A MONTH later.... - Micron $MU : from $724.66 ---> $1,141.75 [+57.5%] 📈 - SK hynix : from €1,080 ---> €1,580 [+46%] 📈 - Sandisk $SNDK : from $1,407.61 ---> 2,200.67 [+56%] 📈 Sorry not sorry for the former Memory Bulls turned... clowns, and a real sorry for the Memory Panicans 😌

Great explanation about the selloffs in Biotech but can apply to other Sectors too 🧐 Remember that investing is not only about theoretical valuations but too what the Market assumes and likes/dislikes And pls., never never take leverage or invest in commodities 🤓

AMD $AMD sells the chips inside an AI deployment. How much of each new gigawatt becomes AMD revenue depends on what goes into the racks, what customers pay and when the equipment ships. Our new report and financial model are live: AMD Revenue per Gigawatt: What the Hardware Is Worth. The operating analysis is free. We follow the opportunity from customer deployments through Instinct GPUs, EPYC processors and networking, then into earnings and cash flow. Cooling systems, buildings and equipment supplied by partners sit outside AMD's invoice. A gigawatt also needs a definition: power delivered to computing equipment differs from total facility power. Our Base forecast puts AMD component revenue at about $24.9 billion per gigawatt of IT capacity in 2026, rising to $38.7 billion in 2030. Those are Northwise estimates built from GPU quantities, selling prices and attached CPU and networking content. They are assumptions to examine, not prices AMD has promised to realize. The shipment schedule matters just as much. Equipment installed in one year does not generate another complete hardware sale every year it stays switched on. We model new installations separately from replacement and upgrade purchases, and count customer workloads once when a cloud provider hosts them. By 2030, our Base case reaches approximately $388 billion of total revenue and $96 billion of free cash flow after cash capital expenditure. The chart shows the wider operating range: roughly $191 billion of revenue and $44 billion of free cash flow in Bear, versus $469 billion and $133 billion in Bull. Turning those sales into cash requires inventory, supplier commitments and customer collections to arrive in a workable sequence. The report tests lower GPU prices, higher memory costs and delayed deliveries, because a larger hardware market alone does not settle the economics for shareholders. These are conditional Northwise forecasts dated September 23, 2026, not AMD guidance. The free analysis includes operating forecasts, cash flow and stress tests. Valuation, expected returns, entry prices and the downloadable Excel workbook are part of Northwise Premium, which funds the research. Report link in the first reply. Happy to discuss the assumptions and where you disagree.

Macquarie: HBM revenue to outgrow DRAM in 2027. $MU $DRAM $EWY $SKHY

Advanced Packaging on Board✅

Now $NVDA loves Glass substrates too 😌 en.sedaily.com/international…

A must read. Phison CEO always delivers. Memory & Storage to moooon $MU $SKHY $SNDK

this is how we are optimizing InP lasers right now btw, adjusting the epi parameters like sliders, and wait months for the "frame" to "render"

Daily reminder for (New)Nuclear bois and of course special mention for SMR scammer clowns Time to build more Nuclear was 30 years ago. Good luck now with the HUGE money needed, shortage of specialized workers and inflation in materials

got to love a good sense of humor

4 years ago, Wiz was the fastest company to scale from $1M to $100M ARR in 18 months. In the same amount of time, Higgsfield AI just reached $1B. AI is not only coding, audio and video are growing at unprecedented speed. Higgsfield AI is enabling creators and businesses to produce high-quality, cinematic video from text and images. Their enterprise adoption is up 115% MoM, driven by an enterprise team of fewer than 20 people. This incredible growth has come with positive gross margins since the beginning of the year and 300% net revenue retention. On the other hand, ElevenLabs is best known for highly realistic text-to-speech, voice cloning, and AI-generated voice content. Since they shifted toward enterprise customers, they’ve landed deals with more than 60% of the Fortune 500. Revolut and Klarna have deployed their audio agents for customer support, reaching 40M customers and citing 8–10x faster resolutions. We are playing on a different level compared to traditional SaaS, but many seem to still ignore it

