p909

@HayekAndKeynes What if you get the IPO allocation but then buy puts at another broker?

They are trying to get retail to hold the bag a bit longer Robinhood blocks future IPO access 60 days if shares sold within 30 days; Fidelity bans 6 months first offense, permanently third offense; Schwab restricts ~6 months first offense; Sofi blocks 180 days plus $50 fee for sales before day 120; E*Trade flags and restricts for set period

@OnodaCapital to make it short, one is renting out office space, another GPUs(though NBIS will have some most DCs on their BS), both are capital intensive businesses. Main dif is lifetime of the asset and how you value adjacent services. Not sure semantics matter, it they can turn a pos EBIT.

@hegen84 EQIX is not remotely the same business model jfc

Have only seen this level of retail replies on ASTS

@mgio90 @dampedspring Ehm…nope, there was no bubble in recent history 50 years), which popped with low P/E. In 2008 P/E was low, but I don’t remember anyone saying it was an equity bubble, it was a credit bubble. But you may have other examples I missed.

@hegen84 @dampedspring I see this nonsense reaponse repeated ad infinitum... Low Fwd PEs has been a sign of every major market top (and therefore bubble) in history. Equities top when earnings estimates peak and fw PEs look cheap.

Basic picture 1. Economy is strong due to wealth and capex in the rear view mirror and in the immediate future 2. Capex is a massive wave of funding needs which will compete with consumers sources of spending (selling their assets to dissave) 3. The GDP pie is not big enough for everyone to get the share they need to consume and earn investment profit 4. Equity asset values are in a bubble regime

@dampedspring don’t want to be the party pooper, but looks like the heavy lifting of your case is done by looking only at US gdp as source of profit growth. While it’s around 40-50% for the AI cohort. The current earnings projections still look aggressive, but not as much as you make the case

What do I mean with "it's a bubble " and there's "not enough pie" Enjoy dampedspring.com/wp-content/upl…

@degentradingLSD @lambdabraham Yes, exactly. They do the swaps with the banks and forget about it, they don’t manage the exposure themselves. Though now sure how their options exposure fits in here.

@hegen84 @lambdabraham look at the prospectus of 7709 - it is meant to be a passive vehicle.

@lambdabraham @hegen84 I dont think any of the swap counterparties are interested to warehouse 11B of SK HYNIX delta

@degentradingLSD @lambdabraham not sure how you want to manage your delta risk, if you are not delta hedging intraday, would create a massive basis risk. There are EWY and KOSPI 200 available for trading, probably OTC options/derivatives and stocks trading on SK too. But they def do not wait till KR open.

@lambdabraham @hegen84 None of the bulge bracket investment bank facil desks are allowed by risk management to run even 100m of outright delta risk for any period of time...which means if 7709HK unwinds, that flow will go through into the market.

@teortaxesTex RU had a pretty decent per capita GDP growth (2,5% pa), since 2014 (start of sanctions), on par with IDN, with BR hovering near 0% and SA and ARG contracting. EZ @ 1.2%.

Insane cope. the EU is stagnant and lame, yes, but they're stagnant and lame at 2-3x or higher GDP per capita. (No, Russian PPP per capita is not really $50K). For decades, since 1999 when he was PM, Putin's agenda was to "reach parity with Portugal". Long forgotten now.

@degentradingLSD Forgot to add, there are also KOSPI 200 futures exposure can be hedged with after KR markets close.

@degentradingLSD From what I see the fund has only 9% Sk Hynix of total NAV, rest is cash (collateral) so the exposure is gained via swaps, which the fund has with dealers. It’s likely that they already hedged some of the downside via DRAM, EWY today.

These are my calculations - let me know if this is wrong - otherwise we are looking at 2.4B of selling on SK Hynix just from 7709HK alone. Without including in all the other leverage within the KR ecosystem

@degentradingLSD EWY at 145 would imply a 45% move in KR semi stocks on Monday…unlikely. 175 is -15% already.

i think i might be crazy but i think $EWY prob hits 145 on monday.

@labubu_trader @AntonLaVay Actually if only Sk/ Samsung move and all else in EWY being flat that would imply at -17,5% move.

@labubu_trader @AntonLaVay Doesn’t EWY level imply Sk HYNIX/ Samsung should be down more than 20% on Monday?

