DanielRomero

SK Hynix will start trading on Nasdaq on July 10 of this year
➡️ The company plans to raise up to $30B
➡️ It will cause share dilution of around 2.4%
Overall, very bullish. A decent capital raise, plus access to immense liquidity. A win for shareholders
Samsung should be next

UBS lifts $AMD to $670 from $455
Analyst Timothy Arcuri sees EPYC CPU racks gaining strong adoption, citing:
• AMD’s edge in core count
• Strong multithreading performance
• The strength of the x86 software ecosystem
He expects AMD to take outsized incremental share, given $INTC roadmap and supply problems
Key Estimates
• 2026 CPU server revenue held at $16B
• 2027 raised to $23B, from $21B
• 2028 raised to $29B, from $27B
• 2030 CPU server revenue raised to $50B, from $41B
UBS assumes a 60/40 x86-vs-ARM split in the standalone segment, while still conceding that $ARM takes around 70% of the h

$MU has now significantly surpassed $NVDA margins
From the last reported quarters:
Micron gross margin: 84.9%
NVIDIA gross margin: 75.0%

$MU up 16% after-hours
Spent the whole week listening to people say, "Doesn't matter what they say on the call, the market will spin it negative."
"If margin guidance comes in high, it’s bad. If it comes in low, it’s bad as well." Blah blah
Bears sound smart, bulls make money

The Memory Party Continues
Extremely strong earnings from $MU
So bullish
Here’s the info you need to know:
Q4 guidance blew past estimates: $50B in revenue, 86% gross margin, and $31 EPS, versus expectations of $44B and $26. Mehrotra tied the strength directly to memory’s changing role: "Micron's record fiscal Q3 financial results and even stronger outlook for Q4 reflect the strategic value of memory in the AI era." At $31 EPS, Micron is annualizing north of $120, representing a modest 10x P/E
Gross margin hit 84.9%, a company record, up from 39% a year ago. On the SCAs underpinning the durab

The cyclicality of memory is slowly dying
$MU now has 16 strategic customer agreements, and 14 of them represent around $100B of minimum contractual revenue
Note that this is a floor. Actual revenue will still depend on dynamic pricing
This gives $MU visibility and stability while leaving room for upside
Absurdly strong business model

Seems like Bernstein had a lot to do with the recent weakness in memory stocks
Probably one of the most bearish estimate sets we’ve seen so far
Others are not modeling such a sharp EPS drop
Estimates like these are actually quite bullish for the sector, because they leave a lot of room for upside surprises

$AMD powers 4 of the world’s top 10 supercomputers, more than any other chip company on the latest TOP500 list

There's blood on the streets
Take care

“Even with these wafer additions, we expect DRAM to remain undersupplied through 2028.”
If that becomes reality, you really want to be long memory here
$MU could be trading at 3x 2028 earnings today, while SK Hynix and Samsung could be even lower than that
And that’s without buybacks

Two thoughts:
1. If DRAM prices remain even just stable, earnings are going to explode
2. CXMT is even larger than I thought. Massive company

A standard correction, and people get scared
Samsung and SK Hynix now trade at 7x and 9x 2026 earnings
If these companies went any lower, they would make their entire market cap in profit before 2030
This is a selloff driven by liquidity and margin dynamics, not structural

TrendForce says the DRAM shortage is expanding into older DDR2
DDR2 contract prices are expected to rise 55% to 60% in Q2 2026, followed by another 35% to 40% in Q3 2026
The three major DRAM makers are shifting capacity toward HBM and server DRAM because AI infrastructure demand is stronger and more profitable. That reduces supply for older, mature-node DRAM products.
As DDR4 supply tightens and prices rise, some brands and ODMs are lowering specs to control costs, moving from DDR4 to DDR3 or even from DDR3 to DDR2
That is pushing shortage pressure into older memory generations
━━━━━━━━━━━━━━━

Taiwanese server manufacturer Compal has received NVL72-Ready certification for its Coolant Distribution Unit
Not sure if they can compete with the big boys, but it is interesting

$INTC CEO Chen Liwu:
Intel’s goal is to generate a "10x return within 5 to 10 years" for investors
In an interview with the podcast "No Priors," Chen Liwu emphasized that while the company still lags behind $TSM and "must remain humble," people will only begin to see Intel’s true potential between 2030 and 2032
When asked about his decision to keep investing in the foundry business, Chen said that wafer foundry is very important to the United States and extremely important to the entire industry. Intel’s most advanced processes, such as 18A (1.4nm), are already being planned for 1nm and 0.7nm

Glass substrates hype is ramping up
While most people focus on TGV, Metallization is the hardest step in building a glass substrate and the one actually holding back yields right now
TGV is essentially solved: the lasers are qualified, and the process is under control
$LPK CEO told me as much directly
What isn’t fully solved is filling those holes with copper while maintaining structural integrity
━━━━━━━━━━━━━━━
Here’s how the process works
Once the vias are drilled, using TGV technology like $LPK LIDE, they’re still empty holes in glass
Metallization is the set of steps that turns them into