Shelly
121 posts
Shelly
@countshelly
engineering driven bottleneck pumping
Katılım Şubat 2026
248 Takip Edilen15 Takipçiler
Shelly retweetledi
This is my new 30% position.
Insanely well-made article that shows what an incredible opportunity $CGEH is.
I found this name, because I started getting worried about $NBIS running into capacity issues.
The 30-50% data centers will be delayed headlines made their rounds.
This pain is real, not FUD.
We're increasingly going to see the friction of digital demand running into physical world scaling constraints.
The pain is happening NOW. Which is why I think the feasible solutions will all quickly be sold out into 2030.
This company is ideally positioned and if they now get deals, it will be transformative for the stock.
Will share my thesis and thoughts in detail. But for now I recommend reading this article for people who prefer deep research, details and just want to first learn a ton about the company.
Longview Research@Longviewres
English
Everyone is searching for the next CPO bottleneck whilst ignoring one of the more obvious trades out there:
Chemicals.
It's the boring and unglamourous part of the AI infrastructure buildout, but one too many don't have on their radar.
You have:
1. $ENTG: The purity play.
Sits directly in the middle of the of $TSMC's Arizona fab.
-> A single contaminant can destroy a $30,000 package just like that. Just as $AEHR is critical to chips, $ENTG is critical to chips. It just hasn't moved yet.
2. $MKSI: The packaging chemistry play.
Supplies the specialty chemicals that enables advanced semiconductor packaging.
3. $ROG: System level materials play.
Indirect play on optical transceivers and high-speed switches.
4. $ESI: Direct advanced packaging play.
Most direct play on advanced packaging with organic growth now the highest since COVID.
I have a deep dive into the chemicals thematic out soon.
Make sure you're subscribed (for free).
English
@yieldchad who is even the target user for this slop?
people that want leverage but are too retarded to open an ibkr account and buy options?
English
$MU sold the chunk I bought in the low 300s, flipped it short
English
agree - plus an earnings surprise soon which proves memory is non-cyclical ($SNDK 2027 LTAs driven by kv cache offloads, i.e. deepseekv4) and this trade runs for a lot longer than even smart hedgies expect. don't wish for a drawdown this time, anon
Bucket Shop Capital@bucketshopcap
@ALEXEIMARTOV Tactically, I expect semis to start underperforming for a bit when people think the capex is starting to plateau…that is what is 6-12 months out, but it won’t be the end of the trade. Ppl will chase again as hyperscalers hike capex again. End of trade is 2029-30.
English
@UncleAlpha007 would usually agree but a 49% IRR in this market is just not that attractive
how would you live with the fomo?
multiple times now i've seen a small cap, decided it was too overpriced/greedy, played it "safe", then watched it run up 10-20x
English
Typically when it feels like everyone is going to get rich, it’s time to sell. Life doesn’t work that way.
Per Goldman’s data center report, I do think this is a once and a lifetime window to get insanely rich during this $7.6T build out.
The best strategy, in my opinion, is to stick with a concentrated group of names that are destined to win.
That’s why $SOI $SLOIF is my largest position. Sales go to $3B by 2029-2030, ebitda to $1.2B (25x multiple common) = $30B enterprise value. That’s 5x-6x from here.
Ren@Ren_aramb
Goldman just put a number on the AI buildout: $7.6T from 2026-2031. They explicitly flag optics as the next “buy out the store” chokepoint after memory. The numbers behind the thesis: +Compute alone is $5.1T of the $7.6T total +Annual AI capex grows from $765B in 2026 to $1.6T in 2031 +Data center cost per MW jumped from $10M (cloud era) to $15-20M (AI era) +A single $50K accelerator depreciates $10K/yr but goes economically obsolete faster than the schedule +NVIDIA’s GB300 NVL72 packs 72 processors per rack, linked by hundreds of thousands of km of cabling In the repot they state: similar episodes of intense, short-term pricing pressure are likely to recur across other critical components such as interconnect, optics, storage, and packaging.” Meaning the same dynamic that just sent memory parabolic is queued up across the entire physical layer of AI infrastructure. The “AI factory of the future” data center will pack 576 GPUs per rack at 500+ kW, requiring liquid-only cooling and millions of GPUs deployed at the >1 GW scale. None of that scales without optics. Memory just ran (still going). Photonics is just warming up. Full disclosure: Goldman is confirming what’s been playing since the late last year. A lot is already priced in. The alpha is front-running institutions, not reading their reports. But $7.6T over 6 years doesn’t get fully priced in 6 months. Plenty of runaway left.
English
My M4 Max MacBook gets 3,756,165 tok/sec in pure C, compared to ~50,000 tok/sec with the FPGA.
Try it yourself: github.com/AlexCheema/tal…
luthira@luthiraabeykoon
We implemented @karpathy 's MicroGPT fully on FPGA fabric. No GPU. No PyTorch. No CPU inference loop. Just a transformer burned into hardware, generating 50,000+ tokens/sec. The model is small, but the idea is not: inference does not have to live only in software 👇
English
@vikramskr @KeithBolognese What’s he’s really asking you is - what’s the next bottleneck that’s not already priced in
English
The entire AI value chain is one big mexican wave from one bottleneck to the next.
English
datacenters should maximize financial return; we should not waste any time or money making them look nice
Mike Bird@Birdyword
Many people do not seem to want data centres built near them, despite the fact that they don't cause that much traffic and often generate a lot of local tax revenue. I suspect it's partly because they're ugly! My proposal:
English
@awakenowzone It’s ran 140 percent in last month. Surely it reverts?
English
Here are $INTC’s valuation metrics compared to other semiconductor peers. When you look at Intel through this lens, it becomes clear that the stock is still undervalued across nearly every major metric relative to the industry.
This isn’t just a book-value story (though Intel is significantly undervalued on assets alone). The discount shows up in real operating metrics as well, EV/EBITDA and operating cash flow... P/E is obviously distorted due to massive capital expenditures tied to long-term growth.
If you believe in Intel’s story, don’t let naysayers or clueless analysts tell you otherwise. + Remember this is the current valuation before any real IFS income.
Intel (and LBT) may still have a long road ahead to fix the balance sheet and fire up IFS, and that’s exactly where the $1T upside comes from. But even ignoring all that, today’s valuation data says $INTC is a great deal.
Table ranked from best to worst by EV/EBITDA.
English
English
There is a real chance that for something like Figma you see very little seat loss because the product already existed mostly at the taste layer
And then you enable through AI models the ability for non-technical designers to also join the collab / network effect
There is a chance it expands the TAM honestly
English
@aminvesting5 @ContrarianCurse yes
it is a tool that helps the user create image files and other digital assets. it achieves this by giving the user lots of buttons they can click.
PROBLEM: the frontier labs made it really easy to output images without learning any of the buttons. the kids won't learn figma!
English
@countshelly @ContrarianCurse Two questions boss, 1. Have you ever used figma? 2. If yes, in your own words, explain in one sentence what it is?
English
