The AI Jobs Story Nobody Talks About

You've probably thought about which jobs AI is most likely to replace. Translation? Customer service? Low-skill work? Or maybe one day, yours? Bloomberg published an analysis in June that asked the question from a different angle. The headline was "The AI Doomsday Scenario Nobody Talks About." The scenario it was asking about: government tax revenue. The work AI is currently best at happens to be the most highly paid: software development, financial analysis, legal work, banking operations. In most developed countries, these workers pay the largest share of income taxes. In 2024, 66% of US federal revenue came from labor income, with high-earning white-collar workers contributing far beyond their share of the population. If this group's wages fall or their job openings shrink, even a modest rise in unemployment could punch a much bigger hole in government budgets than most people expect. The harder part: that hole tends to appear exactly when spending is most needed. Job retraining programs, unemployment benefits, universal health systems -- all of these run on the assumption that there's enough tax revenue to fund them. When revenue shrinks and spending needs rise, fiscal pressure builds fast. Goldman Sachs released a report around the same time that added a different piece of the picture. For AI to move from software into real factories and power grids, it runs into an unexpected bottleneck: not enough electricians. Goldman estimates that the US power sector will need to add more than 500,000 workers by 2030. Training takes three to four years before someone can work on-site. The current apprenticeship pipeline is not producing nearly enough. Without electricians, data centers can't be built. Without data centers, AI infrastructure can't run. Two things are happening at the same time: AI's easiest targets are high-paid white-collar jobs, but AI's biggest dependency is the blue-collar skilled trades that nobody worries about replacing. The tax base is quietly shifting. The workers who build the stage for AI don't have a fast track to more of them. "Will my job be replaced?" is one question. A less-asked one is: "Will this system still be able to catch me when things fall apart?" That second question isn't one you can ask AI and get a useful answer. Next time someone analyzes how many jobs AI is taking, ask them whether they accounted for tax revenue -- and ask about the electricians too.