阿修AX

Let's not talk too much about ZhiPu. Nearly 70% of cornerstone investors haven't left, confirming the previous judgment.

Focus on storage. SanDisk fell sharply today, making many people panic. What's the view? There's only one core reason—Samsung's stellar earnings, which ironically became a signal to sell. It's a case of killing leverage after a big rise, not a fundamental collapse.

The panic is excessive. The supply-demand gap will last at least until 2027. The logic hasn't changed; a drop is actually an opportunity.

$KNOWLEDGE ATLAS(02513.HK)$Sandisk(SNDK.US)

Zhipu will have 25.6816 million shares (5.76% of the total) unlocked on July 8th. The unlock time, quantity, and target were announced half a year ago — it's something the whole world knows, so why do you think only you know about it?

When everyone is anxious about the same thing, it has long been priced in. Unlock ≠ sell-off, the premise is that they are willing to sell. Zhipu's cornerstone investors include a large number with state-owned backgrounds. They can slowly offload shares through block trades or over-the-counter agreements before the unlock date, with absolutely no need to dump shares on the unlock day itself.

The biggest risk in the market has never been the "widely known" negative news, but the "you think you know" expectation gap. There is no expectation gap regarding this unlock event.

Moreover, you need to think clearly about one thing: the unlock changes the shareholding structure, not the company's value.

Zhipu's models are iterating, and commercialization is advancing. These are the things you should really be focusing on. The unlock is just a gust of wind; once it passes, the ship sails on.

Most importantly, if one day, when Zhipu is competing with other large models, its development path goes completely wrong and it falls far behind, that's when it should be sold.

$KNOWLEDGE ATLAS(02513.HK)

I mentioned the need to reduce positions in the current market yesterday, and I believe some friends have already taken profits. Given how the market has moved today, I'll add a few more words. If you think I talk too much and find it annoying, feel free to scroll away.

First, don't make decisions during a sharp drop.

When emotions run high, your actions are most likely to go wrong. You'll regret selling, and you'll regret buying. In fact, doing nothing is the best action at this time—it has the lowest decision cost, requires no consideration, and avoids the consequences of impulse. The crazier the market, the steadier you need to be.

Second, when sudden bad news hits, don't rush to buy the dip.

Many people see bad news and rush in, thinking they're "getting a bargain." But be careful; some negative news can trigger a slow, grinding decline—a little drop today, another little drop tomorrow, like a frog being slowly boiled in water. Don't buy the dip during a sharp drop, and don't buy the dip during a slow decline either. Wait for the situation to become clear. Entering in batches or all at once is more prudent than acting in panic.

Third, when you're not in good condition, don't make any decisions.

This one is just for those friends who trade short-term. When you're physically tired or mentally stressed, your judgment declines. One mistake could wipe out all the gains you've worked so hard to accumulate, or even put you in the red. Decision cost is also a cost. If you're not in the right state, rest first. The market isn't going anywhere.

Ultimately, the most important thing at this stage is not to catch every fluctuation, but to protect your capital and your mindset. Friends holding Micron, SanDisk, and GigaDevice can take the money they've secured and feel at ease; the positions you still hold, feel at peace. Leave the rest to time.

Get some good rest and fight again tomorrow.

$Micron Tech(MU.US)$Sandisk(SNDK.US)$GIGADEVICE(03986.HK)

Some small suggestions for the current market situation.

Regarding storage stocks, such as Micron, SanDisk, GigaDevice, etc. My thought is: if you hold positions, you can reduce a small portion, but it's not recommended to clear them entirely. The fundamentals haven't changed, the supply-demand gap still exists, it's just that the short-term rise has been significant. Taking some profits off the table will make your mindset much more stable. As for how much to reduce, it varies from person to person, whatever feels comfortable.

Regarding CPO (Optical Communication), such as Zhongji Innolight, Xinyisheng, etc. Some wrongly-sold targets have recently started to stabilize. You can use the small change from the reduction to gradually observe and add positions. Once market sentiment recovers, the elasticity is often good.

Regarding large models, Zhipu has been oscillating at high levels recently, but the medium-term logic remains unchanged. I personally prefer to wait for a sharp drop before considering it, rather than chasing highs. You can add positions now, but instead of constantly watching the screen, it's better to set a psychological price limit and wait for it to be reached.

In extreme situations, you can look at defensive allocations. If you want a more stable direction, assets like China Yangtze Power are worth a second look at this stage—low volatility, stable dividends, at least they can make your account less anxious.

Finally, the most practical thing to say:

Waiting is always essential in investing. Having some cash on hand and being able to sleep at night is more important than anything else. Don't buy just for the sake of buying, and don't sell just for the sake of selling. At this stage, less action and more rest might be the best action.

Eat well, sleep well, good health leads to good judgment. The market is always there, opportunities are always there.

$Sandisk(SNDK.US)$Micron Tech(MU.US)$KNOWLEDGE ATLAS(02513.HK)

South Korean stocks experienced a V-shaped rebound today, while GigaDevice's trading volume topped the A-share market.

