Askeladd

Zero Alpha, Levered Beta | Chronic Schizo Bozo | Serendipity Maximizooor

[Askeladd @MaxLeveredBeta]

Not sure how tf it happened but it feels like I've slipped my disc with pull ups. Yeah wtf

[Askeladd @MaxLeveredBeta]

All fun and games when its up and "did you listen to me anon"

[degentrading @degentradingLSD]

@DarkCryptoLord grab ur balls and buy.

[Askeladd @MaxLeveredBeta]

@Merridew__ So one last attempt for the highs and everything cracks in July?

[Askeladd @MaxLeveredBeta]

Can we resume climbing the wall of worry now?

[Askeladd @MaxLeveredBeta]

@zephyr_z9 Might wanna convince your head of trading?

[Renny @rennyzucker]

I don’t get how you look at the print, then look at the results of the customer base that are driving it, look back at the print and think “yeah buying this here is a good idea”

[degentrading @degentradingLSD]

pre mkt thoughts 24 Jun 26

US yields were firm and unchanged for the day. In Asia trading, $KOSPI staged a nice rollercoaster to rebound up 3.5% at this time of writing while $NKY is almost unchanged for the day. When we talk about KOSPI and NKY...we inevitably veer into the memory sector. To really get a comprehensive look at this, we need to of course talk about $SNDK and $MU in NY trading yesterday. Both tickers were down ~13% before we get $MU earnings tonight. $SK HYNIX has staged a very strong rebound into todays session despite multiple curve balls thrown at it. This illustrates to me that the technical selling pressure may be abated very soon.

Before this bout of weakness, i was waiting for $MU earnings to come before getting long - given that $MU has always fared poorly post earnings. However, baring a big miss from $MU, i would lean on the balance of probability that $MU and the memory complex in general would stage a strong rebound through July. $RAM - the 2x levered long ETF for $DRAM is also launching tonight. I believe this would get extremely strong traction as well and in the short term be fuel for the memory rally.

Memory pooling names like $ALAB, $CRDO, $PENG have also been very firm through the drop yesterday. $ALAB sits comfortably at 400, 60% higher from the previous low.

The star of the show has to go however to the neoclouds, especially $NBIS. First off, the index inclusion for $NBIS has happened, typically names rally into inclusion and ease off thereafter. $NBIS looked like it was going to roll over into the open before staging a 15% rebound that brough it almost to the top of the range - this on a day where $SMH was down 6% is impressive strength. On the fundamental side, the case for compute has been talked about multiple times - i will briefly go over it here again (look at previous notes) - In a world for a sovereign buildout thesis, there will be a duplication for build out capacity, this expands the TAM expectations. Next, with competition from the Chinese Labs, OAI and Anthropic have to compete fiercer....compute and design are the parameters to push on.

On the neoclouds, i also believe that $SHAZ will be filed in the 13G as Situational Awareness exceeds a 5% stake. This is notable to me because, $SHAZ is a tiny cap stock without much of a buildout. However, why did Situational Awareness lead the equity deal for them? Is there something that Leopold sees in this? Regardless of the case, this funding (equity +convertible debt at ~5% interest) basically has raised enough cash for the 102MW of contract that they were guiding for mid 2027....which should bring about ~1-1.3B of ARR which on comparables would lead to a range between 6B ($CRWV) - 10B ($NBIS) - the stock is ~3B now (after factoring in the dilution)

This is highly speculative and i could be completely wrong about it.

$MU earnings expectations

EPS ~19.8
Rev ~34.5B

Expectations are high, especially on a raise for forward revenue guidance.

To conclude, i think baring a miss in $MU earnings, much of the technical selling in semis is over. I would be calling for neoclouds and the memory pooling sectors to lead in the rally with memory following closely behind.

Good luck!

[Askeladd @MaxLeveredBeta]

@Evan_ss6 Generational entry?

[Askeladd @MaxLeveredBeta]

@bubbleboi Bubblehu akbar.

[degentrading @degentradingLSD]

mkt thoughts 22 Jun 26

First off - the entire hyperscaler complex has sold off dramatically - led by $GOOG at -7% with most of the other names down 5-2%. This is contrasted against the receivers of capex - namely memory and memory pooling solutions. $MU and $SNDK are up ~4-6% with memory pooling solutions like $CRDO up 11%, ALAB up 2%.

In neoclouds, $NBIS was the most resilient name, hovering near 300 before weakness across the entire complex led by $CRWV (-~7%) dragged it down.

