Jensen Huang just called nuclear energy "wonderful" for AI.
The market rewarded companies that build AI.
The next wave of wealth? Companies that power it.
Stocks to watch:
$CEG - nuclear operator, PPAs with Microsoft & Meta
$GEV - grid infra for AI data centers
$SMR - only NRC-certified small modular reactor
$OKLO - Sam Altman-backed micro-reactors
$CCJ - world's top uranium miner
Every time Jensen talks, I break it down here.
Follow so you don't miss it.
$BE spiked from $10 to $300+ in 2 years up 3000%, but its not the real AI bottleneck.
It's actually $MU $WDC $SNDK $DRAM.
These will 10x-20x from here:
1/ Bandwidth is the new clock speed.
By 2035, Next-Gen GPUs need 1,024 TB/s of memory bandwidth. That's 64x what $NVDA $AMD $AVGO delivers today. Without HBM scaling, the GPU is just an expensive paperweight.
2/ HBM stacks are multiplying faster than GPU dies.
Rubin (2026): 8 HBM4 stacks. Next-Gen (2035): 32 HBM7 stacks. Memory is growing 4x faster than compute dies. The bottleneck isn't silicon it's stacked DRAM.
3/ Capacity is what makes trillion-parameter models possible.
288 GB on Rubin. 6,144 GB on Next-Gen. That's a 21x jump in HBM capacity in under 10 years. No memory = no frontier AI. Full stop.
4/ Power budgets are being eaten alive by memory.
Total system power goes from 2,200W to 15,360W by 2035. A massive chunk of that is HBM. Whoever solves memory power efficiency wins the datacenter.
5/ The interposer is becoming a city.
Rubin's interposer: 2,194 mm². Next-Gen: 9,245 mm². Advanced packaging is now a critical supply chain choke point.
$MU $SNDK $AMAT $WDC aren't just memory plays they're infrastructure.
♻️ RESHARE this post and write 1 comment, I'll DM you my favorite data center stock under $10. There is only 1 left and its exploding.
Mike Brown was fired the day after calling out De’Aaron Fox for making a stupid play back in 2024 that lost them the game lmao
Turns out bro was RIGHT the whole time 😂🤣