these aren’t all loans, and the checks aren’t backed by nothing. there’s real revenue, real usage, real infrastructure, and a tool with an absurdly large economic surface area. after spending long enough in this industry, one thing becomes really obvious, once a tool shows THIS much potential, not reinvesting EVERY dollar you earn, and EVERY dollar you can raise, would be the stupid move.
that doesn’t mean every valuation is sane or that every company survives. a crash could absolutely kill startups, crush valuations, make stocks wildly volatile, and strand some data-center capacity. on my side, i worry a lot about 2023-era and older gpu data centers, for instance... but that’s not the same thing as everyone abandoning llms and going back to before ai, because that is just not possible, have you seen the revenues of anthropic and openai? money coming in! in! IT IS HUGE, not as huge as what they’re spending, sure, because THAT is the bet, that’s how silicon valley has always worked, just at a MASSIVE scale this time. spotify was founded in 2006 and only achieved its first full profitable year in 2024. nearly two decades of burning money before they finally turned green.
openai, anthropic, and the rest are silicon valley opportunists, obviously. their bet isn’t merely that ai will become profitable. their bet is that we’ll become so deeply dependent on the systems they’re building that, by the time the market settles and the excess gets flushed out, they’ll own the layer everything else runs through. a bubble can burst. the weaker companies (like my own startups) can and will die. valuations can collapse. none of that puts the genie back in the bottle. it just decides who owns the bottle