牛逼！ 博主通过尾盘的判断，学到了。

@jukan05 Not sure if margins are sustainable, but that wasn’t discussed on td ER, wasn’t it? Otherwise it feels like Groundhog Day, we’ve seen almost same px action before/ after dec25 ER and Qs about margin sustainability.

Broadcom’s margin is around 60%, which isn’t even that far from NVIDIA’s margin. Let’s stop saying things like, “You need Broadcom’s design capability to compete with NVIDIA.” TPU v8i has already succeeded, hasn’t it? Do you really think Broadcom’s margin is sustainable?

@penguinvesting Not sure prepay rate can stay same, there are not that many cp rdy to pay in size (MSFT/ META), Ant and OAI don’t have the cash. Re dilution, think they will raise again via converts, which could lead to higher debt ratio, but lower IR than you model.

🚨My $NBIS 2027 Capex Estimate 👉I believe $NBIS will do between $50-60 billion Capex in 2027. Yes, that’s shocking, especially since it’s roughly their market cap. Let me show you how I got to the conclusion and what will be the implications. First, let me address that I believe guidance will come in at somewhere between $40-50 billion and they will likely tend towards the upper end by year-end with the possibility of a raise at some point as companies usually do nowadays. I like to be cautious, my calculation will assume $60 billion needed to finance, if they do less, that’s obviously less risk. Personally, I’d be very surprised if they don’t land in the $40-60 billion range. 👉What points towards this high spend? To put it simply, capex doubled in 2026 across the four major hyperscalers from ~$315 billion to ~610 billion in the first half of this year, with now guidance pointing towards $800 billion. A bunch of analysts and even $NVDA expect them to easily exceed $1 trillion next year and that seems to be confirmed by the CFO of $GOOGL who said they will significantly increase their capex next year from the current ~180 billion. Some analysts expect up to ~$300 billion from Google. While technically that’s a “deceleration”, I think it’s only due to the law of large numbers and we can expect easy doubles from smaller players. From this alone, we can see the overall trend, but let’s also look at a closer peer of Nebius, CoreWeave: $CRWV history with capex in the recent years: · 2024: ~8 billion · 2025: ~14.9 billion · 2026: ~31-35 billion (guidance) · 2027: ~21-62 billion (analysts’ estimates) Between 2024-2026 we see a clear doubling pattern similar to the hyperscalers. For 2027, analysts expect a wide range for CoreWeave but clearly, anybody under $40 billion assumes serious financial struggles thanks to their debt. All in all, I have no doubt that in normal circumstances a double is coming for $NBIS too. In fact, I think a better question to ask is, what will limit the spend? If the demand is out there, they sell everything they build, then why not spend as much as you can? This leads to my next point: 👉Capex is limited by 1. Infrastructural bottlenecks (like power and components) 2. GPU and other hardware procurement bottlenecks 3. Workforce bottleneck 4. Time bottleneck (permits + physical build out) 5. Capital bottleneck We don’t have clear visibility into the first 4 bottleneck, therefore I cannot estimate them. However, taking into account their 3.5 GW already contracted power, their expansion plans in Asia, and their overall diversification across continents, I think it’s more likely they’ll be able to navigate it for quite a while. However, what I can say is that capital likely won’t be a bottleneck up to $60 billion (more on that later). Keep in mind, other factors also influence the capex spend, like the price increase in datacenter components, and the higher price of the Vera Rubin GPUs that will represent more and more of the Neocloud spend as they become available in H2 2026. $NVDA also confirmed that building a 1GW AI Factory is becoming more and more expensive. Initially it was around $30B, now it’s rather $50B and it’s going to $80B soon. Of course, Neoclouds can pass these cost increases onto customers and therefore charge a higher revenue / MW. I generally think this is a theme we’ll see to continue. Higher capex than expected, higher revenue than expected. The current capex guidance of Nebius is $20-25 billion for 2026, so a natural double already puts us to $40-50 billion but that doesn’t account for anything I mentioned above, for example this year they only face a single-digit headwind in component prices due to early procurement, but it will affect them more next year. That’s why I believe spend will likely exceed 50 billion and rather tend towards 60. 👉How will they finance it In Q4 2025, they did blow us away by saying 60%+ of their initial guidance (16-20) was covered by prepayments and cashflow. Counting with the midpoint, that’s 10.8 billion covered at least. Note that, they expect only $3-3.4 billion revenue in 2026 with a 40% adjusted EBITDA margin. Why is that important? It’s important, because adjusted EBITDA is the approximate cash the business generates after counting for operating expenses (already existing datacenters). So, it’s basically the cashflow from which they can fund the capex. The 40% of the midpoint revenue (3.2) is only 1.28 billion. It means that roughly 1.3 billion or less comes from adjusted EBITDA and the remaining 9.5 comes from prepayments this year. In 2027, they will have a meaningfully higher revenue (at least $12 billion assuming they achieve their year-end $7-9 