The first episode of the Working Notes podcast centers on AI and the current buildout. Here’s a rundown of what I go over:
AI can be genuinely useful and still be a terrible investment at the wrong price.
The technology can survive even if the first capital structure collapses.
We have seen all this before (and I provide a few examples)
In each case, the build-out was real, but the speed, financing structure, and
In my view, the key question is not “is AI useful?” but “who is holding the risk?”
A lot of today’s AI financing is circular, opaque, and dependent on assumptions about demand that nobody can really audit yet.
When the major AI companies go public, the S-1 filings will matter more than the headlines.
The back half of those filings, especially the footnotes, will tell us where the pressure really is: revenue quality, useful lives, commitments, depreciation, stock comp, reseller arrangements, and demand durability.
I am not saying AI is useless or the utility is a mirage (albeit it is currently oversold and overhyped). I’m arguing something narrower: a real technology can be financially pumped beyond what the underlying business can support, and I think that’s what we’re seeing here. What matters is demand durability and if the companies, as they are currently operating, will ever be able to turn a profit, and who gets left holding the bag if they can’t.