In my pre mkt thoughts, i made the case to reduce leverage going into end of Jun. Looking at the current flow and landscape, i think this is time to bid some short term protection, The overall fundamental story is still intact, however technical flows looks set to turn into sell pressure for the +1wk horizon.

Good luck!

[bubble boi @bubbleboi]

This is almost like saying any company associated with PCIe will go bananas. This misses the forest from the trees. I am bullish CXL but you need to understand what it actually is to know where the value accrues.

CXL is just a new protocol over the same physical layer PCIe sits on. It’s electrically no different. CXL 1.1/2.0 rides the PCIe 5.0 PHY at 32 GT/s and CXL 3.x rides the PCIe 6.x PHY at 64 GT/s, same connectors, same slots. A CXL device literally negotiates at boot via Alternate Protocol Negotiation, falling back to plain PCIe if the other side doesn���t speak CXL. The entire difference is at the link/transaction layer and semantics. Unlike PCIe you get three sub-protocols with CXL multiplexed on Flex Bus, CXL. I/O (basically PCIe), CXL.cache, and CXL.mem for byte-addressable load/store.

And to back up the point I’m making is you can buy CXL IP from basically anyone. In fact Cadence, Synopsys, even fucking Rambus sells it. There is no moat in just selling IP and the real IP is the SERDES anyway, which is lower bw and easier to make than the top of the line stuff you’d get from Broadcom/Marvell. CXL maxes out at 32G/64G PAM4 per lane riding the PCIe PHY, vs. the 112G/224G SerDes people are pushing for Ethernet/scale-out. It’s in a completely different league than what CXL needs and that’s why so many people sell it.

The point here is just like PCIe, you wouldn’t be bullish on a company because it has PCIe IP, you’re bullish the companies using it to solve problems. Additionally, CXL has all kinds of problems itself like down clocking vs. PCIe equivalent, and also the market currently is using it as a “memory expander” which is retarded. You can stick terabytes of DDR next to your socket. Adding it over CXL is just adding complexity and worse performance. The numbers back all back this up and it’s basically a NUMA node hope in latency that you’re paying to bolt on capacity you could’ve gotten from a DIMM slot.

Additionally, CXL really isn’t better than PCIe alone on latency or BW. It’s on the same PHY, so it has the same ceiling. To get the value out of CXL you need tons and tons of CXL cards all networked together, and if you look at Astera they understand that. They’re building CXL switches and a software interconnect layer that actually makes it usable with their Scorpio fabric switches as well as the COSMOS software suite which is the actual product. These aren’t major revenue drivers right now but they will be, it’s not a retimer business long term & it’s not licensing CXL to AI ASIC designers either they will just go with a plethora of other options. It’s the whole stack, from IP, switches, software, your own cards you become a one stop solution for scale up networking that isn’t locked into a walled garden.

If you look at their financial results it shows the value add they made $308M in revenue last quarter, +93% YoY, into a scale-up switch market projected at $20B by 2030. This is just switches, retimers, & IP the orchestration software will be huge eventually too and will have near 100% margins. You buy the switch, you buy the cards, you buy the cables, & man that software just makes it so easy I don’t want to switch now I got a huge data center to run it ain’t worth it.

That’s the thesis.

There are literally no good CXL plays other than Astera for that reason. Marvell which you would think should be competing is just a pump and dump by the management team to take your money and put it in their pockets and wall street is pricing them as an asic design house not realizing the asic design business is just IT outsourcing.

So trust me on this, I’ve spent a long time thinking about it and made a lot of money from it. Most of you are retarded, broke, and easily swayed while I am none of those things.

[Evanss6 @Evan_ss6]

Intel is spiritually the Dogecoin of this bull run except it actually has a chance to be wildly successful

[Askeladd @MaxLeveredBeta]

@bubbleboi bubble pls i'm only holding astera labs

[Askeladd @MaxLeveredBeta]

@aleabitoreddit sive and xfag coded

[Askeladd @MaxLeveredBeta]

@SStricklandMMA What is “would of”? Am I trippin or is it supposed to be “would have”?

[Askeladd @MaxLeveredBeta]

@Stockmadam @kevinxu What a menace hahaha 🤣

[Askeladd @MaxLeveredBeta]

@CKCapitalxx It’s a low float play, calm down on the valuations

[Askeladd @MaxLeveredBeta]

@SemiAnalysis_ $mu hbm, power semis, now $intc?

[Askeladd @MaxLeveredBeta]

Suddenly feeling underallocated to neoclouds, even though I get the bear case of capping margins during compute crunch.