Looking for mispriced opportunities. Semis, AI infrastructure, defense. YouTube: High Water Investing
Los Angeles
Joined October 2026
- Tweets1.6K
- Following593
- Followers196
- Likes455
Nvidia's board just launched an independent probe into whether Jensen Huang was targeted by a foreign intelligence effort. $NVDA the board says no reason to believe secrets were compromised, Jensen is cooperating, and he stays CEO. Earnings land Nov 17, and now this!
📢 𝐉𝐔𝐒𝐓 𝐈𝐍: Statement from $NVDA NVIDIA Board of Directors
stockwhiz.ai/news/statement-…
Memory is the constraint inside the constraint now. $NVDA supply commitments jumped $160B in one quarter, and the CFO said it plainly: mostly memory. $MU $SKHY SK hynix is sold out through 2029. The queue for HBM now stretches past the queue for GPUs!
$MU $SNDK
Memory is the constraint inside the constraint now.
$NVDA's $279B in supply commitments jumped $160B in a single quarter. The CFO said it plainly: mostly memory.
SK hynix is sold out through 2029. The queue for HBM stretches past the queue for GPUs.
Whoever holds the wafers sets the price. We wrote the longer version on Substack. Subscribe and tomorrow's lands in your inbox.
BCG asks who pays the $2.6 trillion AI infrastructure bill, with installed AI capacity jumping from 17.8 GW to 120.3 GW by 2030. $VRT that is nearly 7x the power in five years. Somebody has to build it, cool it, and pay for it!
BCG: Who Will Pay the $2.6 Trillion AI Infrastructure Bill?
> We estimate that installed AI capacity will grow from 17.8 GW in 2025 to 120.3 GW in 2030, which is 14% more than Brazil’s highest peak energy demand and only slightly less than Japan’s.
> Once a gigawatt of compute capacity is online in 2026, our model estimates that a data center will need to earn $15 billion per gigawatt to cover its costs, including both the cost of debt and an acceptable ROI for equity investors.
> Although we have modeled the required AI revenue to be $2.3 trillion in 2030, our total system or end-market economic requirement in 2030 is $2.6 trillion.
> The International Data Corporation (IDC) forecasts that companies will spend $7.3 trillion on IT in 2030, excluding China and AI.
> $42 billion to build 1 GW of AI compute in 2026, core IT representing $28.8 billion of the cost.
> Inference % of compute expected to increase from 66% in 2026 to 72% by 2030.
$AMZN $NVDA $AMD $GOOGL $MSFT $META $ORCL
JPM says memory order visibility keeps extending, but the market is pricing a downcycle as early as 1H27 on AI funding and power fears. $MU $SKHY the stocks now trade under 3.5x while EPS has not budged. The debate is not demand, it is whether the cycle breaks first!
J.P. Morgan on memory:
"The order visibility is extending as we progress through the year, but we believe the market is anticipating AI funding risk and power bottleneck driven memory downcycle as early as in 1H27E. This puts memory in a tricky spot as investors debate between "too-cheap to ignore" vs. "it is cheap for a reason." We believe EPS cuts are unlikely to happen in the next 6M as the memory S-D gap is too wide to narrow over a short period of time and memory earnings will have a buffer protected by LTAs and prepayments."
$MU $DRAM $EWY $SNDK $SKHY
The squeeze on Iran keeps tightening on land too. The US just asked Iraq to shut most of its border crossings with Iran and keep only two open. The naval blockade was only half the story, and Tehran's overland lifeline is getting squeezed next. How much more pressure can the economy take before talks move?
US asks Iraq to shut most Iran border crossings, Iraqi foreign minister says reut.rs/4sVXFIE
Iraq just reached an understanding with the US to keep its oil exports flowing through Hormuz via a shipping lane close to Oman! Exports remain ongoing even with the strait under threat. If Baghdad's barrels keep moving, the supply scare loses a lot of its bite, no?
Trump just told CNN a return to war with Iran is 'highly unlikely' if talks keep progressing! After weeks of tanker mines and blockade headlines, that is the cooling off signal oil traders have been waiting for. Does the risk premium finally start coming out of crude?
这句否认比看上去软。台积电的声明里没提3D DRAM，也没提英伟达，只说自己的使命是做全球逻辑芯片产业可信的技术和产能供应商。报道点了一个负责人，先进封装副总裁Hsu Kuo-chin，他以前负责美光在台湾的业务。
The denial is softer than it looks. TSMC never said the words 3D DRAM or Nvidia. It said its mission is to be the trusted provider of logic chips, and left the actual question unanswered. The report named a lead, Hsu Kuo-chin, the VP for advanced packaging who used to run Micron's Taiwan operations.
$50B per GW just to rent the compute, and 3x that is $150B! The whole company runs at an $80B annualized rate. Somebody has to make a return on all this compute, the math only works if inference monetizes fast
Anthropic is paying $50B/GW rate on the several hundred MW it’s renting. Paying 3x that is $150B/GW.
For reference, the entire company Anthropic annualized run rate is $80B. High benchmark case for monetization on the portion reserved for inference is call it $100B/GW per SA and others.
Someone has to make a return on compute. There’s a limit to what you can pay for it no matter your belief on what you can do with it.
$2M per MW! Anthropic just paid $3.4B for gas generators because the grid cant keep up. When the AI labs start buying their own power plants the bottleneck has officially moved from chips to electricity
The FT reports Anthropic paid $3.4B for 1.7GW of Caterpillar gas generators that OnePWR ordered about two years ago, or roughly $2M per MW, which shows what near-term generation is worth when the grid can't keep up.
$ANT $CAT
ft.com/content/41bf5c6a-9735…
86% growth and the market still prices it like the old 20% grower! Doubling AI semi growth in FY27 AND FY28 is not a value trap setup, its a market that stopped believing the guide. How long can that gap last
The last time $AVGO was trading this cheap, the company was growing by 20%
Now it just grew by 86%, while projecting to double AI semiconductor growth in FY2027 AND FY2028.
Unreal.
🧵 Stock thesis #37: Vertiv just posted Q3 sales of $4.31B, up 61%, with adjusted EPS up 78% and margins at 25.2%, above the top of its own guidance. The stock sits ~52% below its high. The market prices a cyclical. The order book says toll booth. That's $VRT.
The market prices the orphan Emerson threw out. The deposits say every AI factory pays this toll booth. I'm long $VRT.
This is my personal research, not investment advice. Do your own homework. Full thesis: highwaterinvesting.substack.…. Follow @HighwaterCap, I dig into one mispriced company every day.