It's laughable that Apple's price hike is affecting memory stocks, because it only dragged down consumer-grade memory demand, while AI memory demand remains unaffected — Micron's Q3 revenue surged 346%, HBM capacity is booked until 2027, and long-term agreements have locked in $22 billion. As long as users are willing to pay for AI, the AI industry's closed loop remains unchanged.

At this point in time, holding or day trading are both viable, but selling memory stocks now is very unwise. Don't be led astray by the noise.

$Micron Tech(MU.US)$Apple(AAPL.US)$GIGADEVICE(03986.HK)

On June 27, DeepSeek V4 quietly updated a framework called DSpark. The technical term is "an 80% increase in inference speed," which in plain language means—the speed at which the AI replies to your messages is nearly twice as fast.

First, let's explain why it was slow before.

For large language models to generate text, it's essentially a "word-guessing game." For every word the AI writes, it has to re-read and re-calculate all the previous words to decide what the next word should be. Writing 100 words means digesting what it has written 99 times. Academically, this is called "autoregressive generation"—the next word can't move until the previous one is calculated.

So every time you ask the AI a question, it replies word by word, like squeezing toothpaste—it's not intentionally slow; the technical principle dictates it can only work this way.

So how does DSpark solve this?

Here's an analogy to help you understand:

Before, it was like a senior professor personally writing word by word—high quality, but painfully slow.

DSpark's approach is: first, let a fast but average-skilled "assistant" draft, guessing several words ahead in one go, then present them all at once to the senior professor. The professor glances over it, directly keeps the correctly guessed words at the beginning, and starts writing a correct one from the first mistake.

This is called "speculative decoding"—draft first, then review quickly.

But there's a catch here: the faster the assistant guesses, the more likely it is to guess blindly later on. The paper calls this "suffix degradation"—the first word is okay, but by the fifth or sixth, it's basically making things up.

The brilliance of DSpark lies in this: it assigns a "reliability score" to each guessed word, then dynamically decides how many words to send to the senior professor for verification based on how busy the system is. When the system is idle, verify more; when busy, only verify the most reliable batch.

What's the result?

For responses of the same quality, speed is increased by 60% to 85%. A reply that used to take 10 seconds now comes out in five or six seconds. Most crucially—during peak hours, DeepSeek finally stops constantly "spinning."

Why am I quite moved?

The focus of this update isn't that the model got smarter, but that the engineering implementation was done beautifully. Without changing the model itself, it boosted speed purely by optimizing inference efficiency.

Behind this is the result of joint R&D by Peking University and DeepSeek, with the paper personally authored by Liang Wenfeng. Moreover, the entire technology is open-sourced—other AI companies can take it and use it.

Not burning money to pile on parameters, but relying on technology to squeeze out efficiency. This kind of pragmatic effort is worthy of respect.

In investing, the most torturous thing is never "not buying," but rather "could have bought." When a stock you were bullish on rises 50%, doubles, or even goes up tenfold, that voice in your head keeps saying—"if only I had bought it back then."

But I want to tell you a fact: missed trades don't cost you anything.

If you didn't buy, there's no possibility of losing money and no anxiety of being trapped. It goes up on its own, you live your life, and there's no financial relationship between you. What truly costs you is always something else—reckless trading.

Chasing in at highs, panic selling, swapping positions that shouldn't have been touched for assets you don't understand. These are the actions that could wipe you out in one go.

But a crucial premise must be added here: all the above judgments are based on your assets being able to support your daily living expenses for more than three years.

If this money is for next month's rent, your child's tuition, or an emergency fund at home—then any volatility is a matter of life and death for you, and there's simply no talk of "holding calmly" or "waiting for an opportunity." The prerequisite for investing is always using money you can afford to lose, not money you can't.

Don't be upset because you missed out, don't act recklessly because you're itchy. The market never lacks opportunities; what's lacking is you still having capital to sit at the table.

Last night's market gave a very interesting signal: Micron's earnings report was explosive, surging over 15% after hours; Apple announced price increases across the entire Mac and iPad lineup, with its stock plummeting 6%.

But the biggest gainer was neither Micron nor Apple's short positions; it was SanDisk—soaring about 22% in a single day.

Why? The logic isn't actually complicated.

Although Micron is making money, it's being "squeezed from both sides." On one hand, it's indeed benefiting from the AI dividend—Q3 revenue was $41.46 billion, up 346% year-over-year, with a gross margin of 84.9%. But it simultaneously announced its exit from the consumer storage market, no longer doing storage chip business for PCs and phones. In other words, Micron is "abandoning" the big consumer electronics pie, going all-in on AI. And Apple's price hike precisely proves that consumer electronics terminal demand is being crushed by storage costs—which is adding insult to injury for the business Micron just cut.

SanDisk is a "pure AI" beneficiary, benefiting from both sides. SanDisk makes NAND flash and enterprise SSDs, with data centers being its main battlefield. Apple's price hike shows consumer electronics demand is shrinking, but SanDisk doesn't rely on Apple at all—its customers are cloud providers like Amazon, Microsoft, and Google. Q3 revenue was $5.95 billion, skyrocketing 251% year-over-year. The worse consumer electronics get, the more production capacity shifts towards AI, and the fuller SanDisk's order book becomes.

