Moonlight 月光

The Boy Prodigy Investor Leopold really turned life into a movie these past two days

During the day, because of high-leverage bets on AI, his funds were forced to sell most of their publicly traded stock positions

On the weekend, he still held a wedding as usual

And the bride, Avital Balwit, is no ordinary person either—she is the Chief of Staff to Anthropic CEO Dario Amodei, an AI safety researcher, and she has also participated in discussions that helped shape Leopold’s《Situational Awareness》thought process

She even wrote a line from a poem that left a deep impression on me:

“It’s August and the trees are bleeding.”

One stream studies capital, compute, and the AI race; the other studies the risks to civilization, AI alignment, and poetry

The foundation is bleeding badly, but the wedding goes on

During the honeymoon, there are still one-on-one meetings with investors

$NVDAB

The next phase of AI competition may no longer be only between models, but between networks.

What the market ultimately cares about is not who has the best software, but who can bring together more Agents with execution capability, trustworthiness, and economic value.

Recently, I’ve been paying attention to AIMETA @AImetalabs. They are building a verifiable AI Agent execution network for Web3, and have already begun collaborating with infrastructure ecosystems such as CESS, Crust Network, DIN, and others.

$AIMETA will be listed on PancakeSwap tonight, Singapore time, July 31 (Friday) at 21:00.

Be sure to watch @AIMETA and see how they turn AI Agents from content-generation tools into real execution entities that participate in economic activities.

📌 Before trading, please make sure to verify the official contract address:

0x753Be4F15c765DD6Eb455Aa18d12ad0c6Fb85eE3

Beware of counterfeit tokens, phishing links, and fake accounts.

How惨, this AI prodigy has also been taught a lesson by the market😭

The AI-themed hedge fund Situational Awareness, founded by former OpenAI researcher Leopold Aschenbrenner, reportedly emptied all its publicly traded stock positions before the U.S. stock market opened on Thursday, and also sold a large amount of its own Anthropic shares.

The fund manages about $20B to $24B.

As of the end of June this year, its net return has reached as high as 439%.

It was originally one of the best-performing hedge funds this year, and also made Leopold the new generation of stock god.

However, with recent sharp declines in AI-related stocks, major pullbacks in the fund’s heavily held positions—SK Hynix, Nvidia, Micron, CoreWeave, and others—along with a squeeze affecting its short positions in software stocks, caused the assets to shrink rapidly.

It’s currently reported that institutions such as Bank of America, Goldman Sachs, and JPMorgan Chase are helping handle margin calls, while Leopold has also begun seeking new funding from existing investors and lenders.

Once a dazzling AI-themed hedge fund, it has now been forced to liquidate all of its publicly traded stock holdings.

$NVDAB

Dear, recently there’s something quite interesting going on in the AI / U.S. stock market community.

That is, Microsoft and Meta released their earnings reports at almost the same time.

And both companies’ report cards were not bad.

But the market reacted completely differently.

Microsoft surged higher after hours for a time.

Meta, however, saw clear price swings due to its latest capital expenditure guidance.

Seeing this, I think the signal the market is sending is actually very clear.

AI has entered the next phase.

The focus has shifted from “who is best at telling AI stories” to “who can truly make money with AI.”

What impressed me most about Microsoft this time is that its Azure cloud business has, for the first time, exceeded $100 billion in annual revenue.

This number is very significant.

In the past few years, Microsoft has been aggressively buying GPUs and building data centers—many people have been debating whether the costs are truly worth it.

Now the answer is getting clearer and clearer.

These investments are steadily turning into real, tangible revenue.

Copilot and Azure are products that enterprises are willing to pay for over the long term.

Once companies adopt them, the migration costs are also very high.

So Wall Street is more willing to believe that Microsoft’s AI is now in a virtuous cycle.

Continuing to expand capital expenditures also makes it easier to gain support from investors.

Meta’s situation is a bit different.

This time, its advertising business is still performing excellently, and user data also looks great.

But management also announced that it will continue to increase the scale of AI investment going forward.

What the market is worried about is also easy to understand.

Everyone knows AI needs to be invested in.

It’s just that as investment keeps getting bigger, when it will generate corresponding returns is still not very clear.

