A Yale Economist Says AGI Won’t Automate Most Jobs—Because They’re Not Worth the Trouble

Why It Matters

The analysis shifts the AI debate from headline‑grabbing job loss to a deeper concern about wealth concentration, urging policymakers and investors to address compute ownership and distribution.

Key Takeaways

• AGI values compute over human labor
• Most non‑essential jobs will stay human
• Labor share of GDP trends toward zero
• Wealth concentrates with owners of computing resources
• Policy focus shifts to compute ownership redistribution

Pulse Analysis

The debate over artificial intelligence and employment has long centered on the fear of mass displacement. Restrepo’s latest research reframes that narrative by treating computational power, not human skill, as the primary scarce resource in a post‑AGI economy. By assigning a cost to the compute needed to replicate any task, the model revalues work based on its contribution to economic bottlenecks. This perspective explains why high‑impact activities—energy production, scientific breakthroughs, and national security—are prime candidates for automation, while socially intensive or creative roles remain uneconomical to replace.

For workers, the distinction between "bottleneck" and "supplementary" tasks has concrete implications. Hospitality, arts, and many service‑oriented jobs are likely to persist because the computational expense of mimicking nuanced human interaction outweighs the benefits. Conversely, sectors that drive growth, such as data‑center construction, advanced manufacturing, and climate‑tech research, will see rapid AI integration, potentially reshaping wage structures. Restrepo’s model predicts a decoupling of wages from GDP, meaning that while total economic output rises, the average worker’s share of that pie diminishes, concentrating earnings among those who own the underlying compute infrastructure.

The policy ramifications are profound. As labor’s GDP share approaches zero, the distribution of compute assets becomes the defining equity issue of the AGI era. Proposals ranging from universal basic income funded by compute royalties to treating compute as a public utility echo historic debates over land and natural resources. Investors and corporations that secure early control of massive data‑center capacity stand to capture outsized returns, intensifying calls for regulatory frameworks that ensure broader access. For businesses, aligning strategy with compute‑centric growth—whether by investing in AI‑ready infrastructure or upskilling talent for bottleneck roles—will be essential to thrive in a landscape where the value of work is measured in FLOPS rather than hours logged.