billion ARR guide). Adjusted EBITDA margin will also likely improve, for reference, CoreWeave is at 56% at the moment and was already above 60% last year. So, I think assuming a 55% margin on average for the FULL YEAR for Nebius is reasonable, and that puts us at 6.6 billion cash that can be spent on capex. That’s 5x more than in 2026, meaning, to sustain the 60% coverage rate of capex they’ll need proportionally less prepayments. However, I don’t believe we will have proportionally less prepayments for these reasons: 1. They confirmed prepayments are getting more and more popular and significant across all types of customers with some of them even paying 100% upfront 2. Meaningful increases in prices and newer gpu generations means customers need to pay more to lock in future capacity On the other hand, you can argue that as we go in time, it will be harder and harder to keep up the prepayment rate in relation to capex, simply because you need to fill in previous contracts (still spending money on capacity that was already paid for). This is a legitimate argument, however we are still in accelerating phase, so this shouldn’t be a big problem yet in my opinion, and the previous arguments I presented should completely counter it to the upside. Personally, I would not be surprised, if prepayment rate stays the same or increases in addition to the 5x more adjusted EBITDA. However, to stay conservative, I will calculate with the same 60% capex coverage rate they have this year. (that implies prepayment rate decreases slightly) This means, 60% of 60 billion is 36 billion which will come from prepayments and adjusted EBITDA (the 6.6 billion I calculated). So, to simplify, this assumes we’ll get approx. $30 billion prepayment and ~6.6 billion generated by the business. 👉How do we finance the remaining 24 billion? Let’s look at the subsidiaries first. Ofir Naveh said the following: “Most importantly, we truly believe that these stakes will significantly increase in the midterm. When they will do so, they will enable us to continue growing our business in a hyper-based mode. As we plan 2027 and beyond, while using this capital and continue keeping a very disciplined balance sheet. We're really happy with this potential capital injections for the future.“ To me, this suggests they will likely not sell the majority of the stakes yet, but probably ClickHouse will double again in 2027, so it’s hard not to assume they will trim at least bit. I expect them to sell and trim the other subsidiaries faster or more aggressively. For these reasons, I assume they’ll raise only $2 billion in 2027 from subsidiaries, leaving the majority of ClickHouse untouched. That leaves us with $22 billion remaining needed to finance. 👉Debt vs Equity for the remaining part to finance They’ll definitely do a combination. Personally, I think Nebius will be easily above $300 in 2027, which implies a $76 billion market cap. Let’s round that up to 80 for simplicity. This means they can raise $6 billion by diluting 7.5%. I think 7.5% dilution is COMPLETELY ACCEPTABLE and is EXPECTED. Personally, my models for 2030 assumed way higher dilution. So, if they decide to dilute more, I think that’s still easily acceptable and will leave them with less debt, but for now, I’ll count with a 7.5% dilution, and that leaves us with 16 billion remaining. In this case, they’ll need to take on $16 billion debt, which puts them at a total of ~24 billion long-term debt. That might seem scary at first, but it’s not really, considering their revenue in 2028 will easily exceed it by a wide margin. (based on their recent comments on H1 2027 capacity) Again, for reference, $CRWV is already at ~24 billion debt with a revenue guidance of ~12.5 billion. Even in this case, Nebius will have less than half of the leverage of CoreWeave and that doesn’t account for additional debt CoreWeave might take on this year. If we get to $24 billion debt, and we assume the average cost of the debt to increase to 4% from the current cost which is less than 2% on the convertible notes, we still get less than $1 billion interest expense per year. In 2028 their revenue will likely be above $30 billion. 1 billion will be only a 3% expense in relation to revenue. Last quarter, it was ~15% for Nebius, and ~25% for CoreWeave. To me, this doesn’t seem that scary, nor the 7.5% dilution. ✅Summary: In my opinion, they will be aggressive with capex, because demand is real, they have the backlog and they are in a financially healthy position to start catching up to $CRWV and set their path to become a true hyperscaler. And that’s why I accounted for $60 billion. I believe the scenario I presented to you is conservative, because it’s likely that the increases in prepayments, the Vera Rubin GPUs’ better economic profile, and the price increases in older generations will rather improve the situation. But again, to stay conservative, I assumed a similar situation as this year and this showed that even such an aggressive spend would come with fairly low dilution and a manageable debt profile. And if we can get cashflow and prepayments to cover anything above 60% of capex that would be ULTRA bullish in my opinion. Not to mention the case if the stock gets to $400 and/or they dilute 10%. That would meaningfully decrease the required debt, leaving us with an even healthier balance sheet. And finally in 2028, they can start tapping deeper into subsidiaries to continue this mind-boggling growth.

