that capacity is NAND.

Globally, there are fewer than five manufacturers that can produce both of these.

Over the past year, the stocks of these five companies have risen like this:

SanDisk: Split from Western Digital in February 2025. It’s an old company that makes USB drives and solid-state drives. At its peak, the stock price rose by 22 times.

Micron: A cyclical stock that fund managers disliked for ten years. In a single year it surged over 550%, and its gross margin climbed from 18% to 56%. Apple’s gross margin is about 43%—already a widely recognized “ultra-high-profit” level in tech. Micron’s now even higher than Apple’s.

SK Hynix: Up 123% this year. Samsung: Up 94%.

Seagate and Western Digital both hit historic highs.

Then there is Korea.

Together, Samsung and SK Hynix account for more than 30% of the weight in Korea’s KOSPI index. In 2025, they lifted the entire Korean stock market by 76%, winning the annual championship among major global indices.

When the two memory manufacturers’ performance exploded, the stock market of an entire country flew along with it.

The pricing side is even more direct. DDR4 memory chips: in early 2025, they were $1.45 each. By February 2026, they hit a high of $17—nearly 12 times in a year. Kingston 16G memory sticks from Huaqiangbei went from 200 yuan to 800 yuan. If your phone or computer has been getting more expensive lately, part of the reason is inside these few stocks you didn’t buy.

SK Hynix’s net profit in Q1 2026 surged 398%, and its operating profit margin was 72%. Samsung Electronics’ overall operating profit jumped 755% year over year.

If you sell memory worth 100 yuan, 72 yuan is profit and 28 yuan is cost. This isn’t doing business anymore—it’s like mining.

II. Institutions are more irrational than retail investors

In a typical market, institutions are the ones in suits, expressionless, saying, “We are bullish on the fundamentals in the long term.” Retail investors are the ones in group chats shouting “Go, go, go.”

From 2025 to 2026, the storage sector was the first to go crazy among institutions.

Google, Microsoft, and Amazon started placing open-ended orders with Micron—“no price limits, no quantity limits.”

The three words “no price limits” are worth thinking about. It means: you quote whatever price you want, and they pay whatever amount—no haggling. This kind of procurement approach usually shows up in wartime scenes when governments buy arms.

In 2025 and 2026, it appeared on the purchasing of memory sticks by tech companies.

Broadcom locked supply for three years, extending to 2028.

At an investor conference, SK Hynix said, “By 2026, all HBM production capacity has already been sold out.”

All of it. For the entire year.

HBM is high-end memory designed specifically to be used with AI chips. Every time Nvidia sells an AI chip, a matching HBM is needed alongside it. Globally, only SK Hynix, Samsung, and Micron can make HBM, and SK Hynix accounts for about 57% of the market. The meaning of “sold out” is that one of the most critical components in building global AI infrastructure will have no excess in 2026 for the whole year.

Then there are the analysts.

Within three months, Wall Street’s consensus estimate for SanDisk’s 2026 earnings per share was raised by 172%. Citigroup projected that the average server DRAM price in 2026 would rise 144% year over year. Nomura said the super cycle would last at least until 2027, and that meaningful increases in supply would come no earlier than 2028. Melius upgraded Micron to a buy after the stock had already risen by several hundred percentage points, and added, “There’s still 41% upside in the next 12 months,” without even blushing or breathing hard.

DeepMind CEO Hassabis publicly said the overall memory supply chain is constrained, which is limiting large-scale AI deployments. Intel CEO Chen Lihwu said the memory shortage won’t ease before 2028.

Next, SK Hynix secretly submitted an application to the SEC to issue ADRs in the U.S., raising up to $15 billion. A company whose capacity is fully sold out and whose profit margin is 72% decided to raise more money in New York. The reason: the valuation offered by the Korean market is too low, and U.S. investors understand AI better and are willing to pay a higher price.

A-shares followed suit.

