Embaud retweeted
모건스탠리, 삼성전자와 하이닉스 목표가 하향
근데 이번 보고서에서 더 황당한 건 2027년 이익 전망이 아니다.
2029년부터 2035년까지
메모리 기업의 장기 정상화 이익률 자체를 확 낮춰버렸다.
근데 잠깐만!?
내년 이익도 제대로 맞히기 어려운 판에
3년 뒤, 5년 뒤, 10년 뒤 메모리 마진을
어떻게 알고 지금 주가에 박아 넣냐?
메모리는 공급 투자 한 번, 기술 변화 한 번,
AI 수요 한 번에 판이 뒤집히는 산업이다.
근데 2035년까지 이익률을 미리 정해놓고
지금 목표가를 깎는다?
이런 정신 나간 놈들을 봤나.
장기 전망은 참고할 수 있다.
근데 10년 뒤 숫자를 정답처럼 넣어서
오늘 주가를 재단하는 순간부터
그건 분석보다 소설에 가까워진다.
물론 얘네가 맞다면 영웅이 되겠지만;
Embaud retweeted
“No Más.” — Roberto Durán
This was the week it felt like investors finally threw in the white towel on the AI trade.
The S&P 500 was down just 2.7% for the week, but it felt far worse. The real damage wasn't in the index. It was in the hedge fund world, where what looked like a capitulation moment triggered a cascade of forced selling across hedge funds, leveraged ETFs, and CTAs. The Nasdaq 100 (NDX/QQQ) finished down 5% for the week, with more than 100% of that decline occurring in just two hours on Thursday, between 8:30 and 10:30 a.m. New York time.
The magnitude of the deleveraging was extraordinary. Volatility spiked, put-call ratios reached extremes, trading volumes exploded, and hedge funds cut exposure at a historic pace. Hedge fund closure rumors went viral.
And I was buying into the panic.
In this week's video, I explain why I would try and catch a falling knife and why I believe we're witnessing a market structure deleveraging event rather than a breakdown in AI fundamentals or the economy. I go through the data behind the forced selling, the factor signals I'm watching for exhaustion, and why the midterm results may have actually reduced the risk of an AI bubble.
I also explore the changing market structure, the emerging importance of factor rotations, and why declining liquidity headwinds could be a significant positive for AI, Bitcoin, and gold.
The AI revolution is a long-duration investment opportunity being traded in an increasingly short-duration market.
Investing is becoming a counterpuncher's game. The market structure has changed forever and hedge fund leverage has forever peaked.
Watch the full weekly update:
youtu.be/LJo0GGo3Sc0
Embaud retweeted
You are showing DDR4 ram, which nobody use anymore. See ddr5 below. Absolutely no price drops at all.
Misinformation sadly
Embaud retweeted
I respect Jukan, but I feel a little frustrated whenever I see this kind of argument.
You can build any scenario around plausible assumptions, and yes, of course, the outcome will follow from those assumptions. If you assume demand collapses, customers accumulate inventory, and contract terms get renegotiated, then naturally, memory companies’ revenue and profits will decline.
But honestly, this feels like the perspective of analysts who treat memory as just a commodity you manufacture and sell, without properly understanding the technological changes taking place in custom memory.
Who exactly are these customers we are talking about when we say they could cancel LTAs or defer deliveries? They are the very customers working with memory companies to design the next generation of HBM and 3D stacked DRAM.
After jointly designing and validating these products, can customers really cancel those volumes so easily or push deliveries out by one or two years? How would they manufacture their own products, which were designed around that memory? What happens to the launch and system deployment schedules they have already coordinated? Switching suppliers also raises questions of design compatibility and requalification. If you are going to argue that contracts can be adjusted, you also need to explain what that decision would mean for the very customers canceling them.
This is one of the biggest reasons I keep saying that memory is no longer the same cyclical industry it used to be. The parties to these LTAs are already tied together from the design stage. But if you still view memory simply as a commodity where rising prices lead to more production, and inventory buildups lead to price cuts or canceled orders, you will inevitably miss this shift in the technical and business relationships.
One more thing. These arguments about a collapse in memory demand seem to assume that planned AI data center investments will be canceled on a massive scale, or that AI infrastructure expansion will slow so substantially that memory demand falls sharply. What interests me more than whether contracts can be adjusted is why you expect demand to weaken that dramatically in the first place.
I think it would be more realistic to discuss directly whether the AI era we expect will actually fail to materialize, and whether much of the AI infrastructure investment currently planned will disappear. If that is your outlook, I can at least understand why you would expect memory demand to collapse and customers to cut volumes even for products they have jointly developed.
I'd like to believe LTAs are durable too. But take one example.
Analog chipmaker onsemi said this in 2023.
