Feed

$GOOGL is now discussing deals to extend their memory LTAs through 2035 from KeyBanc channel checks. Samsung already has 70% of their capacity allocated through 2031, and this move would suggest a more durable cycle for $MU, $SKHY, and the others. They also expect HBM prices up 80% to 100% in the Q1 2027. Having all your capacity potentially allocated out into 2035 via LTAs would ease a lot of memory durability fears. 595 likes · 58 retweets · 48 replies · 2026-10-15T20:00:10.000Z\ Captured by scrape-follows from @aleabitoreddit

SHOCKING: We have confirmed that multiple Neoclouds have been refused access to GTC Berlin due to them adding AMD GPUs, TPUs, and other chips to their cloud offerings Note that this doesn't apply to hyperscalers, as they have more buying power. But for neoclouds, executives feel NVIDIA has been using high-pressure tactics to keep them NVIDIA-only. As such, neocloud executives are starting to consider offering TPUs or AMD GPUs quietly to avoid getting pressured by NVIDIA and to avoid the feeling that NVIDIA will retaliate.

• $TSM reported 3Q26 revenue of $46.72B, up 41.2% YoY and 16.2% QoQ. - Net income of NT$763.06B and EPS of NT$29.43 each rose 68.7% YoY. - Gross margin reached 66.4%, with operating margin at 58.7%. - Wafer revenue mix was N2 8%, N3 33%, N5 28%, and N7 10%. - HPC revenue grew 22% QoQ and represented 68% of total revenue. - 4Q26 revenue guidance of $52.3B-$54.3B implies approximately 14.1% sequential growth at the midpoint. - Fourth-quarter guidance calls for 65%-67% gross margin and 56.5%-58.5% operating margin. - The 2026 CAPEX plan remained $60B-$64B. - Management expects annual N2-family capacity in 2027 to more than double from 2026. - Management described CoWoS production at 5x reticle size and development toward 9x and 14x. > Dollar revenue growth of 16.2% outpaced wafer shipment growth of 3.2%, suggesting contributions from higher-value process and packaging mix, but the gap cannot be interpreted as wafer price inflation. > Given management's attribution of the third-quarter gross margin decline primarily to the initial N2 ramp, the immediate indicators to watch are yield and cost improvements alongside N2 revenue share. > Subtracting nine-month CAPEX of $40.64B from the annual plan implies fourth-quarter spending of $19.36B-$23.36B, a condition for meeting that plan rather than separate quarterly guidance. > N2's nanosheet transition increases the importance of etch technology that controls complex transistor geometries, suggesting opportunities for $LRCX's GAA-capable tools, although this announcement confirms no individual equipment orders. > CoWoS-L connects logic and HBM through redistribution layers and local silicon interconnect chips, so the economics of larger packages depend on both interconnect area expansion and assembly yield. > At constant wafer area and yield, larger interposers reduce the number of packages produced, so CoWoS expansion should be evaluated through actual good-package shipments rather than monthly wafer throughput. > As screening defects before assembling expensive logic and HBM becomes more valuable, $FORM's known-good-die probe cards could see indirect demand opportunities, although these earnings establish no specific supply agreement. https://pr.tsmc.com/english/news/3346… 3 likes · 2 retweets · 2 replies · 2026-10-16T21:30:17.000Z\ Captured by scrape-follows from @PhotonCap

A small step back to take a objective look. As @pequityresearch showed, demand is essentially being modeled as infinite, it’s not nebius alone and that’s not healthy. I’m bullish on AI. You’re bullish on AI. Almost everybody is bullish on AI. But sometimes we need to go back a little bit to the foundations. At some point, there will be a discrepancy between the demand being modeled and what customers actually want, can afford, and are willing to commit to. I was on a call today with somebody who has extensive experience in large-scale compute deployments. We talked automation, clustering, financing, the whole lot. And the concern I’m about to point out was his concern also. It’s what he’s seeing in the current market. I’ve talked about systemic risk before. Currently, you see more B-tier companies coming into AI neoclouds, but mostly into AI infrastructure, all wanting to enter the gold rush. Inexperienced management teams borrowing at unhealthy rates against contracts signed with B-, C-, or even D-tier customers to get the compute going. As this part of the ecosystem grows, I think the risk to the whole ecosystem grows also. Gartner recently forecast that over 80% of neocloud providers will cease independent operations by 2030. That includes the possibility of consolidation. A lot of people immediately relate that to the Nebiuses, IRENs, and CoreWeaves of this world. But my concern here is especially the large group of upcoming companies underneath, all fishing in the same ecosystem, using the same suppliers, approaching the same lenders, and chasing the same customers. The weakest link matters when it’s connected to the next one. If one company’s problems are enough to make the next company wobble really hard, the flywheel supporting the AI buildout can start working in reverse. This is not me being bearish on AI. It’s taking a realistic view of the landscape and what’s developing there. We will have to see how many projects are being financed on the assumption that demand will simply be there once they’re ready. There’s a difference between securing strong customers before bringing capacity online and sitting on compute while still looking for somebody to take it. That difference says something, especially when financing costs are already running. Then you also have multiple companies building around the same pool of potential customers, all with ambitious scaling plans. Look at MeUs, for example. Great tool, great use case, huge growth. But it competes for a pool of customers that other companies are trying to reach also. And they all have ambitious scaling plans, because that’s the key to their business: reaching scale. We have to be careful about assuming everybody achieves those plans at the same time. Meanwhile, the financing costs are real. If customers take longer to arrive, need less capacity, or cannot pay, infrastructure providers may accept lower-quality contracts just to get a piece of the pie and keep capacity occupied. That compounded pressure can move through the ecosystem and create systemic risk. I’m still very bullish on the longer-term AI buildout. But that doesn’t mean every company entering it has a viable business. And modeling for infinite demand is not healthy, in my opinion. It won’t be the big names that will shake the ecosystem if it will ever happen. It will be a wave of B/C tier companies. $NBIS $CRWV $NVDA $IREN