원문 보기
Welcome to the Wall Street Lounge. Today we're following one investor, Ark Invest's Katie Wood, and what the firm was buying and selling during last week's AI stock selloff. Three tickers directly from the daily trading disclosure. I'll give you an answer right away. Ark bought CoreWeave and Rocket Lab.
She sold Palantir. Wait, she sold Palantir? The same company that actually prints money? That's right. And CoreWeave bought Ark on the very day the shares fell by 15.46%. 381,231 shares in two sessions, about $25.8 million . Oh, it's catching a falling knife with both hands. She doesn't think so. So, the plan is this.
First the disclosed numbers, then all three companies through their actual reports. So can we assess whether this thesis is justified? That's right. And finally, one name I would keep an eye on . First, tell me about the big picture. Here is the data for October from Ark reports. Rocket Lab, about 233,000 shares on October 1, another 169,000 on the 15th, about $25.6 million in net buying.
And CoreWeave? About 189,000 shares on the 14th , another 193,000 on the 15th. Let's say $25.8 million . Both of these stocks were falling rapidly at that time. With Palantir, it was the other way around : three separate sales, on the 1st, 14th, and 15th. About 82,000 shares, approximately $16.2 million.
So they financed the purchase by selling the winner? The reports don't say this, but that's exactly where the money was going. Is this just an October story? No, it all started in September. Ark then also sold Palantir for about $26 million and invested about $38 million in Block and $12.8 million in Rocket Lab.
Two months, same direction. Which makes this a targeted rotation, not a one-day deal. So what really happened last week? Gartner published the report on October 14. It states that over $3.5 trillion in spending on AI infrastructure is at risk of being devalued by the end of 2030. 3.5 trillion. Honestly, this number is too big to comprehend.
Then use something else . To cover capital expenditures, financial costs, and a 12% return on capital , the sector needs about $8.5 trillion in aggregate revenue by 2030. Gartner's own forecast reaches approximately 3.6 trillion. Um, that's not even half of it. And then the line that did the damage.
Over 80% of neo-cloud services will cease to exist as independent businesses by 2030. What is neo-cloud, exactly? How is it different from a regular cloud? This is not Amazon or Microsoft. This is a company whose entire business is renting graphics processors for artificial intelligence calculations.
Coreweave is the largest of them. So, one source of income. That's what Gartner thinks. No diversification and minimal differences from hyperscalers. And Coreweave suffered the most. That's right. That day, the stock fell 15.46%, closing at $65.78. Trading volume reached 107 million shares, compared to a three-month average of 29.5 million.
So, panic. Last week, the Nasdaq closed down 4.5%, snapping a four-week winning streak. The semiconductor index has fallen for eight consecutive sessions. And what is happening outside the field of AI? This is the most interesting part. On Monday, the 19th, the S&P 500 index rose 0.41%, while the Nasdaq added 1.11%.
Oh, so he bounced back. At first glance, the S&P 500 equilibrium index actually fell by 0.32%, and the number of stocks that fell exceeded those that rose on the New York Stock Exchange. Mega-cap companies pulled out of the index. The yield on 10- year Treasury bonds also rose to 5.14%. So what is Wood actually saying ?
On October 16, she was in Seoul for partnership talks and told Bloomberg the following. The fact that people are afraid means we are not in a "bubble." This is a confident statement. She went further. As for the AI, she said we're not even in the first inning. We're still in the parking lot before the baseball game.
In the parking lot. That means the game hasn't started yet. Regarding the cost of borrowing, she said this often happens with young companies and pointed to Coinbase. Everyone who bought back then was generously rewarded. Okay, so her words match her actions. That's right. Although there is a gap between the analogy and the balance sheet.
How? The crypto exchange is developing without significant physical assets. Coreweave needs to build a data center before it can generate its first dollar of revenue. You see, this is a completely different kind of debt. That 's right, and we can verify this with numbers. We will do it. Okay, let's look at them one by one.
Let's start with Coreweave . Of course. This is the largest position Ark has added this month, and the stock has already fallen more than half from its 52- week high. Remind me what this company does. It purchases Nvidia GPUs on a large scale and rents out that computing power, creating a cloud exclusively for AI.
Nvidia itself is the second largest shareholder . How big is it? The active capacity is 1.5 gigawatts . Taking into account contracted capacities, it is about 3.7 gigawatts. Hmm, you measure it by power. For a data center, how much electricity can be obtained is the proposition. Active capacity has increased by about 500 megawatts over the past year.
Who are the customers? Large AI labs and hyperscalers, mostly. They book capacity with long-term contracts, so every deal is huge. That's why the order book is so large. Of course. The downside is customer concentration. Losing one of them weighs a lot. How was the last quarter? This is the June quarter.
