2026年的美股，赚得我有点心慌

"If you make money in A-shares, it proves you have strength, financial luck, boldness, skill, vision, insight, and patience.

If you make money in U.S. stocks, it only proves you have money deposited there."

That's the situation most people trading U.S. stocks find themselves in around 2026.

Those who quietly bought U.S. memory stocks, casually deleted the app, and went on with their lives, only to log back in one day and find their accounts had multiplied several times over.

Driven by U.S. stocks, memory stocks in A-shares have also started to take off.

People in the crypto space have also shifted from discussing memes and altcoins to talking about U.S. stocks: "I live in daily fear of U.S. stocks going up and BTC going down."

New account holders ask the ultimate question in group chats: Why is making money in U.S. stocks so easy?

1. What's Actually Driving the U.S. Stock Market Up?

The main capital narrative in global markets for 2026 is clearly, memory.

Justin Sun was among the first to call out the memory sector towards the end of 2025.

Netizens calculated that if you had bought U.S. memory concept stocks when Justin Sun made his call:

If you bought Micron, you'd be up +222%; if you bought Seagate, you'd be up +256%; if you bought Western Digital, up +280%; if you bought SanDisk, up +515%.

If you had invested 500,000 RMB in SanDisk stock this time last year, you would now have 15 million RMB.

So what exactly is memory?

Memory chips are the components in computers and phones responsible for storing information. There are two main types: DRAM handles short-term memory, temporarily storing data while programs are running; NAND handles long-term memory, holding your photos and files. When you choose between a 128GB or 256GB phone, that capacity refers to NAND.

There are fewer than five manufacturers globally capable of producing both types.

The stocks of these five companies have performed as follows over the past year:

SanDisk, spun off from Western Digital in February 2025, an older company known for USB drives and SSDs, saw its stock price surge up to 22 times its value.

Micron, a cyclical stock that fund managers avoided for a decade, gained over 550% in one year, with gross margins rising from 18% to 56%. Apple's gross margin is around 43%, widely considered a high profit level in tech, yet Micron is now even higher than Apple.

SK Hynix is up 123% this year. Samsung is up 94%.

Seagate and Western Digital are both at all-time highs.

Then there's South Korea.

Samsung and SK Hynix together account for over 30% of the weighting of South Korea's KOSPI index. In 2025, they helped drive the entire South Korean stock market up 76%, securing the top spot among major global indices for the year.

The stellar performance of these two memory manufacturers lifted an entire country's stock market.

On the pricing side, the impact is even more direct. DDR4 memory chips, priced at $1.45 in early 2025, surged to a high of $17 in February 2026, an increase of nearly 12 times in one year. A Kingston 16GB RAM stick in Huaqiangbei went from 200 RMB to 800 RMB. If you've noticed your phone or computer getting more expensive recently, part of the reason lies in these stocks you didn't buy.

SK Hynix's net profit for Q1 2026 soared by 398%, with an operating profit margin of 72%. Samsung Electronics' overall operating profit surged 755% year-on-year.

Selling 100 RMB worth of memory yields 72 RMB in profit and costs 28 RMB. This is no longer a business; it's like owning a mine.

2. Institutions Have Lost Their Cool More Than Retail Investors

In a typical market, institutions are the ones in suits saying, "We are bullish on fundamentals for the long term," while retail traders shout "Buy buy buy!" in group chats.

In the 2025-2026 memory sector, it was the institutions that went crazy first.

Google, Microsoft, and Amazon started placing open-ended orders with Micron: "no price limit, no quantity limit."

The phrase "no price limit" deserves consideration. It essentially means, "Name your price, and we'll accept it, no negotiation." This type of procurement is typical in wartime governments buying munitions.

For 2025-2026, it appeared in tech companies buying RAM.

Broadcom locked in its supply until 2028.

SK Hynix stated at an investor conference: "HBM production capacity for 2026 is completely sold out."

All of it. The entire year.

HBM is high-end memory specifically designed for use with AI chips. For every AI chip Nvidia sells, it needs an accompanying HBM module. Globally, only SK Hynix, Samsung, and Micron can produce HBM, with SK Hynix holding approximately 57% of the market. "Completely sold out" means that one of the most critical components for global AI infrastructure has no surplus for the entire year of 2026.

Then came the analysts.

Within three months, the Wall Street consensus estimate for SanDisk's 2026 earnings per share was raised by 172%. Citigroup predicts the average selling price for server DRAM in 2026 will rise 144% year-on-year. Nomura says the super cycle will last at least until 2027, with meaningful supply increases not expected until early 2028. After stocks had already risen by hundreds of percentage points, Melius Research upgraded Micron to Buy, adding without a flinch that "there is still 41% upside over the next 12 months."