Plus, many people still remember how Reality Labs burned through huge amounts of money back then.

So the market reaction is naturally more cautious.

I think these two companies actually represent two different investment logics.

Microsoft is more like “certainty.”

Mature enterprise customers, stable cash flow, and AI has already started contributing to revenue.

Each step is relatively easy to verify.

Meta is more like pursuing a bigger burst of explosive growth.

It has billions of users worldwide, and if in the future AI agents, smart glasses, or even new AI products find a truly viable business model, its growth potential is still very much worth期待.

So after this earnings report, my biggest takeaway is that the standard the capital markets hold AI to is really getting higher and higher.

When you feel like you’ve lost a lot

Don’t rush to be sad first

Take a look at this professional manager

Only 50 days into the job

He’s already turned the “newbie protection period” into “hell mode”

Did he maybe buy at the peak for $SNDKB ?

Suddenly, it feels like

my account isn’t that miserable either 😭

Two months ago

Legendary investor and former OpenAI researcher Leopold Aschenbrenner

Sold Nvidia $NVDAB

Put a bunch of capital into power and mining sites

(Quick addition: he’s only 25 this year—he graduated from Columbia University at 19. A bona fide prodigy.)

And then today, news broke

That his Situational Awareness hedge fund, which he founded and manages and has a scale of about $20 billion,

After this AI stock-market crash,

has started seeking a capital increase from existing investors and lenders

And even offers some investors the option to directly buy the fund’s investment portfolio assets

Seeing this news,

I suddenly felt like my own investment performance

doesn’t seem so terrible after all

Turns out

even a young stock god

can’t escape a lesson from the market teacher 😭

Today stocks plunged

Not only Korean stocks were wiped out

Taiwan stocks also hit the 7th-largest closing drop in history

I quietly canceled the plan I had to take my family out for an all-you-can-eat seafood buffet

Instead, I’m asking everyone

Should instant noodles be topped with an egg, or is hot water enough 😭

So it turns out that Wang Hong’s recommendation letter was written by Lei Jun!

But it wasn’t Xiaomi’s Lei Jun—it was a Lei Jun with the same name and surname, her Ground/Air Defense Class teacher

What’s most unbelievable is that not a single math department professor was willing to recommend Wang Hong

No wonder that after Wang Hong won the 2026 Fields Medal, she said that for a long time at Peking University she felt discouraged

It felt like, “As long as I can survive, that’s enough.”

Only after she went to France and the United States did she slowly regain her confidence

Did Peking University’s math department back then look down on Wang Hong?

During her application process for studying abroad, the math school didn’t help at all. In the end, it was Lei Jun, a teacher from the Ground/Air Defense Institute, who wrote the recommendation letter

When she graduated, Wang Hong didn’t take a graduation photo with the math school either. Instead, she rushed back to the Ground/Air Defense Institute to take a picture with her first-year classmates

But the moment Wang Hong won the Fields Medal, Peking University’s math department became actively eager to claim her as their pride

Where did all the student support go back then, when students needed it the most?

Dear, many people don’t know that in the United States, almost every state has its own official website for “Unclaimed Property”

If you have some time, you can search your name—you might be able to recover money you’d long forgotten

For example, government refunds, tax refunds, paychecks, bank account balances, insurance claim payouts, deposits, and more—could all be involved

Taking California as an example, the official website managed by the California State Controller’s Office has already helped people reclaim more than $8.38 billion

The search is completely free, with no application deadline. Just enter your name to check

Remember to use only the official website—don’t look for third-party agencies, and don’t pay any query fees

If you’ve ever lived, worked, or studied in the U.S., it’s worth spending a minute to check—you might be surprised by an unexpected bonus

Other states also have similar official lookup websites, so you can search them together

#btc

AMD Chairman and CEO Lisa Su spoke at the AMD $AMDB AI conference

Lisa Su and NVIDIA $NVDAB founder and CEO Jensen Huang are cousins

The two of them lead AMD and NVIDIA respectively, becoming two of the most closely watched Chinese-American business leaders in the AI chip field

After the AMD AI conference, the latest target price adjustments from major institutions

📈 Raised target prices

Baird: $625 → $1250 (upside +139%)
UBS: $700 → $730 (upside +40%)
Roth MKM: $500 → $650 (upside +25%)
Bank of America: $560 → $620 (upside +19%)
Jefferies: $515 → $640 (upside +23%)

