Opinion: The AI debate to remove the white collar – disaster or delay?
Editor’s Note: GeekWire publishes guest comments to encourage informed discussion and highlight diversity of opinion on issues that shape technology and the startup community. If you’re interested in submitting a guest column, email us at [email protected]. Submissions are reviewed by our editorial team for relevance and editorial standards.
Many economists argue that AI will rapidly change white-collar work, depress wages, destroy aggregate demand, and destroy our economy. The idea is simple: if machines can think, why would anyone pay people to do white-collar work? Thought tests like the widely publicized Crisis Intelligence Global 2028 offer a neat story arc: AI capabilities converge repeatedly, workers are displaced from real estate, and the economy slows down. Transportation doomers have the rhetorical advantage of a terrifying punchline.
The latest note from economists at Citadel Securities, The 2026 Global Intelligence Crisis, goes back. Their main argument: just because AI can improve itself doesn’t mean businesses will adopt it at the same accelerating pace. If technology makes workers more productive, that often increases the supply of goods and services, which is generally good for the economy. And there are real-world limits to how quickly companies can replace human workers with AI systems.
The first insight of note is what I would call the compute-cost ceiling. As companies rush to automate everything at once, the demand for computing power increases, and the price of that computing power increases. At some point, it becomes cheaper to pay a human to do a job than to pay an AI to do it, and automation stops. It’s a natural brake on the economy, and one that disaster traders completely ignore.
But hold on. There is a big hole in this argument, and I’m about to call the data center about it.
The commute-cost ceiling argument ignores the fact that commuting costs are falling off a cliff. The famous Moore’s law (costs of computing halve every 18 months) has stopped working after 50 years, but there is an AI version where costs decrease by a factor of 10 each year. Specifically, LLM index costs (on a per-token basis) have been dropping by almost an order of magnitude YoY over the past two years. Andreessen Horowitz coined the term “LLMflation” to describe this trend, documenting a nearly 1,000x drop in costs over three years. To be fair, the cost per transaction is slightly lower because the frontier models burn more logic tokens per query, but it still continues to decrease rapidly. A ceiling that drops 10x every year is not a ceiling. It is a speed.
The Citadel note also invokes John Maynard Keynes’ infamous 1930 prediction of the 15-hour work week, noting that Keynes was wrong because he underestimated the human appetite. People wanted more things. Okay, but this misses the distribution problem. The “people will want more things” argument only works if enough people have money to buy more things. If AI-driven profits flow to the top 0.1%, and everyone else is out of luck, then the economy is in trouble. In other words, Musk and Bezos can’t eat too much, and not enough to keep the economy booming.
So where does this leave us?
The Citadel note is correct that institutional conflicts are delaying the deployment of AI, and that democratic societies will eventually respond by adjusting policy. These are real brakes, buying time. But the potential for cost reduction is strong: algorithmic efficiency, hardware optimization, benchmarking, distillation, and strong price competition among index providers. None of this slows you down.
My conclusion is that the AI displacement doom loop has been exceeded and will take longer to arrive than many expect. But it is coming, and we are facing major economic and possibly social upheavals. Doomers ignore conflict. Optimists think that the conflict will last forever, as the cost curves wear them down year after year. The truth will come somewhere in between, and it will be ugly. Preparing for that is a real policy challenge, and so far, our policy makers seem to have focused their attention elsewhere.