I don’t think investors fully understand the scale of what is happening here. 👉 Based on a WSJ story The AI buildout is on track to become the biggest economic bet in U.S. history. Not the internet. Not highways. Not electrification. Not even the railroad boom. AI. The WSJ estimates that U.S. data-center and AI infrastructure investment could reach $10.3 TRILLION between 2025 and 2032. That works out to roughly 3.6% of GDP per year. For perspective, the railroad boom averaged about 2.2% of GDP. Highways were roughly 1.1%. Telecom and fiber were also around 1.1%. That alone is incredible. But I think the charts underneath the headline are even more important. The five hyperscalers at the center of this buildout - $GOOGL, $AMZN, $META, $MSFT and $ORCL - are expected to spend roughly $4.2 trillion in the four years ending in 2029. That is no longer normal corporate CapEx. It is large enough to reshape the economy around it. And you can already see that happening in construction. Through July, private U.S. data-center construction spending was running about $9 billion above the same period last year. Meanwhile, private construction spending on basically everything else - houses, apartments, shopping centers and more - was about $46 billion BELOW year-ago levels. That chart is remarkable. Data centers keep going up while almost everything else rolls over. To me, that is where this becomes much more than an AI story. Hyperscalers are now competing with the rest of the economy for electricians, construction workers, transformers, turbines, land, natural gas, water and power. The WSJ gives a perfect example. Mississippi was in the running for an aluminum smelter that could have created roughly 1,000 permanent jobs. Then a data center was announced near one of the proposed sites and tied up electricity the smelter needed. The smelter went to Oklahoma instead. That is what crowding out looks like. And it is not just power. Data centers are also pushing up land costs, pulling skilled labor into the buildout and creating shortages across parts of the equipment supply chain. Then there is inflation. Import prices for computers, peripherals and semiconductors were roughly 20% higher year-over-year in August. That matters because the AI boom is not happening in some isolated corner of the economy. It is bidding up the price of labor, equipment, electricity and capital at the same time. And then there is the wealth effect. U.S. households now own roughly $63 trillion of stocks and mutual funds, nearly double the amount at the end of 2022. Think about the feedback loop here. $GOOGL, $AMZN, $META, $MSFT and $ORCL spend trillions on AI infrastructure. That spending boosts construction, wages and demand for equipment. AI-related earnings expectations push stocks higher. Higher stocks increase household wealth. That wealth supports consumption. Meanwhile, hyperscalers keep borrowing and spending to build even more infrastructure. That is an incredibly powerful cycle. But it also creates a risk that I think the market may be underestimating. The more the U.S. economy depends on one massive investment cycle, the more painful it becomes if that cycle ever slows. And a growing share of this buildout is being financed with debt. That is why I think this is much bigger than an AI bubble debate. The question is no longer just whether AI generates enough revenue to justify Nvidia chips or data-center leases. The question is what happens to the broader economy if the biggest infrastructure boom in U.S. history suddenly loses momentum. We are not watching another software cycle. We are watching the construction of a new layer of the U.S. economy. And increasingly, I think one of the biggest mistakes investors can make is treating AI as just another sector. AI is becoming the economy.

$3 TRILLION in Global CSP capex across 2027 & 2028. TrendForce: 68% of 2027 CSP capex goes to memory. But wait. Not one dollar transacted yet. Still 2026. Run the EPS math at 85%+ memory gross margins and try not to drool $MU $SKHY $SNDK $NVDA $TSM Holy Structural Shit

$TRT - Just got off a ~25 minute call with IR/management after earnings. I wanted to understand one thing above everything else. Why did Q4 revenue decline from $16.5M → $14.9M when semiconductor demand supposedly remains so strong? The key takeaway from the conversation is that this does NOT appear to be a demand problem. The sequential decline was discussed in the context of timing/revenue conversion rather than the underlying opportunity disappearing, which makes the backlog much more important. TRT finished FY26 with $24.2M total backlog vs. $11.0M last year, $19.8M of that is semiconductor backlog. And of the ~$14.2M of previously announced AI GPU burn-in board orders, only ~$6.3M had been delivered during FY26. That leaves roughly $7.9M still to be delivered. So the AI orders haven't disappeared. A significant portion simply hasn't converted into reported revenue yet. We also discussed margins. - Q3 gross margin: 15.5% - Q4 gross margin: ~19% That sequential recovery matters. My concern after seeing the FY numbers wasn't demand. It was whether TRT could actually make money from all this new volume. Based on the discussion, high-teens gross margins appear to be a reasonable way to think about the business going into FY27. Then there's Penang. $TRT is adding significant capacity and has already told investors that production volumes are expected to progressively increase beginning in Q2 FY27 as the expansion ramps. I specifically asked management to quantify how much additional production/revenue capacity the expansion could ultimately provide. They couldn't give me a number on the call, but they're going to come back to me with a number that we can use. That number is VERY important. Because if we can estimate the capacity increase, we can start properly modeling what $TRT could look like once the facility is utilized. So my view after speaking with them. Q4 revenue decline? Disappointing. Demand deterioration? I didn't come away from the conversation thinking that's what we're seeing. Backlog? Strong. AI orders? Still waiting to convert into revenue. Margins? Improving QoQ, but profitability still needs work. Penang? Potentially the biggest piece of the FY27 growth story, but I want actual capacity numbers before putting estimates on it. The market's question is legitimate. $TRT just grew annual revenue 72%. Why isn't that growth producing meaningful operating profit yet? That's what management needs to prove next. If the backlog converts, Penang ramps and margins continue recovering, FY27 could look considerably different. If those things happen and profitability STILL doesn't improve? Then we have a much bigger problem. For now, I think FY27 is about execution. And I am holding my position.

"2nm optics" sounds exotic, but the 2nm is mostly in the DSP > The real message behind “2nm optical” is less about the photonics itself and more about $MRVL’s DSP/SerDes capability. > Moving coherent/PAM4 DSP into 2nm while still hitting 400G/lane-class performance and aggressive power-per-bit targets is the important part. > It highlights that $MRVL’s moat in AI optical connectivity is not just photonics, but also the DSP behind it.