Dogs are the best fit for the first longevity drugs.
Why? No complex insurance industry distorting incentives.
We can also observe aging in our dogs more easily than in humans. According to @celinehalioua, this creates faster feedback loops for @loyalfordogs longevity treatments.
Dropped my thesis behind my $PENG position.
PENG is a large position in my portfolio.
The memory shortage driving my $MU thesis is showing up somewhere most people aren't looking.
MU builds it, PENG deploys it.
$GOOGL AI Supply Chain ETF:
Since they're raising $80B to finance AI infrastructure expansion.
Total 2026 AI Capex: ~$185B [2x 2025]
Custom silicon / memory:
- $AVGO: TPU
- $TSM: TPU foundry + advanced packaging
- $MRVL: apparently in talks for inference TPU + MPU.
- Samsung: HBM
- SK Hynix: HBM3E
- $MU: third source HBM
- $SNDK: NAND
Networking & optical:
- $LITE: R64 optical circuit switch, co-developed w/ Google + in production
- $COHR: 400G/800G/1.6T transceiver
- $CRDO: AECs linking TPU clusters
- $FN: contract manufacturer for transceivers
Google cloud compute:
- $NVDA: GPUs
- $AMD: GPUs, smaller exposure
Server ODM & power:
- $CLS: primary TPU board + rack-level integrator
- $VRT: CDUs, liquid cooling + rack power architecture
- $ETN: Switchgear, busways, UPS, electrification
- $BE: fuel cells
- $GLW: subsea fiber
Real estate / colocation:
- $WULF: hosts Google-backstopped AI capacity + Google holds warrants
- $HUT: 245 MW lease w/ Google financial backstop
- $CIFR: Fluidstack/Google HPC hosting
Just for a top level overview of potential direct beneficiaries.
There are ofc plenty of other names located in Taiwan/Korea like Wiwynn (6669) for server ODM flows.
But given the ~$460B cloud backlog, a large chunk of the raise will probably go to more memory since it's ~60% of capex.
Also TPUs since Google's thesis is replacing Nvidia with TPUs e.g. Broadcom.
If you have been wondering why “old school” server companies like Dell and HPE suddenly exploded higher again, this article is worth reading.
The key point is not simply that AI server demand is growing.
The center of the AI infrastructure war has shifted from:
“Who builds the best GPU?”
to:
“Who can actually deliver and operate a full AI factory inside the customer’s own datacenter?”
The article does a very good job explaining:
- liquid cooling
- power bottlenecks
- on premises AI
- sovereign AI
- enterprise AI rack integration
from a very practical perspective.
More importantly, it also discusses:
what exactly the market is re rating in Dell/HPE today,
and what could eventually break this rally.
A very interesting read for anyone following AI infrastructure investments.
Goldman also just raised the bar for its entire memory thesis again.
The big change isn’t 2026. It’s after - towards 2027-2028.
> Goldman now expects DRAM, NAND and HBM supply/demand to be even tighter in 2027 than 2026
> Undersupply conditions now expected to persist into 2028 across all three memory markets
> AI servers and agentic AI driving much stronger demand visibility than prior cycles
> Limited supply growth due to slower capacity additions and rising HBM wafer allocation
> Long-term agreements increasingly reducing traditional memory cycle volatility
> HBM pricing expected to “catch up” further versus conventional DRAM, driving a larger HBM TAM
> PTs raised for Samsung and SK Hynix; Kioxia upgraded to Buy (see post below)
$MU $005930.KS $000660.KS https://t.co/gvV7GGWRoK
Samsung Electronics Eyed for Anthropic Chip Order… Foundry Revival Begins
Samsung Electronics has made an investment in Anthropic, the U.S. AI company behind the AI model "Claude," raising the possibility of a foundry partnership in which Samsung would manufacture Anthropic's chips on a contract basis.
Samsung's foundry business, which has run at a loss for several years, has recently landed a string of major global customers, fueling growing expectations for a return to profitability next year.
On the 28th (local time), Anthropic announced that it had raised $65 billion in its recent Series H funding round, with its post-money valuation assessed at $965 billion (about 1,440 trillion won).
The three major memory semiconductor manufacturers—Samsung Electronics, SK Hynix, and Micron—participated in this funding round as "strategic infrastructure partners."
Anthropic emphasized that "the technologies of these companies play a critical role in the global supply of memory, storage, and logic chips," adding that "partnerships with them will be a great help in reliably scaling computing capacity to meet customer needs."
Anthropic's mention of "logic chips" in its announcement has prompted speculation about a foundry partnership with Samsung Electronics. The process of manufacturing logic chips is foundry work, and aside from Samsung Electronics, neither Micron nor SK Hynix has a foundry division. For this reason, the industry analyzes that Samsung Electronics and Anthropic will newly collaborate not just at the memory level but also in foundry.
This implies a high likelihood that Samsung Electronics will produce the AI chips used in Anthropic's flagship model "Claude." Anthropic, alongside OpenAI, is regarded as one of the world's leading AI model companies and a core AI player. The prevailing view is that Samsung's foundry has secured yet another major customer, following Tesla, NVIDIA, and others.
Earlier, Samsung Electronics won orders for Tesla's next-generation AI chips, "AI5" and "AI6." Samsung's foundry is also handling production of an upgraded version of "AI4." NVIDIA's inference-dedicated language processing unit (LPU) chip "Grok3" is likewise being produced through Samsung's foundry. Samsung Electronics also plans to supply the image sensor to be mounted in Apple's new iPhone next year.
Accordingly, attention is focused on whether Samsung Electronics' foundry business—which has run at a loss in recent years—can regain its vitality. As of last year, Samsung Electronics held second place (7.2%) in the global foundry market, but the gap with first-place TSMC (69.9%) reaches 62.7 percentage points.
Regarding Samsung's investment in Anthropic, the industry views it as strategic cooperation spanning "AI semiconductors, memory, foundry, and AI infrastructure." Samsung Electronics is positioning a "one-stop solution"—encompassing high-bandwidth memory (HBM), advanced foundry processes, and advanced packaging—as its differentiation point. As the AI market evolves into an ecosystem centered not only on HBM supply but also AI chip design, production, packaging, and system optimization, the move is interpreted as Samsung strengthening partnerships by leveraging its competitiveness across the semiconductor spectrum.
A semiconductor industry official said, "Samsung's investment is not a simple equity stake in an AI company, but is interpreted as a signal that Samsung is earnestly expanding strategic relationships with the key players of the AI era," adding, "Expectations are growing as to whether Samsung's foundry business can seize another opportunity amid the expansion of the AI market."
🚨 do you understand what happened to your right to protest..
The FBI just invented a new domestic extremism category called "anti-tech violent extremism" - and it doesn't exist in any public document.
WIRED got 1000+ leaked pages proving they're already using it to track AI critics and data center protesters.
What the leak exposed:
- Photographing a data center is now officially "suspicious behavior" in Virginia
- Fusion centers are surveilling town halls where residents complain about water
- 61% of Americans believe AI will destroy more jobs than it creates
- Even Marjorie Taylor Greene snapped: "how dare the peasants complain"
They built a new extremist category for you before they told you it exists.