When Ai and skilled labor are discussed together it’s often the looming unemployment that Ai will cause as it replaces skilled workers and allows companies to operate more productively with fewer people. The points that are overlooked in this conversation are the hyper-fixation on “professional” jobs and the labor that will go into the infrastructure that will evidently disemploy so many of our low level computer programmers and copy editors. Leaving this eventuality aside for the moment, the current relationship between human capital and Ai infrastructure is more prosaic. As the Center for Strategic and International Studies illustrates, we’re currently experiencing a dearth of skilled electricians, welders, construction laborers and HVAC technicians, all of whom are considered skilled workers. We don’t have the apprenticeship programs available to meet this demand. While governments and corporations address this, a feasible substitute would be the immigration of skilled workers. This might not be politically viable in the near term.
As things stand under current high-level projections, (those in which the Ai revolution continues to be as significant as the second Industrial Revolution) the US will need an additional 140,000 skilled workers; the low estimate still calls for 63,000. Using craft-labor coefficients that allow us to understand the sort of labor needed to build each new megawatt of data center capacity (man-hours in other words) we see each megawatt of Ai data center capacity requires about 1,800 electrician-hours. Ai data centers are in the range of hundreds of megawatts, and as demand grows so will our deficiency. We need to increase apprenticeship programs by about 50% by 2030 and this brings its own problems. Assuming either the state or private corporations wanted to fund these programs (some big tech firms have partnered with local colleges) the question remains, who would teach these courses? The instructors for expanded programs would come from the current labor pool, which would remove them from the workforce. We can’t train the next generation of skilled technicians without removing skilled labor from the workforce now.
To have a chance of changing course and catching up in this Ai competition, we mustn’t overlook the fundamentals. The Ai race requires more than arbitrarily large amounts of financial investment and speculation; these complex computer programs run on ever larger and intensive computers, and this in turn means data centers and the materials they’re made of, the people who build them, and the energy that will fuel them. While discussions about data centers are energy are common, the need for skilled human capital is left out of the public discourse. Overlooking this critical component is costing us time that we don’t have. We are losing the race for Ai domination while forsaking durable, middle class jobs that will themselves resist displacement from automation and generative Ai. Technological advancement doesn’t have to come at the cost of good jobs, but rather create and correlate to them. This analysis from CSIS comes as China opens the world’s first underwater data center, and as the Telegraph reports American executives feeling “shaken” by the advanced automated manufacturing they’ve witnessed in China. One can imagine the peculiar sense of eerie intimidation an executive might experience while touring a “dark factory,” the manufacturing behemoths in which so few humans operate they have no reason to keep the lights on, and so to save electricity, they don’t.
