Thematic Investment Research on the AI Buildout
Not Financial Advice
Joined December 2024
- Tweets9.2K
- Following327
- Followers15.3K
- Likes11.2K
OMG. MOONSHOT USES $CBRS.
Moonshot just dropped Kimi K3.1 Pro. We used it to speed run online shopping.
One prompt → monitor found → checkout → order confirmed.
11.5 seconds.
Watch @Kimi_Moonshot K3.1 Pro in action, accelerated by Cerebras.
I don't need to say much...
A few concentrated bets that matter are way better than holding 20+ companies and not knowing a single one to the core.
@kevinxu maybe you should take notes.
$APP
$APP is one of the BEST stocks to BUY at current levels.
The business will scale even more due to Artificial Intelligence.
Recent rotation into other more attractive plays has also made it go low.
Don't see how this can't double by the end of this year.
Won't keep repeating myself.
we're so dumb that we think US is the only source of demand for data centers.
while it is the biggest, it doesn't mean that AI stocks won't keep going brrr due to the rest of the world's demand
everyone needs this technology, and if US falls behind, which is clear they won't, then others will come in, which they can't afford right, because the ego is so high always to be the #1
even if Jensen comes tomorrow and says there is going to be demand and we shouldn't worry about all this, it will still keep extreme fear in the market the way it is
i think we just need to be sensible and think straight, see that this is the time to buy and make money in the long run, opportunities like these are rare
a bull run is great, but it doesn't teach you the lessons this does and give you what it is right now
$IREN pipeline doesn't mean anything if they simply can't convert that GW into AI factories of the highest quality.
Their business economics and financing structure is moderate and could be a long-term struggle.
Unhealthy stock!
🚨 $IREN's >35 GW Pipeline
They FINALLY did it.... I can't believe $IREN finally revealed the size of their massive pipeline.
For years investors were begging management to reveal the size of their power pipeline.
It only took the stock to crash 58% YTD for them to reveal this very relevant piece of info.
But honestly, wow. No other neocloud, or colo provider for that matter, even comes close to 35 GW...
I bought more today.
Yes, AI stocks.
Keep giving me such opportunities and I'll keep at it.
Thank you for your attention to this matter!
Never seen this much stupidity in the market around a theme as I'm seeing with the memory names.
How can you not be attracted to the insane cash flow these companies will generate over the next several quarters?
They won't sit quiet while the street calls them commodity businesses. And oversupply won't play out like previous cycles, no matter how much you cry.
1. HBM eats a big share of wafer capacity itself, and bringing new capacity online for a structural shift in demand is hard and slow.
2. They learned their lesson. Last time buyers got cheap memory and gave little in return, which hurt the suppliers while buyers were all smiles. Cough cough, $AAPL.
I also see these companies diversifying across the semiconductor supply chain. Samsung already has, and Micron and SK Hynix likely follow.
The cash they'll have makes it very possible. It takes time, but the street is forward looking, so any news on that and they trade at 25-30x PE in no time.
@sama, blame our boy Timmy Cook for the name change ✌️
The street is completely wrong about how it looks at memory companies today.
I think I figured out the reason the next bull run comes for these names, and it’s not what most people think.
Many investors look at Micron, SK Hynix and Samsung as memory companies. That’s why their multiples are so compressed and why they get tagged as commodity or cyclical. But that view is wrong, because it isn’t considering certain important shifts.
In the latest quarter, the big three made $169B in operating profit combined. Take Broadcom, ASML, Applied Materials, AMD and Intel together and you get to $29B. That’s more than 5x what five of the biggest chip companies made. On free cash flow, Micron did $33B in a single quarter, more than double Broadcom’s best quarter ever. And this doesn’t stop in 2026, especially with the long-term deals they’re signing that run through 2030.
Now think about what that cash does. Hock Tan just said the next big bottleneck of the AI buildout is capital, not chips or memory. Guess who is getting insane amounts of cash every quarter?
These companies can reward investors through dividends and buybacks, they can go after acquisitions, and they can push into the rest of the semiconductor stack, from packaging to logic to foundry. Maybe they even go further into the software layer with their own cloud. Who knows.
The push into the rest of the stack is the point I need you all to pay attention to most.
Samsung already has all of it, with foundry, logic, packaging and devices under one roof. Hynix and Micron seem to be moving that way. Micron is sitting on $73B in cash and returned only about 2% of its free cash flow last year.
So this is the main point I want you to take from this post. Stop looking at these as memory companies. Because of how they’re redefining the memory structure, they’re building incredible moats. But it’s not just that. You need to start looking at them as full-stack semiconductor companies. Once you shift that perspective, everything makes sense. The moat protects their pricing power and margins, the cash funds the next expansion, and the next expansion widens the moat again. That’s how you get to $2T and $3T valuations.
And the product itself keeps changing. Memory is going custom. With HBM4 the base die moved to a logic process, so customer IP now lives inside the memory stack. Every stack gets designed together with the chip. Then you have HBM4E, HBM5, and further out Samsung’s zHBM, which stacks memory right on top of the accelerator. High bandwidth flash and even new DRAM architectures are coming. Every new rack needs more bandwidth, and HBM takes way more wafers, so supply stays tight and margins hold.
Then there’s demand on the application level. We’re not talking about one agent anymore. We’re talking about thousands of agents working at once, and every one of them carries its own context, which all lives in memory. Software is getting personalized on the spot, with AI generating the code or the visual you want right there, something Zuckerberg talked about recently. Enterprise, consumer, software, everything is getting tied to AI. I don’t see the exponential rate of increase in token consumption going down anytime soon.
The street many times comes late to understanding the bigger vision companies are working toward. By the time it gets it, the stock has already priced it in. This feels like the time to buy heavy. NFA.
even at $5, I'm not buying $IREN
portfolio destroyer, this one
BREAKING NEWS🚨: According to @theinformation, Microsoft's $5B less in capex in Q3 than it had planned was in part due to delays by $IREN.
We would not be surprised by this, considering that IREN continues to come up in our conversations with neocloud customers, usually for the wrong reasons. Multi-day power outages, network upgrades, storage failures, air quality controls leading to link flaps, XIDs everywhere. The #1 worst site in the industry, according to users.
