NVDA: The best business in the world has nowhere left to put the money

The print itself was the usual: EPS beat, data center $89B, up 117%, now 92% of the whole company, next-quarter guide $108B against a $104B consensus. the number that actually matters came on the call: management pointed to roughly 70% revenue growth for fiscal 2028. from a company that has roughly doubled three years running while lowballing every single quarterly guide, a 70% out-year number reads less like deceleration and more like the usual sandbag. that part will get priced, argued about, priced again.

The part i find more interesting as a value question is the capital account. $26B went out the door this quarter in buybacks and dividends, on top of an $80B authorization from may, plus equity stakes and financing extended to its own customers. this is a fabless company, its own capex needs are tiny. so a business earning some of the highest returns on capital ever recorded literally cannot reinvest at anything close to those returns. the marginal dollar gets recycled at a 2.5% earnings yield via buybacks, or lent down the customer chain, which is its own conversation.

Textbook says a compounder needs two things, high ROIC and somewhere to put it. NVDA has the first in historic quantity and almost none of the second. that makes it less a compounding machine and more a royalty on global AI capex that pays everything out. how do you all think about terminal value for something like that? the earnings power is enormous and the reinvestment runway is basically zero, and i'm not sure the usual quality-compounder framework fits either side of it.