On January 5, 2023, at JPMorgan's CES forum, CEO Hassane El-Khoury was asked how much LTSAs would still hold once the shortage eased. His answer: “They are a legally binding agreement ’cause they’re a liability for the customer because they’re take or pay.” He also said he would not accept a customer reopening price talks just because it could get the same parts 10% cheaper elsewhere.
On October 30, 2023, on the Q3 earnings call, onsemi cut its SiC shipment forecast for the year from $1 billion to more than $800 million after one automotive OEM reduced demand. An analyst asked how this could happen when the company had been saying its LTSAs were legally binding. The same CEO reaffirmed that they were binding, but said he did not want to enforce the contracts just to push inventory onto a customer whose demand had dropped. If it worked for both sides, terms could be adjusted customer by customer, whether by a quarter, a year, or longer.
In its 10-K filings, onsemi also states that it amended LTSAs with some strategic customers by mutual agreement in 2024, and changed delivery schedules and committed volumes on some LTSAs in 2025 as demand shifted.
Microchip is another case. Customers in its Preferred Supply Program (PSP) originally had to place 12 months of orders and could not cancel or reschedule them unless Microchip raised prices. On its November 2025 earnings call, however, the CEO said the company had accepted some cancellations and pushouts so customers could adjust their backlog. On the same call, the CFO separately addressed the multiyear LTSAs the company had signed. Some of those contracts were still in effect, he said, but customers had been allowed to defer taking delivery of committed volumes by one to two years if they needed to.
Every time I see cases like these, I doubt how much memory makers' LTAs will protect earnings in a downcycle. Right now the most reliable thing is the prepayments and deposits themselves. That cash can go into capacity expansion today.
I can picture hyperscalers with no DRAM inventory continuing to buy at current prices. But once inventory has piled up, will they keep taking every committed unit at the same price? I have a hard time picturing memory makers holding customers to the high contract price all the way through.
The LTA itself may never be broken and may stay in place. But if both sides agree to push this year's volumes into next year or two years out, this year's earnings estimates come down. If price gets renegotiated too, margin estimates have to come down as well.
Again, I agree LTAs are sturdier than they used to be. What I question is whether, after demand turns, memory makers can still earn 80% margins on contracts signed during the upturn and keep collecting hundreds of billions of dollars in profit because of those LTAs.
Embaud retweeted
SemiAnalysis is bloody good at figuring out where compute is scarce. Probably among the best out there. But then they zoom into individual names and sometimes seem to forget that knowing where the scarcity is and knowing who actually collects the rent are two very different things.
You can benchmark every GPU, obsess over Kubernetes configs and dunk on $IREN's legacy infrastructure all day. Fair enough, uptime matters. But you're not underwriting a SaaS startup here. You're underwriting a capital intensive business where there are lots of moving parts that can matter just as much as who has the prettiest dashboard.
And the irony is their own macro research increasingly supports this. GPU rents rising, financing slowing marginal supply, existing capacity gaining pricing power.
Meanwhile $IREN is signing contracts with 90% of GPU capex prepaid by customers, recontracting old H100s at twice previous prices and owning the land, power and datacenters underneath.
Doesnt mean the equity is automatically cheap or execution risk disappears. But jesus, at some point you have to connect the technical analysis to the balance sheet because the best GPU cloud isnt necessarily the best investment. And the worst Kubernetes experience isnt automatically the worst owner of scarce megawatts.
TLDR: these guys are good at spotting industry trends, but don't let them pick your portfolio companies because that is a different ball game and they are not up to that task imho
Embaud retweeted
Crypto was never about meme coins.
The death of meme is the birth of real world usage.
Memes were a liveness test run by cabals and larps, who are all out.
Crypto rails will run the $1 trillion stablecoin remittance and payment economy and $2 trillion tokenized equity and commodity economy.
Memes were a liveness test.
Thanks for testing the networks and the internet.
Embaud retweeted
When the masses analyzing compute demand don’t understand the kind of compute necessary to serve basic enterprise software at CSPs it makes sense they don’t understand what’s required to serve enterprise AI workloads at scale.
Embaud retweeted
Microsoft was renting A100s from Oracle for powering Bing and they won't renew that contract
Oracle probably wanted a premium on those A100s and MSFT walked away
Overreaction tbh
Microsoft has committed $80B+ to rent third-party AI compute, according to Bloomberg. Recent reporting and company disclosures show several changes across that portfolio:
>Oracle: Microsoft is winding down its 2023 Bing compute agreement after declining to renew. Older NVIDIA GPUs reportedly no longer met its specifications. The contract’s value wasn’t disclosed.
>Nscale UK: Microsoft canceled a ~$3B Loughton agreement after grid connection delays pushed expected power availability into the 2030s.
