Publicar

1) I love this. Throwback FinTwit moment discussing single names. 2) I’ll just show you guys my numbers for $AMZN to make my point. - Backlog mix to labs overstates their reality bc of duration (OpenAI is $138B but over 8yrs, ramping over time, all committed). - Lab IaaS is lower gross margin but long-term and very low incremental opex, so converts at reasonable mid-high 30s EBIT and low-mid teen ROIC. Fine, not amazing. - Bedrock (token inference and agent platform) is what really matters for AWS. Bedrock is growing faster, will dwarf lab IaaS over time, at much higher margin (see thread below this on Bedrock). - Core enterprise workloads continue to grow at high margins, more On Demand mix here so understated in backlog/RPO. - See my AWS forecast below. Lab IaaS peaks at mid-teens % of AWS EBIT as contracted GW is delivered 2026-2028, before declining to HSD. - Total AWS is driven by enterprise and bedrock, with rising margins and trending towards Jassys vision of a $1T business in the early 2030s (at ~40% EBIT). - A final word: As enterprise demand ramps, I actually think the clouds negotiating power vs labs is increasing, bc the labs have more competition for the incremental GW. Clouds will have higher Rev/GW from enterprise and therefore I suspect more lab infra will be Neo/Colo/insourced, and what they buy from clouds will actually see prices go UP (this is in fact what is happening today).