YTD performance: +91%
After such a run, I’m now looking at what comes next
> 71 companies on the watchlist, mapped out over the coming months
I share it all here
https://t.co/pNFeU6ZfPO
I won't name names
But... What's even the point of the FinX investors who pick 5–10 stocks per popular theme?
Like, there's not a chance in hell you can possibly understand 5–10 stocks in depth across semis, robotics, biotech, cybersecurity, photonics...
"Training costs" (easy to turn on and off) 😂
How can you use a 4-month model development cycle as a bullish argument, yet also ignore the huge training costs?
You can't have your cake and eat it too
Makes no sense
Try turning off your training compute and see what happens
This is so clearly just a forced hit piece because $IREN is a popular company that catches eyeballs
It's easy to attack a new neocloud that has a couple of old "testing" sites and is only now ramping up literally a gigawatt of state-of-the-art deployments
Let's come back to this in a couple of years
Towa to Build Advanced Packaging Equipment Factory in Japan as AI Demand Accelerates
The company will invest ¥5 billion ($31.7 million) in a new 5,000 m² factory in Kyoto, with operations expected to begin as early as 2029
The plant will produce next-generation semiconductor packaging equipment, with Towa citing rising overseas labor costs, the weak yen, economic security, and concerns over technology leakage as reasons for shifting more strategic production back to Japan
Towa remains especially strong in HBM and compression molding, holding more than 60% of the global resin-molding equipment market and facing little competition in high-end compression molding
AI and data-center demand are already boosting results. For FY2026, Towa expects revenue to grow 18% YoY and operating profit to increase 48%
They also increased their financial targets by 2x in like six months
The market isn’t looking just at what $AMD says, but at where things are going with agents
Meta and OpenAI are key customers, and they’re entering an arms race in agents
Agents need lots of CPUs. $AMD has the best CPUs
The writing is on the wall
HBM Growth Pushes Memory Offload Toward New Use Cases
HBM growth is pushing memory offload toward larger-memory and legacy-GPU use cases rather than eliminating it
Next-gen accelerators keep increasing native HBM, from 180GB on Nvidia’s B200 to 288GB on B300/Rubin and 432GB on AMD’s MI455X, letting more workloads stay entirely in HBM
External memory still makes sense for older systems and on-prem deployments where memory capacity, not compute, is the bottleneck. It can extend hardware life without adding GPUs, networking, power, and cooling costs
The main challenge is now data movement. Offload only works if DRAM or SSD data can be prefetched fast enough to avoid GPU idle time
CXL and flash are complementary: CXL for memory expansion, flash for tiering and software-managed offload
$ALAB $MRVL $SNDK $MU