satya nadella just proposed a new model of the firm. here's the breakdown

it's from his recent essay "a frontier without an ecosystem is not stable" – a piece on how companies should position themselves in an ai-driven economy

his core argument: the real moat is not which model you use, but the learning loop you build on top of it. every firm now runs on two kinds of capital – human capital (judgment, relationships, tacit knowledge) and token capital (owned ai capability, private evals, institutional memory). they feed each other, and the loop compounds into proprietary ip that survives any model swap underneath

what's sharp about it: it reframes "ai strategy" away from picking a vendor and toward owning the learning. "you can offload a task, or even a job, but you can never offload your learning" is the load-bearing line

what's worth pushing on: microsoft sells the picks and shovels for exactly this architecture, so the framing is not disinterested. and the compounding arrow is asserted more than proven – if frontier models keep absorbing general capability, the delta a firm's private loop adds may shrink, not grow

still, the sovereignty test is the most useful takeaway: can you swap the model without losing the company veteran? if not, you don't own your ai – you rent it