@Lazarus_Capital they make it sound as if this guy was Bezos’ right hand man, looking at his LinkedIn he looks more a like middle manager.

Anything that helps $IREN improve whatever they call their cloud now should be celebrated. Should’ve done this year but hopefully better late than never. Watch for the improvement on the bottom line, all that matters. Needs a massive step function improvement in pricing to be competitive.

@Lazarus_Capital Think Anthropic is not an easy client, they wouldn’t do prepayment( don’t have the cash) and don’t have good credit like HS. Guess they would pay up, but any contract with them would result in more dillution vs META/MSFT contract.

Keep pumping it anyway!!! Gotta sell my $IREN subs somehow

@Lazarus_Capital @danroberts0101 Think acquisitions r not easy for them. There is little cultural fit and the Iren C- suite clearly lacks technical acumen, any integration might be tricky/ costly. It’s a big Q for me y they abandoned the colo route, which would have been perfect for them.

I know I tease the $IREN cult a bit, but in all seriousness, this is sad. It’s still the same “land and power” thesis. @danroberts0101 is highlighting their Mirantis acquisition for orchestration when peers are already building out their inference capabilities ($NBIS $CRWV). Does IREN have potential? Yes. They have a massive amount of power and the ability to bring data centers online. However they decided to focus on the cloud without any technical expertise and that’s why they’re having trouble selling compute. YR ARR target missed. Q1 $500m ARR target (that bulls said was sandbagged) missed. ARR guidance for the full year revised down, twice. Returns are getting worse and they’re trying to hide it behind headline ARR numbers and “land and power”. If they want to be a serious contender, they need to improve their technical capabilities significantly. Acquiring an orchestration provider in May 2026 is a joke. Especially if you’ve already labeled yourself as a serious cloud player. Can they? Yes. Through acquisitions is one path. They’ve already been diluting the crap out of shareholders, they should at least use that to acquire talent to help with their cloud. It’ll be a wake up call for shareholders but it’ll improve the company. They really do have potential. If you disagree, tell me exactly where I’m wrong. Explain to me how this amazing vertical integrated company is only guiding to $3 per B300 hr. It’s actually less than $3 but close enough.

@EndicottInvests It’s a colo deal and delta one was supposed to deliver the power/ engines. It is reasonable to assume that they either will reduce their rate or compensate NBIS otherwise. We don’t know the margin impact, but it’s for sure not the full BE rate. Unlikely, but it could even be pos.

Here is my read on this $NBIS x $BE deal: It's neutral. Why? The bull case is real. Vineland was stuck in NJ DEP air permit review for the Bergen reciprocating engines. The first batch was supposed to activate in April. It didn't. Fuel cells skip combustion, which means virtually no air pollution, minimal noise, and a much lighter permitting path. This unsticks Vineland and de-risks the end-of-year ARR guidance. Bloom delivered Oracle's Abilene system in 55 days, so the time-to-power is real. The bear case is also real. This is expensive power. Up to $2.6B in aggregate service fees over three 10-year phases for 250MW guaranteed / 328MW installed. The structure is opex, not capex (Bloom owns and operates), but a lease liability likely shows up on the balance sheet. The margin question is whether $NBIS can pass the power premium through to customers. In the current demand environment, probably yes. In a softer market, it gets uglier. Net-net: necessary deal, good stopgap, not a long-term power strategy though.