重度 Grok Bot 用户推演 SpaceX 2028 年万亿营收

Farzad 🇺🇸 🇮🇷 @farzyness原推发布于 2026.10.21

在 X 上打开原推

As someone who's been using AI agents EXTENSIVELY this year - and most recently have become completely and totally Grok Bot pilled - it's quite easy to see how SpaceX can reach its $1 trillion revenue target by 2028. There's these thresholds that keep being passed with model capability that FORCES YOU to delegate more to the agents because of how insanely productive they are becoming. Not only that, it enables you to be FAR more ambitious in trying to find the limits of capability of these models, which require WAY MORE tokens. I think we're still EXTREMELY EARLY in how well people are deploying these agents, and I think usage per account will likely 10x into 2028, while the growth of accounts does another 10-100x. Honestly probably way more. I don't think people are grasping just how profound agentic uses-cases are, and the demand that it'll generate. On top of that, because SpaceX OWNS ITS OWN COMPUTE (and is building it the fastest by far, and will eventually throw up in space)... which it can balance between its own agentic harness & customers that rent compute... means that SpaceX serves its models for insanely low costs because their customers are paying for their compute, which they can pay off in 6 months to a year. So they can train insanely good models. Serve them for extremely low costs. And make a TON OF MONEY from everyone else renting their compute because demand is off the charts. Monster company $SPCX