Smith’s Anil Gupta Pushes Back on Bill Gates’ AI Job‑Loss Warnings

Bill Gates’ recent 6,000‑word essay and follow‑up comments to The New York Times have reignited debate over whether artificial intelligence is accelerating toward a future of widespread job displacement. Gates argues that current market incentives reward companies for replacing human workers with machines, and he proposes new policy tools—including a controversial “token tax” on AI compute—to slow automation and fund retraining programs. He also warns that young people entering the workforce may face sharply reduced entry‑level opportunities as AI systems take over routine tasks.

But strategy professor Anil K. Gupta at the University of Maryland’s Robert H. Smith School of Business and co‑lead of UMD‑LinkUp AI Maps, says the evidence does not support Gates’ most dire predictions—at least not yet.

Job-Loss Fears Not Reflected in Current Labor Data
Gupta points first to macroeconomic indicators. “Gates’ concerns about widespread job losses caused by AI may materialize in the future. But there’s no evidence of economy-wide AI-driven job displacement so far,” he says.

According to the AI Maps dataset, U.S. employers posted 4.2 million new jobs in Q4 2025, compared with 4.0 million in Q4 2019 and 3.7 million in Q1 2018. “Job creation has not collapsed. If anything, it has grown,” Gupta says. Federal data tell a similar story: the U.S. unemployment rate stood at 4.1% in July 2026, down from 4.4% in March.

Gupta is especially skeptical that young workers will be hit hardest. “The AI era requires workers with AI fluency. Younger workers—who are exposed to AI tools in college and are more open to new technologies—are more AI‑native than mid‑career professionals,” he says. They are also less expensive to hire, he adds, making them more attractive in roles where AI augments rather than replaces human labor.

UMD-LinkUp AI Maps calculations reinforce his point. Job postings explicitly targeting recent graduates (up to one year of experience) accounted for 12.6% of all U.S. postings in Q4 2025, up from 11.7% in Q4 2022—the quarter ChatGPT launched—and far above 8.7% in Q1 2018. “If AI were eliminating entry‑level opportunities, we would see the opposite trend,” Gupta says.

The AI-fluency of younger workers is already becoming visible. This summer, big banks assigned many interns to work on developing AI applications rather than more mundane projects as in earlier years.

Token Tax Misses the Economics of AI Compute
Gupta also challenges a proposed “token tax.” As AI agents proliferate, token consumption is exploding while per‑token prices have plummeted—OpenAI’s API prices for some classes of models have dropped more than 95% over the past several years. “Given these trends, a fixed per‑token tax would become an increasingly larger share of the cost of inference as token prices fall, while aggregate tax liabilities would rise rapidly as token use expands,” Gupta says. “No government can realistically adjust tax rates every few weeks.”

He adds that a token tax collected through commercial API providers could also disadvantage small and mid-sized organizations relative to large employers capable of running open-weight models on their own infrastructure.

If society eventually faces meaningful AI‑driven labor substitution, Gupta argues that the most powerful solution is already on the table: a shorter workweek. “Until World War II, the typical workweek was six days. We transitioned from six to five, and we can transition from five to four,” he says.

With early applications of four‑day schedules showing promising results, Gupta says such a shift could absorb a 20% substitution of human work by AI. “AI will reshape work. But the data do not support a narrative of imminent, widespread job loss—and certainly not one that falls hardest on young workers.”