Revenue was $2.575 billion, up 112% from $1.212 billion a year ago. More than twice. Adjusted EBITDA was 1.51 billion with a margin of 59%. There is nothing to complain about in this indicator . By this indicator— I don't like how you said it. GAAP operating income was negative $49 million. A year ago it was positive—19 million.
So, revenue doubled, and the company went from operating profit to loss. Does it look better in the adjusted form? Adjusted operating profit was 128 million, margin 5%. A year ago it was 200 million at 16%. So, even with adjusted data , the margin fell by 2/3. Revenue doubles, margins fall. Depreciation, electricity, and interest, which we will talk about now, are growing faster than income.
So, it grows, but loses even more money. The net loss was $626 million, compared to $290 million a year ago. GAAP loss per share was $1.14. What is "eating" it? Interest. Net interest expense was $640 million . A year ago, they were $267 million. Wait. The interest alone is approximately equal to the total net loss.
This is the entire company in one line. The interest of $640 million is greater than the net loss of $626 million. The cost of debt is the loss. Will it get better? Management forecasts interest expense in the 3rd quarter of $860 million to $940 million. They are growing. Are they growing? And they are forecasting adjusted operating income for the same quarter of $200 million to $260 million.
Interest remains three times higher than operating profit . So what are the arguments in favor of buying this? The order portfolio is about $104 billion as of June 30. Plus over $25 billion in new commitments signed at the beginning of the third quarter that aren't in that $104. $104 billion? How many years of revenue is that?
The forecast for the entire year 2026 is 12.4–13.2 billion. So, for about eight years. Okay, that's a real number. CEO Michael Intrator said the quarter marked a significant inflection point as scale began to translate into operating leverage. Is a backlog of orders a guaranteed revenue? These are contracts, so they are firm.
But first you need to build a data center before you can provide services. The larger the order portfolio, the more cash is generated at the start. And, therefore, strength and weakness are one and the same fact. That's right. The annual capital expenditure forecast is between $35 billion and $39 billion.
And what do pessimists pay attention to? On the balance sheet. Total debt, excluding leasing , was $35.1 billion as of June 30. At the end of 2025, it was 21.4 billion. So, almost 14 billion was added in 6 months. In 6 months? Add 16.3 billion in leasing obligations, and you get 51.6 billion. Shareholders' equity is 5 billion.
This is a debt- to-equity ratio of 8.94. Almost 10 times the capital surplus . Capital expenditures in the second quarter were 6.4 billion . Operating cash flow—679 million. That is, about 5.7 billion of negative free cash flow is covered by new borrowings. What is included in this 35 billion? Recourse debt for which the parent company is responsible — 31.4 billion.
Non-recourse—3.7 billion. Almost all of this lies with the company itself. There's not much chance of just leaving. And the order of payments matters . If it comes to that, creditors and landlords are first in line. Shareholders are the last. Total liabilities of 72 billion against 5 billion in capital mean a fragile stake for shareholders.
How did the market evaluate this balance? Their bonds were trading well below par, at the level of distressed assets, and credit default swaps hit a record high. Shares fell from $77.81. on the 13th to $65.78. 14th and up to $61.39. On the 16th . And ARK bought it out? On the 14th and 15th, about 382,000 shares, worth about $25.8 million.
Then, on October 19, Goldman Sachs upgraded the stock to a “buy” rating. Wait, they raised the rating? At the same time, lowering the target price from $139 to $106. It's not because the company has become better, it's just that the shares have become cheap enough. What does Nvidia think? They own a stake in this.
Nvidia is the second largest shareholder, and Jensen Huang said that CoreWeave's cost of capital is decreasing. Which is the exact opposite of what the bond market is saying. Record high credit default swaps on one side, the largest client and investor on the other. Wood chose the other side of the issue.
The second name is Rocket Lab. They also bought into the bad news, and the third delay is serious. What happened here? On October 14, Rocket Lab announced that the first flight of the Neutron rocket is being postponed to 2027. This is the third delay. Because of what? Air pockets, voids, found in the forward dome that covers the first stage fuel tank.
This is a carbon composite part. Does emptiness mean it will fail? Beck said that doesn't mean the dome will break, but they won't fly without a full safety margin. The spare dome is already in production, and rockets number two and three are being built in parallel . In August, they announced that the degree would be taken to the site this year.
This plan, like the flight itself, was postponed until next year. Is there a new date? Not yet. Two stages will be completed by the first quarter of 2027, the third during it, and only then will they set a target for launch. So, no one knows when it will fly. And Neutron is the company's way of entering the medium-duty market.
Until this is resolved , stocks have a "ceiling." What does real business look like? Revenue for the second quarter was $234.1 million , up 62% from $144.5 million—a record quarter. So, it is growing. The order backlog reached a record $2.36 billion, up 137% year-on-year . The launch portfolio alone has over 90 missions .
