Yves here. In posts and via links, we have pointed to many reasons why AI is set to come up short against its economic and performances promises. On the list are the ginormous data center needs, from political opposition to lack of enough water for cooling to grid capacity limitations. This post describes an additional one: too few skilled workers.
By Robert Rapier, a chemical engineer with 25 years of international experience in the chemical, oil and gas, and renewable energy industries who holds several patents related to his work. Originally published at OilPrice
- Goldman Sachs estimates U.S. data center power demand will more than double from 31 GW in 2025 to 66 GW by 2027, requiring a massive buildout of grid infrastructure.
- The U.S. power sector needs roughly 510,000 additional workers by 2030, according to Goldman, while a wave of experienced construction workers approaches retirement.
- The labor crunch gives pricing power to infrastructure contractors like Quanta Services, MYR Group, MasTec, and EMCOR, though it also limits how fast they can execute projects.
Most discussions about artificial intelligence focus on chips, data centers, power plants, and electricity demand. Those are all important. But another bottleneck is beginning to emerge, and it may prove to be an underappreciated challenge.
The AI boom needs electricians.
It also needs line workers, substation technicians, grid engineers, mechanical contractors, welders, construction crews, and commissioning specialists. These are not jobs that can be filled instantly with a software update or a new financing round. They require training, experience, and a steady labor pipeline that the power sector does not currently have in abundance.
That is an important reminder that the AI boom is not only a digital story. It is also very much a physical infrastructure story.
From Chips To Construction
The first phase of the AI buildout has been dominated by the race for computing power. Investors have focused on semiconductors, cloud providers, and the companies building massive data centers to support artificial intelligence workloads.
But every one of those facilities must be connected to the grid. It must have transformers, substations, backup generation, cooling systems, transmission access, and workers qualified to build and maintain that infrastructure.
That is where the problem becomes more complicated.
Reuters recently reported that the rush to build data centers is intensifying shortages of power and grid workers, including electricians, line workers, and other engineering, procurement, and construction roles. The issue is not just that demand is rising. It is rising while a large share of experienced construction workers is approaching retirement.
This creates a different kind of constraint than the ones most investors are used to thinking about. A utility can raise capital. A hyperscaler can sign a power purchase agreement. A developer can order equipment. But if the trained workers are not available, projects can still be delayed.
The Scale Of The Need
Goldman Sachs Research has estimated that U.S. data center power demand could rise from 31 gigawatts in 2025 to 41 gigawatts in 2026 and 66 gigawatts in 2027. That would more than double estimated data center capacity from the end of 2025 to the end of 2027.
Meeting that demand will require a massive buildout of generation, transmission, interconnection, and backup systems. Goldman has also estimated that the U.S. power sector will need roughly 510,000 additional workers by 2030 to satisfy rising demand, with Europe needing another 250,000.
Those numbers help explain why the labor issue could become a limiting factor. The power sector is not simply competing with itself. Data centers, utilities, renewable developers, manufacturers, industrial projects, and grid modernization programs are all chasing many of the same skilled workers.
The Bureau of Labor Statistics projects employment of electricians will grow 9% from 2024 to 2034, much faster than the average for all occupations. It also projects about 81,000 electrician openings each year, many of them tied to workers leaving the occupation or retiring.
For electrical power-line installers and repairers, the BLS projects 7% employment growth over the same period, also much faster than average, with about 10,700 openings per year.
Those are good jobs. But it takes time to train a qualified electrician or line worker, and the most experienced workers are often the ones needed for the most complex projects.
Costs, Delays, And Utility Bills
A shortage of skilled labor does not mean the AI buildout stops. It means the buildout may become more expensive and uneven.
Projects with the strongest sponsors, best locations, and clearest utility partnerships will likely move forward. Others may face delays, cost overruns, or longer interconnection timelines. The same pressure could also affect transmission upgrades, renewable projects, natural gas plants, and grid-hardening work.
This has direct implications for energy policy, utility customers, and investors.
If utilities must build more infrastructure to serve large data centers, someone has to pay for it. Regulators are already wrestling with whether the costs should be borne mainly by the large customers driving demand or spread more broadly across the rate base. Labor shortages add another layer to that debate because higher construction costs eventually show up in project economics.
That is one reason the data center boom is becoming more than a technology story. It is moving into utility regulation, construction labor, power markets, and local economic development.