DeMingli hit the daily limit; Bawei Storage surged sharply; Jiangbolong rose 41%. For small manufacturers like Shanyong Chipchuang, net profit in Q1 was projected to increase 6714% to 8747%—four-digit percentage gains. The topic in financial group chats shifted from “Can the CSI 300 still be bought?” to “Which should we buy: Micron or SK Hynix?” Two months earlier, people who didn’t even know how HBM was put together started explaining in group chats the working principles of high-bandwidth memory.

Even many dating groups discussed storage stocks.

III. The most ironic scene

On February 24, 2026, Citron Research announced a short position in SanDisk, providing three lines of logic.

First, storage is a cyclical stock. In 2008, 2012, and 2018, every time profitability hit a high, it ended with a crash. Current capacity is already twice the peak level of 2018, and releasing supply is only a matter of time.

Second, SanDisk sells commodities.

“Nvidia has a moat; SanDisk is just a commodity.” Nvidia’s moat is the CUDA software ecosystem. Almost all AI models run on it, and replacing it is extremely expensive.

As for SanDisk’s solid-state drives, Samsung can build essentially the same thing tomorrow—possibly even cheaper.

Third, major shareholder Western Digital and others are unloading SanDisk shares on a massive scale at discounts below the market price of 25%.

Selling your own shares at 75 cents on the dollar. One possibility is that they need money urgently. Another possibility is that they believe it will be cheaper later. In both cases, there is no such thing as being confident about the outlook.

Two trading days later, SanDisk rebounded and continued to set historic highs. Citron’s report spread across various financial group chats and became meme material.

One question everyone skipped was this: those shares sold at a 75% discount—where did they finally end up?

IV. Is making money in U.S. stocks as easy as breathing?

The three most profitable storage companies in the world, at the moment when profits were at their highest, all collectively chose not to expand production capacity.

In 2025, SK Hynix’s HBM-related capital expenditures fell 50% year over year. The official explanation was concern about oversupply in 2027. Samsung’s DRAM production capacity growth in 2026 is only about 5%, far below the pace of demand growth.

Total industry capital expenditure growth was only 14%, whereas in history, during every boom expansion period it was usually 30% to 50%.

These three companies control 92% of global DRAM production capacity, yet they all chose not to expand. In any other commodities market, there’s already a name for this: supply-side coordination. OPEC has done this with oil, resulting in the 1973 oil crisis. The concentration in the storage-chip market is even higher than OPEC’s. The combined market share of the three companies is something that 13 oil-producing countries cannot reach.

Investors interpret “manufacturers exercising restraint in expanding capacity” as a positive. Logically, prices can be supported for longer. But what this structure means for the buyers on the other end is not covered in any analyst report.

This rally can be explained by two stories, both defensible.

First: AI demand for storage is undergoing a structural change. In the reasoning era, AI models need to remember increasingly longer context, and the memory required jumps by orders of magnitude. The three major storage manufacturers control 92% of production capacity. New factories can only start production as early as 2027. Until then, the gap won’t disappear.

Second: It’s the same as every time in the past. The narrative during the 2000 internet bubble was “the internet changes everything”—and that was true. The 2008 subprime narrative was “housing prices won’t fall nationwide,” and even that made sense based on the historical data at the time. The real problem has never been whether the story is right; it’s whether prices have already front-loaded (overextended) the story.

The storage industry has an iron law that has held for 30 years: prices rise slowly, while prices fall quickly.

In the 2018 super cycle, from the top to a 50% haircut, it took less than two quarters.

Nobody knows on which day the top will be this time. This includes those selling at 75% of the price—or more precisely, especially those selling at 75% of the price—because when you sell “chips,” you’re selling the chips themselves. And selling chips to people who believe the story is the most efficient way to make money.

The last time you bought a phone and upgraded the memory from 128G to 256G, you paid an extra 300 to 400 yuan. That extra few hundred yuan, distributed through an industry chain where profits are taken layer by layer, ultimately ends up as part of SK Hynix’s 72% operating profit margin, part of Samsung’s 755% profit growth rate, and part of all the stocks you didn’t buy.

And of course it all finally funnels into the moment when you open every social app and see someone else asking their soul-searching question: Why is it so easy to make money in U.S. stocks?