CRWV 的资产负债表在新云厂商中属于最差之列：流动性约 117 亿美元，而 12 个月内到期的账单为 180 亿美元；股票和债券下跌使融资更加困难。他对 64% 的五年违约概率表示怀疑，预计不会破产、也不会引发系统性危机，但会出现稀释，并称当前低价是买入机会。

There is a vicious cycle of pessimism in the markets right now, and this is largely related to the uncertainty surrounding CoreWeave and its debts. I mean, we've all probably heard about these charts and how terrifying it is, with everyone saying, "Oh my God, CoreWeave's credit default swaps are skyrocketing."

The reason people are panicking about this is that it reminds us of 2008, when people were trading credit default swaps (CDS) and we saw those big crashes that ended up making 100-to-one profits in the real estate market.

CoreWeave is a modern inference cloud. About 25% to 26% of their active energy is dedicated to inference, but 74% of their contracted energy is located in third- or fourth-tier rural markets, which is very good for inference.

More inference means good news for CoreWeave because 74% of their contracted capacity is in inference-friendly markets.

This will help us understand what is happening with CoreWeave and its vicious cycle of debt.

If I wanted to run a 700GB model and load it with enough space to work, I could do that using nine H100 cards. This is the type of hardware CoreWeave will have.

If you want to train a model like this, that's great. If you want to use it for deduction, that's fine. You can also use CoreWeave. But it is less attractive because, again, you are building in training markets rather than conclusion markets.

Thus, CoreWeave faces a double-edged sword, as proxy computing and even large-scale models that people want to provide can now be run internally on Mac Studio machines. So, it seems like we need less Korwef for this likely purpose.

Furthermore, there is a fear that we may not need training as much; we just need more agents anyway. Therefore, advanced models will become less important, and certainly H100 or even Blackwell chips may also become less important.

Let's get out of the Korwef deals. This is what some hedge funds and institutions are doing, and I believe this is leading to a decrease in the price of Corwef.

But Koroyef faces problems that go beyond the mere fact that it is heavily exposed to training activities. They have a problem with their balance sheet. Their budget problem indicates that if we combine their cash flow with their receivables, assuming everyone pays their bills, we arrive at $8 billion.

If you now add another $3 billion to $3.7 billion of debt they have raised, their total available liquidity will amount to approximately $11.7 billion, assuming everyone pays back.

But they have $18 billion in bills that must be paid within the next 12 months. So, they don't actually have liquidity in the short term.

Now, let's combine these two things. The shift towards generative AI and a slight move away from training are both hurting Korweef, in addition to it probably having the worst balance sheet among new cloud companies.

So, what happens next? The stock price is falling. Okay, so what happens when the stock price drops? People get the impression that if the stock price falls, the company becomes more vulnerable, and its debts may become more dangerous.

The worse the Korwef stock performs, the more difficult it becomes for them to raise cash by issuing new shares. But the problem is that whenever the stock performs badly, the bonds also perform badly because people think that someone is expecting a problem on the balance sheet.

Therefore, the bonds lost 20% of their value. When bonds fall by 20%, and trade at around 80 cents on the dollar, people start saying, "Oh my God, Korwyff won't be able to raise the money to pay their bills.

They're already in a tight financial position. Now, they simply won't have enough money to pay their bills."

In other words, it will become difficult for them to sell shares because the share price is falling. It will be difficult for them to obtain loans because the value of bonds is declining in the secondary market.

All four of these things combine to deliver a severe blow to CoreWeave.

Because you must remember, the worse CoreWeave performs, and all four factors put it in a worse position, the better the swaps perform, and the more insanely their value rises.

As this graph climbs higher and higher, companies like LSEG—formerly Refinitiv—come out to say that this company now has a 64% chance of going bankrupt or defaulting within 5 years.

I don't know if it's that bad. I know their balance sheet is bad, but they are still above their initial public offering price. Therefore, they will still be able to offer some shares.

I think what's happening is a vicious cycle, where the stock price falls, credit default swaps rise, and as they rise, the value of bonds falls. As bond values fall, people panic.

Credit default swaps are increasing. In addition to the other factors I mentioned, this indicator continues its steady rise.

And then people will say: Oh my God, we knew that. That balance sheet is really bad. Or worse than previously thought. We could see the poor balance sheet from afar, but this vicious cycle leaves people in a state of shock.

Credit default swaps are rising because traders are speculating on them. This necessarily means that the company's performance is getting worse. Thus, it enters a vicious cycle where credit default swap contracts rise higher and higher, and the higher these contracts rise, the more bonds and stocks are sold.

The more of those sales there were, the higher the credit default swap contracts rose.

And so, yes, this is the case where a company that may not have a 64% probability of bankruptcy appears to have a 64% probability of bankruptcy according to the market. In fact, CoreWave literally said, "CoreWave's power is improving every day and I am optimistic." "Their cost of capital is decreasing."

Or, their strength may improve every day as they build up supplies, computing capabilities and energy provision. That may be true. But this statement: "Their cost of capital is decreasing."

I don't know where you got that from, Jensen, because if the secondary market collapses, the cost of new capital rises. So, this is not good. But they own 47.2 million shares.

Next on the list is Magnitar Financial, and it all seems like we're rewriting the dot-com bubble story, isn't it? I...don't misunderstand me. I think it's just a bubble and it will burst someday in the future.

Anyway, if you put all of this together, yes, in the end this will be a big bubble that will burst, collapse and fade away, but I don't think CoreWave is the black swan.

Core Wave? This is just a side note. This is a vicious cycle. It's like the opposite of "Game Stop". And I want to be clear here. I do not have any financial positions in CoreWeave.

I always like to make my biases clear beforehand. I don't care at all if the value of this thing reaches zero, because even if that happens, and interest rates fall, it will still be beneficial for real estate.

If there is a bias, it is on this side. So, to reach zero. But I actually take the opposite approach here, and I say that I don't think so.

As you know, Bernstein describes them as the most vulnerable technology cloud. This may be true. At least in the public market. But for me, there are still no clear indications that they will definitely go bankrupt.

There are many "rescue friends" and phone calls they can make before they reach bankruptcy. They will simply relinquish their shares and dilute the value of shareholders' stock.

That's what I expect. Therefore, I don't see this as a systemic crisis. Rather, I see it as an opportunity to buy when prices are low. This is not the beginning of the end.