>Nscale Texas: Microsoft declined an expansion worth tens of billions of dollars. Its existing Texas contract remains in place.
>Nscale West Virginia: Microsoft stepped away from a letter of intent for 1.35GW of capacity.
>IREN: Its $9.7B, five-year Microsoft contract covers 200MW of IT capacity. Deliveries have been delayed following Microsoft-requested network changes and supplier timing. The contract remains active, and IREN says the delay carries no financial penalties.
These developments raise questions about how tight supply remains for particular GPU generations, how much external capacity Microsoft will need, and what rental income and residual values older chips can support after their initial contracts expire.
The evidence on older GPUs remains mixed: Oracle says demand is strong and reported renewing four-year-old GPU contracts at a 20% premium. Recontracting prices and utilization will help establish how broadly that demand holds.
Embaud retweeted
Always met someone with the exact story that I am trying to pitch
Today I met with a veteran investor who argued that memory stocks are not cheap even at a P/E of 3. Below is an edited version of our conversation.
Memory LTAs are basically meaningless. The best the LTAs being signed today can do is get customers to shoulder part of the memory makers' capex. Once prices fall, customers will gladly forfeit their deposits (prepayments) and walk away from the LTAs.
Historically, whenever prices fell more than 60% from the peak, customers broke their LTAs even after signing them.
So even if a 2027 shortage is certain and the stocks trade at a P/E of 3 on 2027 operating profit, 2028 operating profit is uncertain, which makes it hard to justify a higher valuation.
He went on to argue that even if 2028 operating profit were certain, we would still have to account for 2029 operating profit.
I pushed back and asked why the market doesn't discount TSMC the same way. Why does TSMC get such a high valuation? Wouldn't TSMC also take a big hit if the AI boom fades?
His answer: the problem was how violently memory prices rose in the first place.
If memory makers had raised prices in a healthy way like TSMC, it might be a different story. But because they hiked prices like a commodity, the market has no choice but to treat memory stocks like commodity stocks.
Even if memory makers stay in shortage through 2030, he said, the market will price in the risk of oversupply in 2031.
He closed his argument this way: even at 3x 2027 earnings, if memory prices fall in 2028 by enough to break LTAs, that 3x multiple on 2027 works out to 7.8 to 10x on 2028. So current valuations are not cheap.
Embaud retweeted
Data centers in Loudoun County, Virginia: “We get $1.2 billion a year in local tax revenue, about 40% of all of our tax revenue...on 3% of our land”
Buddy Rizer, the county's economic development chief, explained how “data center alley” grew:
“We would have never looked at having 55 million square feet of data center, 230-some buildings. That was not the plan. The plan was to utilize the fiber loop, the internet exchange, to utilize a reclaimed water line, and buildings that were already built. That was the plan, right? And so that made a lot of sense.”
“But, you know, fast forward, today we are largely dependent on that. We get $1.2 billion a year in local tax revenue, about 40% of all of our tax revenue. And while that looks overwhelming, data centers take only about 3% of all of our land in the county. So it’s quite a return on investment. You know, 40%, $1.2 billion, on 3% of our land.”
SpaceX’s Arizona data center, per a county supervisor: “We could be reducing property taxes across the board in Cochise County, the entire county, by 20%”
Frank Antenori, whose district includes the site, brought up the county’s tax math during a live Q&A:
“Here’s just something to just chew on. Arizona has an expenditure limit requirement that is placed on counties. If we exceed that expenditure, we have to rebate back to the taxpayers a certain amount.”
“So, if we’re receiving revenue in from SpaceX, by law, we have to offset that to the taxpayers. We have the potential to reduce, based on the projected revenue they’re creating, as much as 20%.”
“Let me just say it again. We could be reducing property taxes across the board in Cochise County, the entire county, by 20% to offset for the revenue here, because we can’t spend all that money. So, we have to give that money back to the taxpayers. The law says we have to.”
“So, you could see a potential significant impact to local property taxes. What that is, I don’t know. We’re going to have to do the math on it. But the law, if we have this large increase in property tax revenue, we have to balance those rates. So that will reduce dramatically the property tax burden of every resident in the county.”
“The construction sales tax, which also is the TPT, the transaction privilege tax, the retail tax, and the jail excise tax. So that will be part of the construction tax. We could literally use that money to pay the jail off in 2 to 3 years, as opposed to the 13 to 20 years.”
“So this could be a revenue to eliminate the half-cent sales tax in 2 to 3 years. We can get rid of that half-cent sales tax.”
Embaud retweeted
$IREN
Remember the sell-off in IREN from around $48 down to $17?
A big part of the pressure started after @SemiAnalysis_ put out critical posts about reliability issues at the Prince George site, saying customers were disappointed with multi-day outages, network problems, and other failures.