Who Benefits?
For investors, the most obvious beneficiaries are not necessarily the AI companies themselves. The bottleneck will likely be electrical contractors, grid builders, equipment suppliers, and utility infrastructure companies.
Firms such as Quanta Services, MYR Group, MasTec, EMCOR, Eaton, and Vertiv sit much closer to the physical buildout than most software companies. The caveat is that labor shortages cut both ways. They can increase pricing power and backlog, but they can also limit how quickly projects can be completed. A second caveat is that shares of most of these companies have already notched huge moves up over the past year.
The Big Picture
AI may live in the cloud, but the cloud has to be built, powered, wired, cooled, and maintained. A model runs in the cloud. A chatbot answers a question. A search result appears instantly. But behind that experience is a chain of physical assets.
Chips may get most of the attention. Power plants and natural gas turbines are also getting more attention as electricity demand forecasts rise. But the labor force may become one of the most important constraints.
This is not an argument for or against AI or data centers. It is an argument for understanding the full supply chain behind them.
The companies best positioned for the next phase of the AI buildout may not only be those with the best chips or the largest data centers. They may also be the utilities, contractors, equipment suppliers, and infrastructure companies with access to skilled labor and the ability to execute large projects.
The AI boom may be digital at the surface, but underneath it is an old-fashioned construction challenge. And in that world, electricians and line workers may be just as important as algorithms.
I was reading that the situation has gotten so bad, that Meta is setting up a sort of training university for workers interested in this field with, I would imagine, guaranteed jobs at the end of it. At Meta projects of course.
The Thermodynamics of Capital: Artificial Intelligence, Energy Crisis, and Ecological Crisis.
I liked this article. It is, however, about 20 pages long showing that the Marxist Socialists can’t say anything briefly. I’ll put the first paragraph of the conclusion below, but the entire article is good.
Conclusion
The social and historical analysis developed in this article leads to a conclusion that the dominant discourse on AI and sustainability systematically evades: the ecological crisis of AI is not a problem of insufficient innovation or inadequate corporate responsibility, but a structural expression of capitalism’s irresolvable tension with the biophysical limits of the planet. The specific processes driving this crisis—the monopoly arms race, the Jevons dynamic, and the systematic displacement of ecological costs onto the Global South—are not technical malfunctions awaiting engineering solutions. They are the normal operations of capital accumulation in its monopoly-digital phase, registered in thermodynamic terms as dissipative processes of Prigoginian nonreversibility: permanent, compounding, and beyond the reach of market correction.
I see that the link was not added.
Heah tiz
https://monthlyreview.org/articles/the-thermodynamics-of-capital/
No comment.
Why do all blue collar jobs start ridiculously early in the morning? I probably could have been ok at some of them otherwise.
What do you mean by “early”, 05:00?
“Why do all blue collar jobs start ridiculously early in the morning?”
I don’t know why ALL of them do, but a lot of them start in the wee hours of the morn so that you’re done before outdoor air temperatures get too high. A lot of data center infrastructure work is outdoors (or indoors before air-conditioning equipment is installed, commissioned, and energized), and it can get brutally hot while you’re having to engage in significant physical labor. In severe conditions, it can be a legitimate safety issue.
I work in the “utility infrastructure” sector that Rapier mentions. Most of my career has been spent behind a desk, but I had at least one trip to the field that involved significant labor in a building with no AC. At a job site located next to a swamp. It was incredibly muggy, and we definitely wanted to get out of there before 3PM.
The concept of siesta is a very sensible one. It would be counterproductive, and dangerous, to do hard physical labor during the hottest part of the day, when one is digesting one’s lunch anyway and is liable to be drowsy and sluggish. Work hardest during the early morning, and the late afternoon/early evening, when the sun is not nearly as hot. Save the least useful and most dangerous time of early afternoon for a refreshing rest, so one can renew oneself and not be totally worn out.
Unfortunately, the idea of a siesta is not compatible with the Protestant/Puritan work ethic that claims idle hands are the devil’s work. So in America, you are told it’s better to fast through the lunch hour, so you can get more stuff done, and to work from sunup to sundown without a break.
What truly matters in America, after all, is not the worker as a human being, nor even the quality of the work, but the performance of industriousness.
