DanielRomero

$INTC Is Getting Back Into Memory
The company is developing HBM alternatives: ZAM and XBM
Intel is studying next-generation memory architectures such as Z-Angle Memory (ZAM) and Cross-Batch Memory (XBM)
ZAM changes how the memory stack is built. It uses thinner memory layers and denser vertical connections, which allow for more memory capacity, higher bandwidth, and lower power consumption in the same area
XBM changes how the memory connects to the GPU. Instead of HBM’s extremely wide interface and expensive interposer, it uses high-speed serial links such as UCIe

$SKHY believes HBM is not the final answer to the memory bottleneck
VP of Memory System Research Kim Hong-il said SK Hynix is researching other solutions to ease the memory bottleneck, suggesting the company is preparing for the next stage beyond HBM even as it leads the market

I’ll be covering SEMICON Taiwan this year
Quite excited about it

Maybe I’m wrong, but why wouldn’t $TSM hike their prices like memory companies?
They want to preserve their good relationships and don’t want to incentivize others to move to $INTC or Samsung
But that’s happening anyway
They have a quasi-monopoly
Why not take advantage of it and make an insane amount of cash before Samsung and Intel get to large scale and achieve competitive yields in advanced nodes?
It seems almost cowardly not to make the most out of your pricing power when there’s an opportunity to do so
And even if your price hikes make others seem like attractive options because of the re

$GOOG making up 12.6% of Berkshire's portfolio feels crazy
Imagine telling Berkshire investors in 2023 that Warren would become AI-pilled
(Yes, he said it was his choice to buy Google.)

Both Elon Musk and Lip-Bu Tan want $SPCX and $INTC to move into memory manufacturing
If that's not a sign that high memory prices will last longer than most expect, I don't know what is

$AMAT plans to invest to double its system manufacturing capacity by 2028
You heard that right: 2x capacity in 2 years
The demand every company across the semiconductor supply chain is seeing is mental

Terafab has started construction
Insane speed

Jukan does a great job but the timing to capitulate was just SPOT ON

$AMD to raise as much as $5 billion through a debt offering, according to Bloomberg
AMD plans to use the proceeds for general corporate purposes, including potentially repaying existing debt, according to a company filing. The company has $875 million of bonds maturing next month

$SNDK calling a 75% operating margin sustainable has to be one of the most bullish memory datapoints you can see

What a day for $NBIS
+34% is crazy

One important aspect of the $NBIS call:
The shift towards short-term contracts
While hyperscaler contracts are a great financial bedrock, to extract the maximum upside, short-term contracts are more profitable
This has been $NBIS's strategy since the company went public
> In early calls, they said they wanted "hyperscaler contracts to survive if a neocloud winter comes."
What does that mean?
If you rely on small companies to fill up your capacity, these companies are the first to go under if there's a recession or AI downcycle
The company wanted to secure its future by signing profitable deal

$NBIS COMMENTS ON VINELAND
"To date, we have delivered all capacity tranches to Microsoft under contract, including two tranches over the past month."
Now waiting for the earnings call

$NBIS Key insights
ACV per MW jumped from $12M to over $20M for Q2 deals and over $40M for Q3 short-term deals
Nebius signed four Q2 deals averaging over $1B each, with prepayments covering 50–60% of capex and an estimated 22-month payback period
AI Cloud adjusted EBITDA margin reached 50%, up from 45% in Q1 and 24% in Q4
Year-end contracted power guidance increased from over 4 GW to 5 GW
Nebius ended Q2 with $8B in cash and secured $775M of asset-backed financing at SOFR + 2.5%

$NBIS Just Reported
The website is down, but we still got the numbers
🟢 Revenue: $582.3M vs. $574.7M est. (+1.3% beat)
🟢 Adjusted EBITDA: $236.2M vs. $173.1M est. (+36.5% beat)
🟢 Adjusted EBITDA margin: 40.6% vs. 30.1% est. (+10.5 pp)
🟢 Adjusted EBIT: -$23.5M vs. -$202.0M est. (loss 88.4% narrower)
🟢 Adjusted EPS: -$0.12 vs. -$0.70 est. (loss 83.1% narrower)
A huge profitability beat