📌 Maintained target prices

KeyBanc: $725 (upside +39%)
Cantor Fitzgerald: $700 (upside +34%)
Rosenblatt: $665 (upside +27%)
Barclays: $665 (upside +27%)
Stifel: $635 (upside +22%)
Wells Fargo: $615 (upside +18%)
Bernstein: $600 (upside +15%)
Citi: $575 (upside +10%)
RBC Capital: $540 (upside +3%)
Morgan Stanley: $410 (downside -21%)

【How to Make $600,000+ per Month Using a Spread Strategy?】

Dear, I just saw a trader’s recent share—by relying solely on cross-market spread trading, they accumulated more than $600,000 in profits in one month.

What I care about most is the long-standing huge price spread between SK Hynix ADR and the underlying shares in South Korea.

As of July 25, compared with the South Korean ordinary shares, the U.S.-listed SK Hynix ADR $SKHYB is still about 28% to 30% higher after conversion.

Many people’s first reaction is that this looks like an arbitrage opportunity.

But reality is not that simple.

For this spread to persist for so long, there are mainly three reasons.

First, the U.S. and Korean markets face different investor bases. Many overseas funds would rather directly buy ADRs with better liquidity.

Second, the additional conversion quota for SK Hynix ADR is nearly used up. As arbitrage supply is constrained, the high premium is easier to sustain.

Third, the trading hours differ between the two markets. The U.S. market reflects AI- and semiconductor-related news first, while the Korean underlying shares usually only adjust at the next day’s open.

Many people choose to short the ADR, buy the Korean underlying shares, and wait for the spread to revert.

But the real difficulty is that the spread could remain for a long time.

Borrowing costs, funding/interest rates, and transaction fees will keep eroding profits—and 30% may not even be the limit.

After going through the full case, my biggest takeaway is actually not the arbitrage strategy.

Every time you see a large price spread, you should first think about why the price deviates, why it can keep persisting, and whether your capital can hold out until the spread returns to normal.

The place where many excellent traders truly widen the gap is often risk management—not the speed of finding opportunities.

When I was a kid, our family was in poor circumstances.

While others learned with GPT, Claude, Kimi, and DeepSeek,

I could only use Doubao.

Every day, I listened as she spoke with a straight face and made up nonsense.

Over time, somehow I developed the skill to tell whether information is true or false.

Only later did I realize:

Poverty didn’t limit my growth.

It just turned the beginner tutorial into a hard mode 😂

Apple and Micron have recently gone head-to-head over memory chips

According to The Wall Street Journal, the two sides have already taken the battle all the way to the White House

Apple wants the U.S. government to allow products sold overseas to continue using chips from China’s CXMT and YMTC, and its reasoning is straightforward:

lower costs and ease global supply tightness

But as the largest domestic memory chip maker in the United States, Micron $MU strongly opposes this, arguing that if restrictions are loosened, Chinese manufacturers will further impact U.S. domestic industry

The arguments from both sides are also quite interesting

Apple says Micron’s profit margins are too high, chip prices have been driven up noticeably, and much of the new capacity is being prioritized for AI customers, leaving other companies unable to buy supply

Micron responded that during the market downturn in past years, major customers including Apple kept pushing down procurement prices, causing the industry to be reluctant to expand capacity for a long time; the current supply shortage is actually the result of years of accumulated factors

Over the past year, memory chip prices have risen by about fourfold

The most intriguing part of this whole matter is

Apple has long been criticized for charging steep fees for memory and storage upgrades

Now that upstream suppliers hold the pricing power, Apple is the one complaining that prices are too high

It can only be said that when the role in the industry chain changes, the stance changes as well

In addition, The Wall Street Journal also mentioned that the relationship between Micron CEO Sanjay Mehrotra and Apple has long been less than cordial

It is said that this grudge can be traced back to when he was still at SanDisk. Back then, business negotiations were extremely intense, and senior executives on both sides rarely even met directly

$MU $AAPL.US

After delivery regulations went into effect, delivery riders, consumers, and merchants have all been affected

Overall, while there are some added safeguards, there is also a great deal of controversy

