资管舵

Livermore thought cotton would rise, but the market hadn’t started yet—it kept hovering around 12 cents.

He couldn’t wait, so he bought 20,000 bales. After he finished buying, the market kept falling. Near the bottom, he cut his losses, losing $300,000.

A few days later it climbed back up. He started itching again, bought another 20,000 bales, and when it fell back down, he cut again at the very lowest point.

He repeated this cycle five times within six weeks, losing $2 million. Livermore got furious and told the manager to dismantle the cotton quoting machine.

Then two days later, the main upswing in cotton kicked off. It climbed all the way with basically no pullbacks.

He knew it in his bones, but he just couldn’t control himself.

You understand its logic, but before it has a chance to play out on its own, you think you can push it along. The result is that you become the very resistance.

Livermore’s third bankruptcy happened because he believed insiders’ information from so-called experts. He met cotton tycoon Thomas. Livermore originally had his own judgment, but Thomas talked with him a few times, fed him inside information, and convinced him to go long on cotton.

At that point, however, the market had already begun to weaken. Livermore went against his own sacred rule—trust the ticker tape, not rumors—and listened to other people’s messages.

Then cotton crashed. He didn’t cut losses right away; instead, he added more to average down his cost, buying more the more he lost. That’s how he ended up bankrupt for the third time. He sold two yachts to pay his debts and still owed $1 million.

He said this was the biggest blunder of his entire trading career, and that he had to stick to his trading discipline without fail.

The Purpose of Modern Finance.
Is always to prevent ordinary people from completing primitive accumulation.
Consumerism is responsible for brainwashing.
Financial instruments are responsible for harvesting.
To achieve good results in investing,
you must first have a level of understanding above that of the average person,
be able to want to do what ordinary people can’t do.

？

A single chart to understand OpenAI’s ledger: compute, infrastructure, and power commitments total about $1.43 trillion.

The top three amounts are Broadcom at $350 billion, Oracle at $300 billion, and Microsoft at $250 billion, stepping down from there into the hundreds of billions.

The headline is even more striking—its scale is roughly equal to the combined total of Google plus Oracle. Note that these are multi-year commitments, not cash outlays in the current period; both of the Broadcom and AMD figures are estimates.

Commitments can sound very grand—whether they are delivered depends on financing rounds lining up one after another.

Goldman’s assessment is very straightforward: the world is facing a shortage of copper and silver.

Seeing the chart makes it more intuitive. This is global copper inventory: dark blue is the United States, light blue is everything outside the United States. In just a little over two months, that U.S. line was stacked from 500 thousand tonnes up to nearly 1,500 thousand tonnes—more than doubled. The portion outside the U.S. is also rising, but the slope is much gentler.

The total is going up. From afar it looks like there’s no shortage, but the bulk is being piled up inside the U.S., while the portion that can actually be called on overseas doesn’t look as roomy on the numbers—this is essentially what Goldman means when it says supply is tighter than inventory figures suggest.

Copper and silver are materials that can’t be avoided by power grids, AI data centers, and solar PV. Moving inventory from one place to another doesn’t mean production increases. In the short term it can suppress prices; in the long run, you still have to look at how much the mines produce each year.

Oracle's CDS just hit a new high: the 2056 maturing bond yield first broke 8%, just one step away from junk status. If it gets downgraded, the $120 billion worth of bonds will be automatically removed from the investment-grade index. The trigger is that it issued a force majeure notice to its new data center in New Mexico. Not just it—Google, Microsoft, Amazon, Meta, Nvidia, and SpaceX's CDS are also trending higher. What the market is worried about isn't any one company anymore; it's what funds will keep the entire AI infrastructure buildout going.

The daily chart has exited a large descending channel, and inside it there is also a bullish flag pattern; it even claims that retail investors are still hesitating. In the past five days, a “giant whale” has accumulated more than 470 million XRP, equivalent to about $724 million. Then it throws out four target levels: $1.70, $2.20, $2.70, and $3.65.

Put it out to look at—it’s logically consistent and the chart is drawn nicely. But a few things need to be clarified. Chart patterns are an abstraction and summary of the past, not physical laws; if a level is broken, there can still be a fakeout breakout and then a turn back. As for the so-called whale accumulation, accumulation definitions and on-chain data interpretations vary across platforms; the same batch of data with a different algorithm becomes another conclusion, so it can’t be independently verified. Target prices are the script he sets for himself, not a promise the market makes.

It’s fine to look at others’ judgments. But treating those target levels as the basis for your own position is a different matter entirely.

SpaceX plans to build a supercomputing facility in central and southern China. The official line is that it will be one of the world’s most advanced AI training clusters: over 2.5 million square feet of space, with millions of GPUs packed inside, and more than 2 billion watts of computing power.

2 billion watts is roughly the electricity load of a mid-sized city. For a facility of this scale, the bottleneck is no longer in the chips—it’s in where the power comes from, where the heat goes, and when it can truly be energized and begin running. None of the power, site, or grid-connection requirements can be solved instantly just by throwing money at it.

