As long-time capital markets people, we think in terms of "house views": What we believe, and what thus drives our interpretation of events. These can and do evolve, but it is also important to know them so one can be clear when one is drifting away, basking in the warm seas of confirmation bias, or seeing actual falsification.
And, to be clear, we delight in discovering when we are wrong. Our greatest fear is being wrong and never finding out, so we aggressively look for falsifying data. As the writer Kathryn Schulz wrote in Being Wrong, her classic book on the topic, being wrong feels exactly like being right—until you find out you are wrong
So, with thousands of new subscribers, heading into Q4 2026 feels like a good time to lay out our house views—subject to evolution, revision, falsification, and the odd moment of tub-thumping triumphalism.
Here is the house view, in plain language, bulleted for easy scanning. We will revisit in the future as events evolve.
- AI is real and consequential. Adoption is extraordinarily fast, many productivity gains are real, and the technology will matter enormously.
- The current AI boom is wildly distorting the economy. Massive spending on chips, data centers, power and networking is now large enough to essentially be GDP growth, to distort profits, make trade data useless, break electricity grids, and distort capital spending.
- Headline economic data is misleading. Without AI capex and its growing externalities—and recently energy shocks—the U.S. economy would look much weaker, plausibly even contracting.
- The boom is debt-financed. AI infrastructure is no the largest borrower in investment-grade and high-yield markets, which is causing spiking funding pressure in other markets, from corporate to sovereigns.
- The basic financial mismatch is ugly: companies are borrowing long-term to build infrastructure producing a commodity—compute/tokens—whose performance-adjusted price is falling 70-80% year-over-year. Usage therefore has to rise hundreds of percent into perpetuity to stand still, let alone produce Wall Street growth while funding towering debt.
- This can be true even if AI succeeds technologically. The internet was transformative; telecom investors still destroyed vast amounts of capital. Great technology does not imply good infrastructure economics. It generally implies the opposite: back to fiber, railroads, canals, etc.
- AI capex is crowding out other investment. Corporate cash flow can look strong partly because firms are delaying ordinary capex, having been squeezed out, while an unusually concentrated group pours money into AI.