Spend $2 Billion, Make $8 Billion: The New Presidential Math

On Monday, I joined to (once again) discuss the ongoing economic fallout from the Iran war, the implications of AI, and the political economy of corruption. Also, what it means to be human.

The Iran War, redux

We’re now twenty weeks into a “four-to-six week war,” and the economic costs are far grimmer than most coverage suggests. Markets can clue us into the scale here — every escalation drops American stocks, while de-escalation lifts them. The moves look small, but that’s because much of the chaos has already been priced in. Once you adjust for that, the market’s judgment tells us that this war has erased something north of a trillion dollars in the future profitability of American businesses.

Platypus Economics is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Some thoughts on AI and competition

Next, we discussed the AI price war — Anthropic and OpenAI are burning cash to undercut each other, and Charlie wondered whether it signals a bubble. On this, I was happy to deliver good news: this is glorious. This is exactly how market competition is meant to work. Rival firms are in a race to create better products, and they’re each feeling pressure to offer lower prices and hence a better deal to consumers. When we write in our textbooks about the value of competition, or Adam Smith describes the “invisible hand,” and folks talk glowingly about the value of markets, this is what they’re after: Robust competition delivering better products at lower prices to buyers.

It’s an interesting lesson in business journalism. This is a good news story for everyone except the AI labs who are seeing their margins compressed. (Okay, their margins may be negative right now, but think of this as part of a longer-run competitive strategy and they are still valued as if their strategies will be profitable.) But nothing is ever good news in business journalism, and the interests of you and I — consumers — are barely…