“even if you assume all incremental compute they've gotten has gone towards inference, their margins are at a floor of 72%... at the start of the year...…”
Anthropic margins float at a 72%+ floor as token demand vastly outstrips compute.
“I think in the next 6 to 18 months, we'll start seeing real breakthroughs in robotics that enable few-shot learning, i.e., there's a pre-trained robot…”
Robotics gets few-shot learning breakthroughs in 6-18 months as cheap implementation unlocks new models.
“Dario used to gloat about how OpenAI was being too aggressive on compute and Anthropic was more sensible in their scaling. And now Anthropic is like, fuck, we…”
OpenAI is far from cooked; aggressive compute buying lets it serve the demand Anthropic can't.
“As demand skyrockets, prices are going up for everything on the supply side. Whether it be the NGPUs, their prices are going up. In addition, their useful…”
NVIDIA margins stay fat as GPU prices rise and useful lives extend far beyond 5 years.
“You see ASML is completely sold out. And they need Carl Zeiss to expand faster. Everywhere along the chain, everyone's either sold out and margins are going…”
ASML completely sold out; Carl Zeiss is the bottleneck on expanding lithography supply.
“Memory can only grow capacity low double-digit percentages a year, right? 20%, 30% a year. Even less for NAND, a little bit higher for DRAM... the true…”
Memory prices double or triple again from here; capacity can't grow fast enough until 2028.
“I think logic also has humongous capacity problems. TSMC just had their earnings. They keep upping CapEx... they're not raising prices fast because they're…”
TSMC could spend $100B on CapEx by 2028, whipping its equipment suppliers harder than anyone expects.
“the semiconductor wafer fabrication equipment supply chain is one that I still think is, it's gone up a lot, but it's still very underappreciated.”
Semiconductor equipment makers get the tail-whip from TSMC's surging CapEx; still underappreciated.
“all the slop code you and I are generating that is now running on some Vercel instance... all of that requires CPU. And so CPUs are completely sold out, and…”
CPUs completely sold out as RL training environments and AI-generated app sprawl drive demand.