会飞的鱼flyfish

I've lost a lot since July. Selling is the real skill. Since things have come to this, there's nothing else we can do.
At 3:00 PM Beijing time, an announcement will be released! After the close, please watch for market volatility.
After President Lee Jae-myung recently called for additional measures to support the stock market,

the related measures will be announced at 3:00 PM this afternoon. Previously, the government will hold a market outlook assessment meeting at 4:00 PM.

Because the announcement time is after the stock market closes, it may affect after-hours trading and the market the following day. The market

I don’t believe the stock market can be like the crypto circle—drawing a Christmas tree in such a short time. The fundamentals are still looking good.

But the market is indeed too fragile; a short essay can cause such huge swings.

The de-leveraging process in South Korea won’t be easy, and the holding experience has worsened a lot
#三星 #海力士 #MU

Meta’s one line, “I’m not playing anymore,” and the market immediately freaked out and plunged.

This shows that what’s truly fragile right now isn’t AI itself—it’s market confidence.

In the past, when people bought AI, they believed big companies would keep burning cash, keep expanding Capex, keep buying GPUs, HBM, memory, and optical modules.

But now, as soon as one major player slightly lets up, the market will immediately start asking:

Will AI demand peak?
Can compute capacity orders keep going?
Have these overvalued assets run up too much?

So this isn’t AI ending.

Instead, AI trading has moved from the “faith stage” to the “order-validation stage.”

AI will keep going, but the market won’t blindly fund every AI story anymore.

Next, the only companies that can truly ride out volatility will be those whose orders, profits, and cash flow can all be delivered.

Samsung and Hynix fall more than 5% during trading; it is rumored that the Korean government is again studying the redistribution of semiconductor excess profits. It feels like everyone is eyeing this fat prize.
I still like the memory stocks. They won’t just stop here—prices started rising in the third quarter, and the next earnings reports should look even better.
- U.S. investment bank Jefferies expects that DRAM and NAND flash prices will rise 40% to 50% quarter-over-quarter in the third quarter of this year.

1. New factory construction in South Korea: short-term impact on sentiment, but in the long run it won’t really affect anything (5-year expansion, doubled)
Seeing this news made me laugh
2. The three major storage companies have been sued. Based on how things develop, we’re just worried it could get blown up into something big and end with a conclusion.
That is: big consumer-end companies like Apple want to
make more money themselves and don’t allow storage to earn so much—so first let smaller suppliers file lawsuits and see how it goes. Wouldn’t it be a matter of character and ambition among princes and commoners?
3. The market is generally discussing the impact of Changxin Storage. This is something where different people see different things—no clear answer. The memory market currently is
still very short; another Changxin probably doesn’t matter. But it will affect short-term structure—especially if Apple switches suppliers, which would cause major volatility. However, in the long run, it’s still upward.

Latest news on SK Hynix: The American Depositary Receipts (ADR) are set to launch in July, aiming to raise up to 45 trillion KRW (approximately $29 billion) through the ADR listing.
#dram #海力士 #mu

$HYPE calculating pe is still not as good as $UNI , I don't get why people choose HYPE over UNI,

Is it because there are too many whales in UNI? Always dumping?

#UNI #HYPE

According to sources from investment banks and the semiconductor sector on the 16th, SK Hynix is gearing up to launch a shareholder return plan valued at approximately 1 trillion won in the upcoming fourth quarter, which includes a stock buyback and cash dividends. It is reported that the buyback will account for about 2% of the total share capital, similar to the volume of American Depositary Receipts (ADRs) set to hit the market next month. SK Hynix plans to complete the ADR listing by mid-next month, pending approval from the U.S. Securities and Exchange Commission (SEC). An insider from the investment banking sector stated: "After formally announcing the U.S. listing plan, SK Hynix received positive feedback during its roadshow aimed at global institutional investors, boosting the company's confidence in its future operations." They added: "Once the ADR listing is complete, the company is expected to unveil a shareholder return plan worth as much as 1 trillion won around the release of its third-quarter earnings report."

Stop worrying about whether there's a bubble or not; the real top isn't something a retail trader like you can predict. If you're worried about a bubble now, do you think AI isn't useful? Think back to when AI really started making waves—it's just this year. If you haven't been in the game for a few years to validate the market, listening to those folks saying it's all over is just plain foolish!

Hynix closed up 15.75%, as I mentioned yesterday, every dip is a buying opportunity. Old Huang said yesterday: "When stocks dip, you should feel excited because you can buy in at a cheaper price!" Right?
#美股 #内存 #dram

June 8th Tech Market Highlights

1. Tencent Bond Issuance: Establishing a $30 billion medium-term note program, looking to raise $3 billion for investments in AI and refinancing.