Well, those same Prince George H100 customers that SemiAnalysis highlighted as having major problems have since renewed their contracts with IREN at roughly double the previous price.
teapawt.net/Agrippa_Inv/status/211…
I’m fully aware that no serious investor takes @SemiAnalysis_ seriously these days, but just to add insult to injury:
Let me get this straight...
Semi gloated about the leak that $IREN was facing delays with $MSFT and immediately used it as another opportunity to throw shade at the company.
Except they apparently failed to realize that it was $MSFT that requested networking / topology changes from what had originally been agreed, which is what pushed part of the deployment into early next year.
$IREN Co-CEO @danroberts0101 was very explicit about this on yesterday’s call.
He also explained that accommodating a change of this magnitude so deep into construction would be much harder for most other neoclouds, because they rely heavily on third-party data center providers and therefore have another party in the middle whenever something needs to be changed.
$IREN, on the other hand, owns and operates the entire infrastructure stack itself.
That vertical integration gave the company the flexibility to accommodate $MSFT's last-minute request, and Dan made clear that the relationship between the two companies remains “fantastic.”
In other words, $IREN was flexible and agile enough to accommodate a major last-minute change in $MSFT's requirements, even this deep into the deployment process.
And somehow @SemiAnalysis_ found a way to spin that into a negative for $IREN?
This really shows how disingenuous this “publication” has become....
How were independent retail analysts with a near-zero budget in @jiahanjimliu and @FransBakker9812 able to figure this out WEEKS ago, while one of the largest AI-focused “research companies” failed to understand what was actually happening?
Don’t they have sources at $MSFT?
Apparently they aren’t quite as well connected as they like to appear...
Who in their right mind is paying these clowns for their coverage? 😂
Embaud retweeted
$IREN
@SemiAnalysis_ your assessment of IREN’s Prince George facility deserves some serious scrutiny.
You portrayed it as one of the industry’s worst facilities, plagued by reliability issues and poor infrastructure.
Yet IREN just recontracted older H100 GPUs at DOUBLE their previous pricing, on a multi-year deal. The previous contract was explicitly tied to Prince George.
If the facility is really that bad, why are customers willing to pay twice as much for aging GPUs and commit for years?
Technical criticism is ok. But ignoring actual commercial outcomes while delivering such a damning verdict is not.
Perhaps it’s time to reassess your conclusions rather than dismiss the evidence that challenges them.
Embaud retweeted
Okay, this is actually far more reasonable than the headlines made it sound.
Gina Hinojosa wants to let ALL already permitted data centers finish construction. And yes, NEW data centers will still be built, just under stricter rules protecting consumers.
She’s not trying to kill the AI buildout. She wants Texas to benefit from it without ordinary people footing the bill.
I can absolutely get behind that.
Embaud retweeted
$MSFT : “Hey $IREN, the specs we gave you last year need to be changed.”
Iren: “No problem, we got you homie, little delay.”
MSFT: “Thanks big dog.”
@SemiAnalysis_ : "OMG WE HEARD FROM OUR DADS EX CRACK HO THAT MICROSOFT BROKE UP WITH IREN BC HORIZON 2 BLEW UP. TRUST ME BRO"
Embaud retweeted
Getting more and more optimistic on @texasforgina
She has a Republican and a union electrician on her 4 person transition team.
And love it that she refuses to engage on culture war issues and is staying focused on making the everyday lives of Texans better.
Texas Governor-elect Gina Hinojosa: “I’m not going to be the governor for just Democrats”
She was asked at her first press conference why she’d go bipartisan after voters gave Democrats a mandate:
“Because this is about my campaign promises, what I ran to create as governor. We are forging a new way forward. People are tired of the divisions, of the fighting, tired of the culture wars. This is a time to come together in a way that is about just delivering for the people of Texas."
"Ernest Bailes, Representative Ernest Bailes and myself were close allies delivering for our public schools. In fact, I count him as one of my best allies for public schools. That is his commitment. There is really little light between our positions on public education, and as such, he serves as a tremendous leader on this transition team."
"And look, there are lots of issues with whom my good allies are Republicans. So I’m just about delivering for people. This is not about being a Democratic administration. I’m not going to be the governor for just Democrats. I’m going to be the governor for all the people of Texas.”
Embaud retweeted
$IREN says $MSFT requested a “network design change” across its Horizon sites and contract allows delivery extensions without delay credits for customer requested changes.
With the first site already live since August I see this more as Microsoft fine-tuning its rollout than backing away.
Embaud retweeted
Manyağın teki amerikan borsasindan 2 trilyon doları tek tweetle sildi.
Oy hakkı denen şeyin default olarak verilmesi delilik ✍️