More crucial is the capital flow. The market is making a clear judgment: whoever has pricing power wins. Micron has pricing power but is simultaneously cutting its consumer business; Apple has pricing power but is also passively passing price hikes to consumers; while SanDisk—neither carries the burden of consumer electronics nor stands at the forefront of AI data center demand.

So, Micron's good earnings prove the storage price hike is real; Apple's price hike proves consumer electronics are being squeezed. These two things combined are most beneficial precisely for that "neither reliant on consumer electronics, yet fully embracing AI" SanDisk.

This is probably the answer the market gave with a 22% gain.

$Sandisk(SNDK.US)$Apple(AAPL.US)$Micron Tech(MU.US)

First, a question: After Micron's earnings report last night, the stock rose over ten points. Did you make money?

If the answer is "no," don't rush to slap your thigh in frustration. Let's calmly break it down and see exactly which step got stuck.

First scenario: You had no idea Micron's earnings would be so good.

You didn't know how strong the memory sector was in Q3—Micron's quarterly revenue was $41.46 billion, a staggering year-over-year increase of 346%, with net profit reaching $28.24 billion, up nearly 14 times year-over-year, and gross margin soaring to 84.9%. If you haven't even seen these numbers, it's normal you didn't profit. The information wasn't in your hands, so the opportunity naturally had nothing to do with you.

Second scenario: You knew the earnings would be good, but still didn't make money.

Knowing ≠ doing. There are two things missing in between:

One is position management. You're fully invested in AI hardware, with no cash on hand, no hedging tools. You hold on during the dips and can't add more during the rallies. Without room to maneuver, you can't seize even the best opportunities.

The other is expectation gap. Market expectations were for revenue of $35.84 billion, but the actual was $41.46 billion, exceeding by $5.6 billion. What you need to judge isn't "whether the performance is good," but "whether the performance is so good that it exceeds everyone's expectations." This judgment is what the capital market truly rewards.

Third scenario: You knew the performance was excellent and have been holding all along without moving.

Then I'd say—it's actually right that you didn't make money on this move.

Because it's not that you didn't understand, but that you chose a bigger logic. Micron has signed 16 long-term supply agreements, locking in $22 billion in customer cash deposits and financial commitments, with remaining performance obligations of approximately $100 billion, covering 2026 to 2030. The company expects HBM supply tightness to persist beyond 2027, with Q4 revenue guidance at $50 billion (±$1 billion) and gross margin expected to further improve to 86%.

To chase a "small profit" of over ten points, giving up the subsequent sustained upside potential—that's picking up sesame seeds but losing a watermelon. True alpha comes from faith in the logic and the compounding of time.

So, stop dwelling on it.

Micron is just the beginning. Marvell, SanDisk, SpaceX, and Zhipu are still to come. Every stock will have its moment like this. What you really need to cultivate isn't "catching it every time," but "thinking clearly about what you want."

Which level you're on determines which level of money you make. By the way, the biggest beneficiary of Micron's stellar earnings isn't Micron itself, but SanDisk, GigaDevice Innovation, and other memory stocks. It might seem counterintuitive, but think about it a bit deeper and you'll know why. I'll share more tomorrow.

$Micron Tech(MU.US)$Sandisk(SNDK.US)$GIGADEVICE(03986.HK)

Recently, global stock markets have plummeted, South Korean stocks triggered circuit breakers, the Nasdaq fell sharply, and AI hardware stocks collectively suffered heavy losses. Various analyses are flying around: some say funds are seeking safety before Micron's earnings report, some say profit-taking is pouring out, and some say it's a ripple effect. They all sound reasonable, but looking back today — these are all noise.

What is noise? Nobel laureate in economics Daniel Kahneman explains it very thoroughly in his book "Noise: A Flaw in Human Judgment."

Bias is that "habitual thought" you always have in your mind — for example, "memory stocks have risen too much, they should fall." This thought has a fixed direction, and you know where it comes from.

Noise is different. It's like standing in the middle of a wet market, with people shouting from all directions, and you can't make out what anyone is saying. Today, this analyst says there will be a surge tomorrow; tomorrow, that media outlet says the bubble is about to burst; the day after, there's news that institutions are selling. These voices are chaotic and directionless, with no pattern. The more you listen, the more panicked you become, and the more panicked you are, the less you know who to believe.

Kahneman said something very interesting: Where there is judgment, there is noise. As long as you are making decisions, noise will definitely exist. You can't avoid it, but you can choose not to listen.

The most frightening thing about noise is not its destructive power itself, but that it interferes with your judgment on truly important matters.

Micron releases its earnings tonight. Market expectations are for Q3 revenue of $35.2-35.5 billion, a year-on-year increase of about 279%; net profit is expected to be $24 billion, a year-on-year increase of over 1100%. This is what's truly worth paying attention to — the company's performance, not the market's sentiment.

Learn to ignore market chatter and wait for the earnings report.

$Micron Tech(MU.US)