2. Jensen Huang's Speech in Korea: Stated that AIGC is just getting started, "The market dip should excite you, it's a chance to buy at a discount"; the KOSPI plummeted 8.29% that day.

3. US Semiconductor Bounce: The Philadelphia Semiconductor Index rose by 5.61%; Intel jumped 11.19%, and Micron climbed 9.87%.

4. Google x Intel Collaboration: Google has ordered over 3 million TPUs (to be delivered before 2028), Intel saw a pre-market surge of 14%.

Jensen Huang says the AI build-out is just getting started, and with the stock market dipping, you should be stoked because you can scoop up some assets at relatively cheap prices.
In the AI space, the only player with a clear moat, with orders booked out for years and a cost share of 50%, is DRAM. So why are you still sweating it?

【Why did Marl gain half an Ethereum market cap in just two days? Here's the scoop from Lao Huang: The next bottleneck for AI data centers is defined as connectivity/interconnect, which means interconnection, optical interconnect, switching, and custom XPU】
1. connectivity/interconnect = interconnection (the core of the bottleneck)
Training large AI models isn't just one GPU doing the work; it takes hundreds or thousands of GPUs tied together to form a cluster.
Computing power relies on GPUs/XPUs, while data moves back and forth via interconnect lines.
Previously, the bottleneck was: not enough GPUs, or poor chip performance.
Now the bottleneck has shifted to: data transfer speed between GPUs is too slow, leading to network congestion → this is the interconnection bottleneck.
2. Optical interconnect (optical modules)
Traditionally, data signals are transmitted via copper wires, which have a low speed limit and generate a lot of heat;
Optical interconnect = using light to transmit data in fiber optics, the only solution for high-speed interconnection in supercomputing/AI data centers:
Servers → switches are connected via optical modules, and large-scale AI clustering relies entirely on this.
As computing power piles up, the demand for optical modules skyrockets, and specifications evolve rapidly, becoming a critical bottleneck.
3. Switching (switches)
Think of it like your home router, but data centers use gigantic 10G/100G switches:
All GPU data is aggregated and forwarded through switches; the massive data throughput for AI places explosive demands on switch bandwidth and latency.
If the switches can’t keep up, no matter how many GPUs you throw in, they’ll just be spinning their wheels and not utilizing their full power.
4. Custom XPU
GPU: General-purpose graphics cards from Nvidia
XPU: Various types of customized dedicated computing chips (DPU/TPU/ASIC, etc.)
To tackle interconnection congestion, manufacturers are no longer just stacking general-purpose GPUs:
They’re creating DPUs (data processing units): specifically designed to handle networking tasks, freeing up GPUs to focus solely on AI calculations;
Developing proprietary TPU/ASIC chips, with high-speed interconnect interfaces integrated directly into the chip;
Using custom chips to optimize interconnection efficiency from the hardware level.

With a market cap of 700 billion, someone just dumped 150,000 lots today, pure rug pull, finally got played! That's some serious moves! These code-copying clowns really don't know which idiots are buying, the government needs to step in. Letting these jokers harvest retail investors won't lead to any R&D; they'll just be scheming to move their assets overseas.

The SEC has officially greenlit blockchain firm Paxos as the first approved securities clearing and settlement agency, allowing it to process securities trades.

On Capitol Hill, the ones who really get blockchain aren’t the lawmakers; it’s those regulators who just figured out how to use mobile payments.

Bezos's Blue Origin's New Glenn rocket exploded during static fire testing.

Well, there you go, Bezos isn't just sending rockets to the moon, he also added a little "fireworks show" to the mix.

After Trump shouted "buy Dell" on May 8, the stock has surged 57%, and yesterday the U.S. signed a $9.7 billion government contract with Dell.

A tweet that pumps the stock leads to a market cap in the hundreds of billions; it turns out the action right outside the White House is the most stable play.

The S&P 500 just closed at a new high, hitting 7,563 points.

The pace at which global central banks are printing money is more stable than the issuance rate of meme coins.

US Treasury Secretary Yellen is publicly urging Congress to pass the Crypto Clarity Act, stating "the most important thing is to bring digital assets to the US."

It seems the US government is finally getting it—rather than letting the crypto industry escape to Singapore or Dubai, it’s better to lay down the rules themselves.

Anthropic just wrapped up a fresh $6.5 billion funding round, hitting a post-money valuation of $96.5 billion.

Another AI player flexes its imagination to score a historic valuation; investors are betting that AI can not only write poetry but also rewrite the algorithms of the global economy.