DeepMind CEO Hassabis publicly stated that the overall constrained memory supply chain is limiting significant AI deployment. Intel CEO Lip-Bu Tan said the memory shortage won't ease before 2028.

Then, SK Hynix secretly filed an application with the SEC to issue an ADR in the U.S., aiming to raise up to $15 billion. A company with all its capacity sold out and a 72% profit margin decided to go raise more money in New York, reasoning that the South Korean market undervalues it, while U.S. investors better understand AI and are willing to pay a higher price.

A-shares followed suit.

De Ming Li hit the daily limit up, Biwin Storage surged, Longsys rose 41%, and smaller player Xiangnong Xinchuan saw its Q1 net profit pre-announce a 6714% to 8747% increase—a four-digit growth rate. Topics in finance groups shifted from "Should I still buy CSI 300?" to "Which one should I buy, Micron or SK Hynix?" People who hadn't known how to spell HBM two months ago were now explaining high-bandwidth memory principles in group chats.

Even many matchmaking groups are discussing memory stocks.

3. The Most Ironic Scene

On February 24, 2026, Citron Research announced its short position on SanDisk, giving three reasons.

First, memory is a cyclical stock. In 2008, 2012, and 2018, every period of high profitability ended with a crash. Current production capacity is already double the previous peak in 2018; supply release is just a matter of time.

Second, SanDisk sells a commodity.

"Nvidia has a moat; SanDisk just sells a commodity." Nvidia's moat is its CUDA software ecosystem, on which almost all global AI models run, making switching costs extremely high.

SanDisk's SSDs could be replicated identically by Samsung tomorrow, possibly at a lower cost.

Third, major shareholder Western Digital is heavily selling its SanDisk shares at a 25% discount to the market price.

Selling its own stock at a 25% discount. One possibility is it urgently needs cash. Another is that it believes the stock will be cheaper later. Neither scenario suggests a bullish outlook.

Two trading days later, SanDisk rebounded and continued to hit new all-time highs. Citron's report circulated in various finance group chats, becoming meme material.

One question was skipped by everyone: whose accounts did those shares sold at a 25% discount ultimately end up in?

4. Is Making Money in U.S. Stocks as Easy as Breathing?

The world's three most profitable memory companies collectively chose not to expand production at the peak of their profits.

SK Hynix's capital expenditure related to HBM in 2025 decreased by 50% year-on-year, with the official explanation being concerns about oversupply in 2027. Samsung's DRAM capacity growth in 2026 is only about 5%, far below the demand growth rate.

The entire industry's capital expenditure growth rate is only 14%, whereas in historical boom periods, it typically ranges from 30% to 50%.

These three companies control 92% of global DRAM capacity. Their collective decision not to expand production has a name in any other commodity market: supply-side coordination. OPEC did this with oil, resulting in the 1973 oil crisis. The memory chip market is more concentrated than OPEC; the combined market share of these three companies exceeds that of OPEC's thirteen member nations.

Investors interpreting "manufacturers exercising capacity discipline" as positive news is logically correct—prices can indeed stay higher for longer. However, what this structure means for the buyers on the other end is mentioned in none of the analyst reports.

Two equally plausible narratives can explain this bull run.

First: AI's demand for memory represents a structural shift. AI models in the inference era need to remember increasingly long contexts, requiring a magnitude leap in memory. The three major manufacturers control 92% of capacity, and new fabrication plants can't come online until at least 2027. The gap won't disappear before then.

Second: It's the same as every other historical cycle. The narrative for the 2000 dot-com bubble was "the internet changes everything," which was true. The narrative for the 2008 subprime crisis was "housing prices can't fall nationally," which was also supported by historical data at the time. The real issue has never been whether the narrative is correct, but whether prices have already fully priced in the story.

The memory industry has an iron rule unbroken for 30 years: Prices rise slowly, but they fall very fast.

During the super cycle in 2018, from peak to halving took less than two quarters.

No one knows when this cycle's peak will be. Not even those selling their shares at a 25% discount—or perhaps, especially those selling at a discount, because what they are selling are chips, and the most efficient way to sell chips is to people who are currently buying into the story.

Last time you bought a phone, you paid a few hundred RMB extra to upgrade from 128GB to 256GB of storage. That few hundred RMB travels up the industrial chain, getting split at each layer, with a small portion eventually showing up in SK Hynix's 72% operating margin, in Samsung's 755% profit growth, and in all those stocks you didn't buy.

And of course, it finally culminates when you open any social media app and see everyone else's existential question: Why is making money in U.S. stocks so easy?