Computing power is the entry ticket to the AI competition, but between piling up hardware and producing results, there are two hard constraints—electric power and time. In the footage, it’s still a construction site; the foundation has only just been laid.

The numbers are what others have provided. When it gets lit up, and whether it can run at full capacity—those are the real milestones.

This is Fidelity’s Bitcoin power-law chart (as of 2026/9/20). Each peak is higher than the last: $1,137, $19,042, $64,337, $122,765—current price $81,218. The teal band is the power-law trend zone, and it also draws a horizontal “line in the sand: $60k.” Based on the copy, the bull market isn’t over as long as you hold the $60k level, with a 2029 target of $300k.

Key features of the power-law model: it fits the past beautifully—always. It connects historical points into a straight line and then extrapolates outward. Whether the price follows the power law is not a theorem, but an assumption. That line leading to $300k is simply an extension of the trend band, not a prediction.

What’s even more worth looking at is the row below: historical drawdowns range from -57% to -102%, and the most recent segment is marked at -71%. Under this model, dropping 70–80% is still considered normal variation—so while it gives you a target price, it also leaves enormous room for interpretation.

You can look at the chart. The lines are drawn by someone else; the target price is extrapolated by the model. Don’t take the conclusion as proof.

This chart labels DOGE signals as "the third cumulative signal after a short-term decline" "upward momentum continues to accumulate." Several sets of boxes are overlaid on the chart, marking the staged accumulation zones, and the most recent segment is also placed within one of the boxes.

The person who drew the chart seems to be saying: after the earlier boxes were drawn, the price also moved upward for a while; this time is probably the same. The logic isn’t anything new—almost all the "accumulation/accumulation" labels are done this way. It looks accurate in hindsight because the boxes are drawn on areas where price has already risen.

The hard part is right now: is the bottom of the decline being formed by someone collecting/accumulating shares, or is it simply a natural, drifting down move with no buyers stepping in? For the same candlestick, both stories can be told. The only things that can be confirmed from the chart are the price range and volume. The boxes were drawn by someone else, and the signals are also someone else’s.

You can read the chart, but the judgment has to come from yourself.

CoinGlass. This Bitcoin quarterly earnings chart highlights the +43.61% for Q3 2026. If you flip back over the past decade-plus, only the +80.41% in Q3 2017 was higher—hence the phrase “the second-best Q3 in history.”

The numbers themselves are correct, but take a closer look to stay calm. The historical Q3 average is +8.74%, and this year’s +43.61% does indeed outperform by a wide margin; but the previous two quarters were -22.2% and -14.09% respectively. When you drop two quarters in a row, a strong quarter doesn’t automatically mean you’ve filled the hole for the year.

The Q4 average of +77.07% looks the most tempting—though that mean is boosted by a single outlier: +479.59% in 2013. And there’s no inherent connection between a strong Q3 and a strong Q4: in 2021, Q3 was +25.01%, and Q4 was only +5.45%.

You can look at the chart, but don’t treat historical averages as a script, and don’t treat “the second-best” as a direct reason to expect things to rise next.

Why not go get the money wallet from Black Ben Village?

Another chart that puts ZEC and BTC side by side. On the left, ZEC is pressed up against the $1,000 resistance line, marked with “HERE,” with “8 years of accumulation” written underneath. On the right, BTC is in the same position and at the same $1,000 horizontal line—but it broke through earlier, marked with “4 years of accumulation.” The caption is even more direct: when ZEC reaches $1,500, it’s like when BTC reached $1,500 back then; the target is $10,000+.

The chart-maker’s logic is clear: BTC spent 4 years building up and breaking through that line; ZEC spent 8 years, and now it’s standing at the same starting point, so history will play out the same way.

But the side-by-side comparison is still the old three-part playbook—the resistance line is the same, yet the order-book structure, liquidity, and buy-side sources are completely different. What powered BTC’s breakout back then was institutions putting in real money; ZEC now relies more on narrative and sentiment. Between “like” and “is” there are several years.

You can look at the chart—the target price is something someone else gave you, but the position size is something you have to carry yourself.

A single chart puts MSTR and BTC side by side for comparison: both start at $15,861. With 100 shares of MSTR compounding at 52% annualized, after ten years it reaches $1.044 million; equivalent Bitcoin at 38% annualized reaches $397k. The conclusion given in the chart is that MSTR leads by 2.6x.

What you really should focus on isn’t which one is more aggressive—it’s those two annualized assumptions: both 52% and 38% come from linear extrapolation of the past decade, and the note at the bottom of the chart even says "not a prediction." Change the starting point, change the range, and the slope changes. Ten years ago, who could have imagined Bitcoin itself would end up like this?

Both storylines can hold up: people bullish on MSTR are betting that it uses Bitcoin as its core holding and can add leverage; people bullish on Bitcoin believe that Bitcoin itself is the asset and doesn’t need anyone to package it. Other people’s viewpoints are theirs, your money is yours—no one can claim the ten-years-from-now answer on your behalf first.