DanielRomero

$META net income is cratering
The trend has clearly broken once you account for the one-off in Q3 2025
One of the reasons is hardware depreciation, which is up 45% YoY compared with 28% YoY revenue growth

How is $META still burning $18B annualized on Reality Labs?
You could probably fund 10 separate companies with that amount and have them achieve the same results
The inefficiency of large corporations

$META doesn’t want to provide 2027 capex guidance.
Which makes all the sense in the world
It wouldn’t be normal to provide guidance so far in advance in such a dynamic supply chain
Nobody knows what the credit markets will look like in 2027 either
I’m sure $META would love to keep breaking capex records, but that doesn’t depend solely on its own decisions

Most investors still don’t understand the magnitude of the Korean crash
> It’s literally the worst month in KOSPI history
> Nothing like this had happened in 46 years
Two million Koreans have been margin-called, and more than 500,000 have been liquidated

July 2026 has been the worst month for semis since September 2000

The state of semiconductor investing:
> MLCC stocks went through an entire boom-and-bust cycle before prices even began to rise significantly
> Taiyo Yuden is the clearest example: down 60% from its highs in less than a month, yet still up 150% YTD
This market is insanity
There are dozens of examples like this one

> Announcement on strikes against Iran
> FOMC meeting
> Microsoft and Meta earnings after the close
What a day
Hopefully, we don’t get mixed signals. I’d rather see either a strong recovery or complete capitulation

Wall Street failing to model HBM pricing dynamics correctly doesn’t make $SKHY’s earnings bad
HBM prices are tied to longer-term contracts, which gives SK Hynix less immediate exposure to surging conventional DRAM prices
That makes earnings less explosive during pricing spikes, but also more resilient
The company still generated ₩93.9 trillion in quarterly net profit, which translates into an annualized P/E of 3x at the current valuation
There’s nothing structurally wrong here, and certainly no demand issue
Yet when the market turns bearish, bullish news gets ignored and imperfections become d

$AMD and $CORZ are partnering to build AI data-center capacity specifically optimized for AMD hardwareaa
> AMD has secured more than 500 MW of U.S. capacity starting in 2027, expandable to 2.5 GW
> Core Scientific will provide the sites, power and infrastructure
> Both companies will collaborate on the physical design so the facilities can efficiently deploy AMD Instinct GPUs, EPYC CPUs and ROCm
> The capacity is intended for AMD customers, including model developers, cloud providers and enterprises
AMD will also receive market-priced warrants in Core Scientific, subject to commercial cond

This would be the buy-the-dip opportunity of the decade
I doubt it happens

The KOSPI is down 11%
It is composed of all common stocks listed on Korea’s main market and has 954 constituents
Yet it trades like a small cap
To see an index representing an entire country’s economy trade like that is mind-blowing, but that’s the Korean way

BofA says the semiconductor selloff is a buying opportunity
Analysts led by Didier Scemama noted that the Philadelphia Semiconductor Index’s roughly 18% peak-to-trough underperformance against the S&P 500 resembles the geopolitical selloffs of 2015 and 2018, rather than the approximately 30% drawdowns typically seen during full cyclical downturns
“Given strong fundamentals, we think the sector looks highly attractive trading on c3x discount to average ’28 consensus multiples and semicaps specifically at 6-7x discount,” Scemama wrote
BofA sees the greatest opportunity in semiconductor capital

Semis going down on:
> Morgan Stanley raising its 2027 and 2028 hyperscaler capex forecasts by 9% and 10%, respectively, now modeling $1.2 trillion and $1.4 trillion in annual spending
> $GOOG raising 2026 capex guidance from $180–190 billion to $195–205 billion, posting negative free cash flow for the quarter and saying demand still outpaces its capacity investments
> $TSM raising 2026 capex guidance from $52–56 billion to $60–64 billion and increasing its full-year revenue growth outlook from more than 30% to above 40%
> $ASML raising 2026 revenue guidance for the second consecutive quar

$AMD released a state-of-the-art, fully open model trained entirely on MI300X and MI325X GPUs using ROCm
Instella-MoE is a 16-billion-parameter mixture-of-experts model that activates only 2.8 billion parameters per token, reducing inference costs while remaining competitive with dense and MoE models with similar or larger active parameter counts
It supports a 64K context window and was trained through six stages, from pre-training to reinforcement learning
AMD also introduced two new efficiency techniques:
- Gated Multi-head Latent Attention, which filters low-value attention outputs.
- Far

Bernstein’s price targets, from a research note led by analyst Gautam Chhugani
$WULF TeraWulf: $36
$CIFR Cipher Mining: $32
$CORZ Core Scientific: $32
$CLSK CleanSpark: $24
$RIOT Riot Platforms: $30
$MARA MARA Holdings: $17
$IREN IREN: $100
Bernstein rates all of them Outperform, except MARA, which is rated Market Perform

Short Thoughts on $LPK
I've been following this company for some time and had the pleasure of speaking with the CEO
Still, I only traded the stock. I never took a long-term position
I don't think focusing too much on the latest earnings is useful. These numbers are largely irrelevant
The key theme for this company is glass substrates, and it will take a couple more years for them to penetrate the semiconductor industry meaningfully
However, I believe the most concerning issue is management credibility
1. Consecutive Overpromising
> The company missed its 2024 guidance.
> Then its 2025 guidan