Every piece, newest first, with the underlying data attached.
A lot of the AI bubble debate tends to focus on valuations, but I think the more useful question is how the gap between AI spending and AI revenue is being financed. Circular deals, growing debt, household money in index funds, etc. could be what turns a future slowdown into something bigger than a stock-market correction, possibly a broader credit problem.
A quick market sizing usually gives you a number without telling you where it came from. Using 2026 data center liquid cooling as the example, I walk through how setting aside untraceable reports and rebuilding the market from physical units can get you to a range you can probably defend in front of a client.
For most companies, the biggest AI risk is the gap between what a vendor shows in a demo and what the tool does in production. Closing it means buying AI the way you would hire: each step up in commitment waits for evidence from your own data, in your own workflow.
