Concerns that artificial intelligence would trigger mass unemployment have largely evaporated since peaking in May, even as the technology reshapes hiring patterns and consumer behavior, according to a new analysis from a16z. News reports linking AI to job losses have dropped to levels last seen a year ago, while announcements of AI-related layoffs hit their lowest point of the year in August and September. At the same time, demand for technical workers has surged: job postings for software engineers climbed 26% since last October, engineering roles jumped 30%, and manufacturing positions rose 13%, according to data from Indeed. The report suggests that rather than replacing workers, AI's arrival has sparked greater demand for people building AI infrastructure and putting the technology to work.

The workforce data reveals striking shifts in who's hiring and what they're paying. Postings for human resources roles—typically a signal that companies plan to expand headcount—have finally returned to pre-pandemic baseline levels after years below that mark. Meanwhile, workers in explosives and ordnance occupations now earn a median of roughly $30 per hour, making them 20% to 25% better compensated than the typical American worker, with North Carolina and New York offering especially high wages. The explosives workforce skews unusually young, with approximately 71% of these workers aged 34 or younger, a demographic pattern that differs sharply from other skilled trades.

In online commerce, AI's influence appears increasingly concrete. Traffic referred by AI tools converts at higher rates, spends more time browsing, views more pages, bounces less often, and generates roughly 50% more revenue per visit compared with shoppers arriving through traditional channels, Adobe data shows. Visa just posted its largest full-year revenue increase in 15 years—excluding mergers and the pandemic recovery year—with nearly $46 billion in revenue on $15.7 trillion in transaction volume, a jump of almost 15% year-over-year. Consumer surveys from Bain indicate growing appetite for major online shopping events, with only 10% of respondents saying they plan to skip Black Friday, Prime Day, and Cyber Monday entirely. Meta's average ad prices in the United States and Canada climbed more than 20% compared with the prior year for two straight quarters, a sharp break from the mostly flat pricing that prevailed through much of 2024 and 2025.

The report cautions that widespread assumptions about AI economics rest on shaky ground. Market anxiety has centered on the so-called "Rack Spread"—the gap between rising GPU rental costs and falling token prices—which some observers interpret as evidence that AI models are becoming commoditized and that massive capital spending won't pay off. Yet the token pricing indexes driving this narrative come from small, biased samples that don't reflect actual enterprise demand, the report notes. Multiplying blended token prices by reported token volumes suggests spending has climbed somewhere between 122% and 160% since May, depending on the data source used. A survey from the Atlanta Fed found mean AI spending exceeded $1,000 per employee, while Ramp's data pointed to a median of roughly $188 per year—a gap wide enough to call the reliability of any single measure into question. The report concludes that no one actually knows total AI token spending, since AI companies don't report earnings and no sufficiently large panel exists to make sound estimates, meaning the proxies now shaping market sentiment reveal far less about true demand than investors seem to believe. For business leaders, the real challenge isn't whether AI justifies current investment levels—it's navigating strategic bets when the only certainty is that today's metrics won't measure tomorrow's value creation.