Knockouttrader
2.7K posts
Knockouttrader
@knockouttrader
Just a little goblin looking for the golden pot sharing information, discussing opinions, trading US and EU shares & options. No recommendations #DYOR
انضم Mart 2019
247 يتبع489 المتابعون
This the type of intel straight from the source that we need in this economy!
Financelot@FinanceLancelot
Stripper who works near several U.S. military bases says the young soldiers are super depressed & blowing all their money because they're being deployed next week.
English
Looks like the potential cybersecurity risks from Claude Mythos might be a bigger problem than expected, no bueno.
Took the L!
Knockouttrader@knockouttrader
Some sellside comments and key takeaways from RSAC 2026 caught my attention on a quiet morning so far. Overall tone: Cybersecurity is entering a winner-takes-most phase driven by AI. Customers are prioritizing fewer, broader platforms with real data advantages. Identity, data security and integrated SIEM stacks are the clear spend vectors, while many legacy tools risk automation and pricing pressure. Deutsche Bank notes AI security is a major driver but value may accrue to only a handful of leaders or remain fragmented. RBC highlights that differentiation is shifting from code to data, context and distribution. Wolfe adds that spending is still early, but already concentrating around platform vendors. $PANW Palo Alto Networks is emerging as the dominant consolidator. Strong platform breadth, accelerating SIEM share and a top-two SASE position give it direct exposure to every major budget shift. It is positioned to capture and consolidate AI-driven security spend as customers rationalize vendors, making it one of the highest conviction names in cyber coming out of RSAC. Discl.: Let´s give it a try in a tricky overall market - long some
English
Some sellside comments and key takeaways from RSAC 2026 caught my attention on a quiet morning so far.
Overall tone: Cybersecurity is entering a winner-takes-most phase driven by AI. Customers are prioritizing fewer, broader platforms with real data advantages. Identity, data security and integrated SIEM stacks are the clear spend vectors, while many legacy tools risk automation and pricing pressure.
Deutsche Bank notes AI security is a major driver but value may accrue to only a handful of leaders or remain fragmented.
RBC highlights that differentiation is shifting from code to data, context and distribution.
Wolfe adds that spending is still early, but already concentrating around platform vendors.
$PANW Palo Alto Networks is emerging as the dominant consolidator. Strong platform breadth, accelerating SIEM share and a top-two SASE position give it direct exposure to every major budget shift. It is positioned to capture and consolidate AI-driven security spend as customers rationalize vendors, making it one of the highest conviction names in cyber coming out of RSAC.
Discl.: Let´s give it a try in a tricky overall market - long some
English
„Israel: Strait of Hormuz closure planner killed“
Going down the supply chain already
English
What does this mean for Broadcom? $AVGO
Aaron@Arronwei3n
MediaTek TPU share from Deutsche Bank
English
Knockouttrader أُعيد تغريده
Canaccord putting some clarity into CLARITY news.
This is positive for $CRCL for sure.
I rarely want to argue with the market but sure seems like yesterdays move was to the opposite direction.
Recall they are the most regulatory-compliant stablecoin globally.
English
While reading a lot of comments about a potential "bloodbath" in HBM/Memory names today, here a short summary and my personal opinion:
Google highlights progress in making AI models more memory-efficient. This raises an important question for the semiconductor and AI infrastructure space: could improved efficiency reduce demand for high-bandwidth memory (HBM) over time?
In theory, more efficient architectures can lower the amount of memory bandwidth required per model or per GPU. That would suggest less HBM needed for a given workload. However, this view misses a key dynamic that has consistently played out in computing.
Efficiency gains are typically reinvested into scale. As models become more efficient, they are also deployed more widely, run more frequently, and expanded in size and capability. Larger context windows, multimodal systems, and real-time inference all increase overall memory demand, even if individual workloads become more optimized.
HBM addresses a specific bottleneck: delivering extremely high bandwidth to accelerators. That requirement does not disappear easily, and alternative memory solutions still face trade-offs in latency, bandwidth, or system complexity.
The more realistic outcome is that improved efficiency may reduce HBM requirements on a per-system basis, but total demand continues to grow due to the rapid expansion of AI use cases.
In the near term, HBM demand remains supported by scaling workloads and infrastructure buildout. Over the longer term, continued architectural innovation could moderate growth or shift how memory is used across the stack. This is not a binary outcome, but a gradual evolution.
But is this new? Sellside as well as Buyside already pricing in some reduced midterm demand so I am not sure about any potential impact all in all.
But nevertheless: Imagine you are that guy that already signed a binding multi-year memory contract with Samsung b/c you (thought you) need more HBM and then this one here drops 1 week later. Development is so fast.
$GOOGL $MU #Samsung #SKHynix
Google Research@GoogleResearch
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI
English
Deutsche Bank out with a Buy Initiation on MediaTek (2454 TT), PT NT$2,000 - calling it the “next ASIC winner” in AI.
Google TPU ramp and hyperscaler demand lead 17% sales / 23% EPS CAGR into ’28. TSMC access and lower-cost design puts MTK in prime position to scale fast.
