“because you are so limited on semiconductors... what is the value I can derive out of this chip today... an H100 is worth more today than it was 3 years ago.”
An H100 is worth more today than 3 years ago; GPU depreciation cycles are longer than bears claim.
“TSMC is not raising prices, but memory vendors are just like sort of to some extent raising a lot of price, right? So they're going to double or triple price…”
Memory prices keep doubling as KV cache and HBM demand swamp slow-to-build fabs. Long memory.
“I think at least this year we're going to see margins for the model vendors go up a lot, right? Because they're so capacity constrained, they have to destroy…”
AI model vendor margins go up this year because they're so capacity-constrained they must destroy demand.
“NVIDIA just did it way earlier than Google or Amazon. And in some cases, Google and Amazon had stumbling blocks.”
NVIDIA locked up most leading wafer, memory and PCB capacity by ordering early; supply secured = margin.
“Google had Nano Banana and Gemini 3, which caused their user metrics to skyrocket. And leadership at Google was like, oh... we have to double compute every...…”
Google woke up late and is now absurdly AGI-pilled, buying power, land and compute aggressively.
“NVIDIA is now the largest customer at TSMC, and NVIDIA is the largest customer at SK Hynix, the largest memory manufacturer”
Memory makers SK Hynix and Samsung capture the margin as AI memory triples and they sign long-term deals.
“by '28, '29, the bottleneck falls to the lowest rung on the supply chain, which is ASML, right? ASML makes the world's most complicated machine. I.e., an EUV…”
ASML is the ultimate AI bottleneck by 2028-29; only it makes EUV and can't scale fast. Long ASML.
“over the last 3 years, TSMC has done $100 billion of CapEx... Nvidia alone is turning some small fraction of $100 billion in CapEx... into $160 billion in a…”
TSMC's tiny 3nm CapEx slice converts into ~$160B/yr of NVIDIA earnings; foundry holds the keys. Long TSMC.
“at some point China is able to scale past us if AI takes longer to get to certain capability levels than... the vast majority of your guests on this podcast…”
Long AI timelines give China years to scale fabs, and Alibaba is the tradable winner of domestic compute.
“Who saw the energy bottleneck before others asymmetrically went to, you know, Siemens, Mitsubishi, and of course GE Vernova and bought up turbine capacity.…”
Gas turbine makers can charge huge prices into a turbine-starved AI buildout. Long GE Vernova.
“Bloom Energy's doing fuel cells. We've been like very positive on them for like a year and a half now because they have like such a capability to increase…”
Power is not the long-run bottleneck; gas engines, fuel cells, batteries can unlock hundreds of GW cheaply.
“electrician wages probably double or triple again, right? And there's going to be a lot of new people entering that field and there's going to be a ton of…”
Electrician scarcity gates the AI buildout, and the MEP contractor that commands that labor prices the premium into backlog.
“the main factor is going to be for reducing the number of people is modularizing things and making them in factories in Asia... these areas are going to ship…”
Modular DC assembly moves value to Taiwan factory integrators; Foxconn and Delta capture more of each buildout.
“they become a less and less relevant customer. And they also will just cut their orders because things in the supply chain are kicking them out, whether it…”
Apple gets squeezed out of leading nodes and hit hard by memory price spikes; phone volumes fall.