It seems like you could do a search and replace on the post above and publish the same analysis on many different industries and workforces in first world societies. Off the top of my head, machinists, farmers, elevator inspectors, nurses, general practitioners, teachers, auto repair technicians, and doubtless many others that I can’t remember.
I think a combination of steadily declining real wages, skyrocketing prices for housing, childcare, food, and other necessities, and the fact that many professions take years to learn and become adept at, puts young workers in a quandary. The decision to commit your life to some particular field is a difficult one under ideal circumstances, and societies no longer make it obvious that you can support yourself and your family in any particular job. Things like outsourcing to foreign countries, artificial intelligence and other trends make the ground under your feet seem even shakier to potential workers.
On YouTube one finds occasional films from the 1950s explaining and touting the virtues of particular work. Eg, I just saw one on Industrial Machinists: here’s what they do, here’s why they’re important to society, here’s how you learn to become one. It was shocking and outstanding that the society was making a modest effort to grow new workers in various needed professions. Needless to say, this kind of thing has gone by the wayside in the current age.
Power generation and transmission line work is well-protected from offshoring due to its regional nature. These careers are also protected from migrant labor and H1B outsourcing, because it is considered critical infrastructure work that requires background security checks.
It’s honest blue-collar work that still pays decent wages. I, and many of my colleagues, were able to support a family, raise children, and save enough for retirement on a single income – a difficult prospect for many careers.
Feeling a lot of schadenfreude reading this, that these capacity and labor issues (whether you do or don’t support data centers) all revolve around the shortsightedness and depthless greed of our elite class. I’m reminded of an line from economist Ha-Joon Chang around the beginning of trump’s tariff wars: “To me, America’s response looks like the tantrum of a country that was once the undisputed hegemon of the global economy now having been reduced to a much weaker position, largely thanks to the actions of its own capitalist class. Outsourcing, offshoring, and not investing are the causes of American relative decline.”
The average age of the electricians on my last job was 63. The divorce rate in my trade stands at 92 percent. Journeyman Industrial electricians usually have to travel long distances in order to find well paying jobs, hence the word journeyman. My last three gigs were fly in jobs, 14 days in and 7 days at home. 250k per year. Young men these days do not want to lose the conveniences of modern life and lose half their wages in divorce court for a pay check. A good electrician requires an apprenticeship plus ten years of experience before they are given a mission critical task. The west is way behind the ball in regards to training, there is a 10 year gap between expectations and reality that will start in a couple of years. I believe the shortage of tradesmen will continue until a cultural shift about the nature of physical work and family formation takes place.
Yeah, I know some field engineers with similar experiences. They can earn stupendous money, but the schedules and hours aren’t long-term compatible with a sane and enjoyable life. Most of them are younger and plan to drop back to something saner after a while, but I do know one “gnarly old veteran” who has done it for 40+ years. He seems to really enjoy it, but I’d personally go nuts living that way for too long.
Here is a highly knowledgeable person with a different view in regards to exactly how much less buildout is needed right away with just a very small amount of data center use adjustment. Worth a read.
And yes the grid even w/o AI/DC is growing and will need those skilled labor from all aspects of the electric grid from home to commercial to distribution to transmission to generation. Those of us in the trades have seen the lack of interest or support or need from the Dems for sure. Was it Obama, go learn to code?
https://open.substack.com/pub/energyempirepodcast/p/governors-are-finally-taking-an-approach?r=um4c&utm_medium=ios
Ah the irony: they want to replace workers with AI and yet they can’t really replace us at all, especially not with AI. I’m a somwhat qualified computer maintenaince technician, BTW, but, other than salary issues and guarantees of a stable job, etc., it seems to me that emigrating to the USA to keep the AI pipedream going for a bit longer would be impossible considering my stand on Palestine (and other issues). The USA has become a no-go zone as of late.
Most office workers do not have the mental acuity to become blue collar workers. A certain ability to adjust to the challenges of the job presented to ‘field engineers’ is just not present in those whose jobs depend on the ability to hide in a crowd.
I have had the pleasure (not) of having to work with ex office workers in the field who were there because ‘that’s where the money was. There was never a learning curve. It was a flat line. The trades require a certain type, as office work does. Stay in your lane. The trades usually pay better because they require an ability not present in the typical office worker.
Circa 2000, the first Internet boom led to telco linesmen making $110K in salary and overtime, so $214K in 2026 dollars.