Potential Read-through to $AVGO?
Broadcom still dominates custom AI (margins/IP moat intact) … but MediaTek is the first real share threat in years - especially on cost and capacity.
If TPU cycles accelerate, MTK isn’t just participating - it’s taking share. Nothing new, but another confirmation from sellside.
English
UBS´ take on $ASML / SK Hynix purchase filing:
The amount seems to suggest approximately 28 tools for 2026-27, compared to our existing estimate of 22 tools for SK Hynix. Assuming all other factors remain constant, this would represent 5-10% potential upside risk to EPS in 2027.
English
Reportedly Silgan is interested in acquiring Gerresheimer - press
$GXI.GY
English
Episode 2 is about "to learn how Yih-Shyan "Wally" Liaw delivered NVIDIA across borders"
$SMCI $NVDA
Supermicro@Supermicro
Watch the latest ServeTheHome video breakdown to learn how Supermicro and NVIDIA deliver turnkey, high‑performance AI infrastructure.
English
AI infrastructure isn’t going optical overnight. UBS with some thoughts on timing and adoption after Nvidia´s GTC
$NVDA made it clear:
Copper still dominates near-term AI GPU interconnects.
Co-packaged optics (CPO) is coming—but in phases:
-Rack-level first
-In-rack scaling later
-GPU-level adoption ~2029+
What does that mean?
The optical hype is ahead of reality.
Near-term winners:
Copper + packaging ecosystem
TSMC $TSM, ASE Technology Holding $ASX, Unimicron 3037.TT
Optical names may take longer to inflect:
Lumentum Holdings $LITE, Coherent Corp. $COHR, Innolight Technology $INLT
Long-term: optics is inevitable.
Short-term: copper still prints.
English
*US DIESEL TOPS $5 A GALLON AS WAR DISRUPTS FUEL SUPPLY CHAINS
When Bloomberg pushes this as a Hot Headline but we in Germany have to deal with around $9.50 / gallon for our Diesel :-D
English
$NVDA roadmap hints at a shift from GPUs to 3D AI processor stacks by 2028E (“die-stacking custom HBM”).
That means huge demand for advanced packaging:
• TSMC (CoWoS / packaging)
• SK hynix (HBM)
• BE Semiconductor Industries (hybrid bonding)
If AI chips go vertical, advanced packaging could become as critical as lithography.
Especially for $BESI.NA that means:
It is key beneficiary of AI’s move to 3D chip stacking. Its hybrid bonding tech connects dies directly (copper-to-copper), enabling ultra-dense interconnects needed for stacked GPUs + HBM. As NVIDIA pushes toward 3D AI processors, bonding steps per chip rise → more BESI tools.
Correct me if I am wrong here but future could be worse for those names above.
Jukan@jukan05
Die stacking!!! custom HBM!!!
English
@BreezeFor1407 Vera Rubin ultra has copper cables, and that‘s why you are seeing some bid up on $APH, $CRDO and other COPPER RELATED plays and concurrently selloffs in optics names
English
@knockouttrader what did he say in regards to optics (causing the sell off)?
English
So we are really playing the copper game, right Papa Jensen?
$CRDO $COHR $LITE
English
So it looks like <€150 on monday was a real gift. Feels like the market kinda overreacted with overall market sentiment being super tricky.
Yesterday´s feedback about sellside´s meeting with $BESI ´s CEO and now Reuters reporting BE Semi might be a potential takeover candidat ( potential acquirer $LRCX )
Stock now trading at nearly €200 in early German premarket session at low volume.
$BESI.NA
Knockouttrader@knockouttrader
$BESI.NA Time to make some gains back again today? UBS has an intg feedback from its call with BESI´s CEO call today: 1) Addressing the impact from relaxed stack height in HBM Post the news from Korea last week, which suggested potential further relaxation of HBM height restrictions, the CEO reiterated that all 3 memory makers were sampling the technology, with one likely to move into high volume this year. 2) Potential for Nvidia hybrid bonding plans announced at GTC The CEO emphasised the large opportunity if other GPU and custom ASICs suppliers were to adopt hybrid bonding in their designs, following AMD. Given Broadcom adoption of hybrid bonding at 2nm, there is a high chance that Nvidia shifts at a similar junction too (Feynman based on A16). Hybrid bonding enables chiplets, which are a means to improve the yield which will worsen as the nodes shrink. UBS believes there could be potential news on this front at Nvidia's upcoming GTC event 16-19 March. More broadly, TSMC currently sit at c60 hybrid bonding tool capacity and he reiterated the new AP7 fab to be more than double that in the next few years. UBS expects follow on hybrid bonding orders from TSMC in coming quarters, driven by Apple, Broadcom and Nvidia. 3) Continued confidences in the drivers of the mainstream recovery On the mainstream side, the CEO mentioned Q4 orders already included some for high end smartphone applications and he would expect more in Q1 given new camera content in the iPhone. For China, demand continues to be strong, especially for 2.5D applications, and Besi are currently sampling their fluxless TCB/Hybrid bonding tools for this region $AAPL $AVGO $TSM $NVDA
English