🛵 Delivery Riders

Benefits

A minimum of 45 yuan per order
When converted, the hourly wage must reach 1.25 times the basic wage, about 245 yuan
The platforms must also provide group accident insurance and liability insurance

Potential Issues

After platform costs increase, stacked orders may start to decrease
Some delivery riders report receiving fewer assignments; actual income may not improve noticeably

👛 Consumers

Benefits

The likelihood of multiple orders being delivered at the same time may decrease
Each order becomes more independent, so delivery speed and meal quality may be more stable

Potential Issues

As platform costs rise, expenses may be passed on to consumers
Uber Eats has adjusted some merchant service fees and membership subscription fees
In the future, there may be fewer free-shipping promotions and increases in meal prices, etc.

🍴 Merchants

Benefits

If a platform wants to adjust commission rates or add new fee items, it must get both parties’ consent
Large brands also have greater bargaining power

Potential Issues

The platform service fee cap can be as high as 35%
Smaller and mid-sized merchants have weaker bargaining power, so profits may be further squeezed

In short:

Delivery riders gain increased pay and insurance protection
Consumers may face fewer stacked deliveries, but may have to bear price increases
Merchants receive more fee protection, but smaller shops still carry a heavy cost burden

So, will this new system truly improve the delivery environment?

It’s only just started—so far, the negative news Moonlight is seeing seems to be more prominent…

Wow, $NVDAB $NVDA Founder Huang Renxun Jensen Huang just shared new news!

This is an open letter jointly signed by multiple AI companies including NVIDIA.

Moonlight thinks the contents here are worth paying attention to.

Key takeaways:

AI will impact every industry, every company, and become part of each country's competitiveness.

Open-weight models can attract more developers, improve safety, promote innovation, and help companies and countries build their own AI capabilities.

Meanwhile, the most advanced closed-source models will continue to push past technical ceilings.

The future of AI is likely not a single model.

Open ecosystems will expand applications, while closed-source models keep driving the frontier of technology—developing both routes together will help the entire industry move faster and go farther.

Dear, every time a new technology explodes, the market amplifies both greed and fear at the same time.

Some people want to be the next winner, while others are afraid they won’t be able to keep up with the times.

So everyone starts抢ing GPUs,抢ing electricity,抢ing data centers—and even stuffing demand from decades into today’s orders.

Recently, Moonlight Research looked into the AI infrastructure field and two highly watched “NeoCloud” compute rental providers: CoreWeave and Nebius.

After the research, my biggest takeaway is that what they offer is no longer just compute services.

They’re more like they’re taking on massive capital expenditures on behalf of large tech companies, and then—using these long-term orders—continuously expanding their financing capacity, buying more GPUs, and forming a growth cycle that keeps compounding.

And NVIDIA $NVDAB stands at the very core of the ecosystem.

It’s not only a GPU supplier, but also an investor—while also providing the most important credibility endorsement in the market.

That’s why, beyond chips, the truly precious resource in the AI boom is how long the market is willing to believe in this growth story.

In past technology bubbles, it was rarely the disappearance of demand that started them.

What really needs attention is that demand still exists—but it gets amplified layer by layer by capital, financing, and leverage, until the whole market finally loses its reason to “get off the train.”

Dear, among the five hottest US stock holdings, which ones do you like?

📌 Google $GOOGLB
AI capabilities are still firmly in the top tier, and Gemini still has plenty of room to grow

📌 Marvell $MRVLB
AI networking and the demand for custom chips continue to grow—there may be an opportunity to break through a trillion-dollar market cap

📌 Oracle $ORCLB
One of the key beneficiaries of the US AI data center buildout; the valuation should still have room to improve

📌 Hynix $SKHY
HBM and AI memory demand remain strong, and it’s still an important beneficiary of this AI wave

📌 Super Micro Computer $SMCI
Demand for AI servers is strong. As long as industry sentiment stays favorable, its performance is still worth期待 this year

The man who once made the moonlight hurt the most

Was—Musk 😭

Back then, when it opened at $SPCXB $SPCX , I bought it and lost a lot

At least I could earn some of it back by shorting it back and forth

But that also makes me wonder—since it’s the same company under Musk’s name

Should we short $TSLAB together too?