What about the financial result? Net loss was $49.3 million, or 8 cents per share. A year ago, this loss was $66.4 million. Cash was $2.13 billion, compared to $829 million at the end of last year. What about the margin? GAAP gross margin is about 36%. 41.5%—non- GAAP. Adjusted EBITDA was minus 8.8 million compared to minus 27.6 million a year ago.
So, losses are rapidly decreasing. The forecast for the third quarter is between $250 and $265 million in revenue. They also closed deals to acquire Kinematic and Motive. And Iridium, right? The acquisition of Iridium has also been announced. This is a launch company transforming into a space infrastructure company.
And the shares were still sold? From $64.22 on the 13th to $55.25 on the 15th and $ 50.84 on the 16th. It was the worst week in the company's history as a public company. What was the 52-week high? 151 dollars. That is, approximately 66% lower than it. Market capitalization fell to approximately $32.8 billion .
And did they buy ARK here too? 233,482 shares on October 1. 168,633—on the 15th. About $25.6 million per month. And just before its first purchase, Rocket Lab signed an agreement with Japanese satellite operator Synspective. 20 Electron launches between 2028 and 2031. This is the largest commercial contract for Electron in history.
What do analysts say? In early October, Citi began coverage with a buy recommendation and a target price of $105, which at the time represented about 50% upside. This, however, was before the delay? Yes, so give a discount. But the second ARK purchase was on the 15th , already after a delay and after the stock fell.
They bought knowing this. The third name is Palantir, and that's what's being sold now. Yes. And not all at once. Three separate sales, gradual reduction . Is it being sold because of weak business? Everything is just the opposite. Revenue for the second quarter was $1.935 billion, up 93% from a year ago and 19% from the previous quarter.
This is an incredible neighborhood. Revenue in the US commercial sector was $764 million, up 149%. In the US public sector—809 million, an increase of 90%. Everything shows acceleration. GAAP operating income was $912 million, with a margin of 47%. Adjusted figure—1.194 billion, margin 62%. This is a company that makes a profit.
Can one big deal inflate quarterly figures ? Contract data says no. The total value of closed contracts is about $3.37 billion, an increase of 49%. The US commercial sector alone —2.13 billion, a 153% increase. How many deals were there? 220 deals worth over $1 million. 98 deals over 5 million, 73 deals over 10 million.
This is scale , not a one-time success. So, it is both growing and profitable. There is an indicator for this - the "rule of 40". Add the revenue growth rate and profit margin. A score above 40 is considered excellent . What about Palantir? 155%. That's right. This is a completely different league. Adjusted free cash flow— $1.22 billion, 63% of revenue.
Cash and short-term treasury liabilities—9.2 billion. The full-year forecast has been raised to 8.15 billion, an increase of 82%. Everything looks perfect and they sold it. Market capitalization is about $467 billion . The average target price among 32 analysts is $215, based on a closing price of $194.15 on October 19.
This is limited growth potential. So, the problem is the price, not the company. ARK did not disclose the reason, and Wood continues to publicly praise Palantir. There is also a structural aspect here. ARK limits the size of any single position. So stocks that are growing are cut, and funds go into those that are falling.
When do they report next? November 2. Revenue forecast for the third quarter is $2.16 billion . Let's compare all three. Coreweave leads in growth with 112%, but it is unprofitable. Palantir grew by 93% and is profitable. Rocket Lab grew by 62% and is also unprofitable. And the risks? Coreweave stands out.
Debt to equity ratio— 8.94. Third-quarter interest expense is forecast at $860–940 million on adjusted operating income of $200–260 million. Interest is three times higher than operating profit . And Rocket Lab? They have $2.13 billion in cash, so funding is not a problem. The risk is only in one issue.
Will Neutron fly? What about the assessment? Palantir is worth about $467 billion by market capitalization. CoreWeave—about 33.7 billion. Rocket Lab— about 32.8 billion. Palantir is in another universe. Note that CoreWeave's quarterly revenue is actually larger than Palantir's, but the market capitalization is less than a tenth.
If you're looking for a reason why ARK is cutting its stake in Palantir, start with this one . So what should you watch out for? Rocket Lab. The delay hurts, but the $2.36 billion order book and more than 90 planned launches haven't disappeared. The company's financial position is strong. The price has fallen, but the value of the business has not.
What about CoreWeave? How should we think about this? There is only one indicator: whether the third-quarter adjusted operating profit will cover interest expenses. If the gap narrows, Wood was right. If it increases, the bond market was right. November is a test. Does this mean that Palantir is not worth investing in?
Absolutely not. This is the strongest business of the three, but at this price, a good quarterly report may not have much impact on the stock. November 2 will show whether the situation will change. Let me summarize. During last week's AI stock sell-off, ARK invested about $51 million in CoreWeave and Rocket Lab combined and sold $ 16 million in shares of the successful Palantir .
Sell the winner, buy what fell. This is contrarianism in the most literal sense. Do your own research before you act. Thanks for watching. If this was helpful, please like and subscribe.
Comments 0