Why Infrastructure, Not Skills, Will Shape the Future of Work

AI will not just augment workers; it will replace many with a few “superhuman” operators
The real divide is access to compute and energy, not worker readiness
Without new policy, AI will create structural labor redundancy

Generative artificial intelligence can change working patterns. It is estimated that it could automate approximately 25% of all work hours. This estimation changes how we should consider public policy. According to a report from the Federal Reserve Bank of St. Louis, workers using generative AI have reported saving 5.4 percent of their work hours in the previous week, showing a 1.1 percent increase in overall workforce productivity (Briggs & Kodnani, 2023). This number is often cited to argue for increased spending on education and training programs, all to prepare people for artificial intelligence. Yet, this idea misses a key consideration: getting ready is simply not about having specific skills. Being prepared also means having avenues. It is about access to money, processing power, reliable electricity, and applicable information. Job opportunities will remain limited if these components are not widely accessible. Instead of simply improving the general workforce, the likely result is job losses. Many positions might start to seem unneeded. A smaller group of highly skilled workers who have access to specialized information and data may be able to produce the same amount of work or more. This writing will discuss this issue. It will explain why just focusing on readiness is not enough. It will summarize the latest information on power usage, the extent of artificial intelligence use, and who is being hired. It will recommend specific policy modifications that treat processing power and electricity as important parts of the job market, not simply as resources for private companies.

The false idea of artificial intelligence readiness: why it hides the truth of division

The idea of preparing for artificial intelligence can easily be turned into training programs. These programs are beneficial and raise the skills of those who can use new tools. Yet, this idea hides two important facts. First, adopting artificial intelligence does not happen uniformly. It requires connecting with others and investing significant money. The operation of current generative models at scale requires specialized chips, engineering teams to integrate the models into daily operations, and reliable electricity supplies. Reports from the International Energy Agency show that data centers already consume a lot of energy and are likely to consume even more, which affects where computing investments are made (US data centers’ energy use amid the artificial intelligence increase, 2025). Companies will either pay more for electricity or simply avoid establishing large computing operations in areas where electric grids are not dependable. As a result, being ready depends on location and available infrastructure.

Second, readiness entails not just individual abilities, but also the resources that infrastructure can provide. Current studies that examine both technical skills and actual usage find that jobs most likely to be affected by artificial intelligence include repetitive positions, lower-paying jobs, and skilled roles such as programmers and financial experts. Instead of employing larger teams of mid-level employees, companies are likely to hire more experienced experts to manage artificial intelligence models, as artificial intelligence boosts the output of senior experts. This change results in fewer chances for new people to get a job and fewer options to receive the training needed to progress into higher-paying positions. The concept of readiness can obscure the fact that access to resources and money is now more important for prospects than completing just training.

Excellent output and the new employment market

Artificial intelligence acts as an effective multiplier when fully integrated. Research indicates large increases in output across areas such as customer service, coding, and marketing (The Economic Potential of Generative Artificial Intelligence: The Next Output Possibility, 2023). Examples show that these tools save time on repetitive tasks, improve problem-solving rates, and help skilled workers manage more work (Jeong et al., 2025). Therefore, small groups can now achieve equal or greater output than larger teams by using software and computing.

This can result in something called structural job redundancy. Instead of only changing parts of jobs, artificial intelligence can significantly reduce the need for certain roles within companies. Studies show only a small change in overall job losses so far. However, it is evident that fewer new hires are being made and fewer job postings are listed in areas where artificial intelligence is prevalent. These shifts can mean paths in a career can fade away quietly. This creates a risk of fewer chances to gain experience, rather than significant immediate job losses.

The inner workings of industries show this shift. Artificial intelligence writes and reviews software programs, which reduces the need for large teams of junior programmers. Artificial intelligence can write drafts and join together information previously handled by multiple helpers in legal and research fields. Small groups with computerized tools can expand their marketing plans across content and marketing. This replaces many freelance jobs and jobs open to new people. These impacts are centered in areas where companies have access to affordable electricity, information, and processing agreements. A division forms geographically as certain areas gain skills and money, while others lose jobs that require moderate skills.

A related worry is that the majority of profits tend to go to a small percentage of companies. Benefits become lasting for those who take the initiative when profits depend on exclusive models, private sets of information, and long-term computing agreements. Traditionally, advantages such as ports or railroads concentrate benefits and slow diffusion. The difference with artificial intelligence is how fast it grows; the combination of models, information, and computing expands and focuses payouts quickly.

What policies can do: electricity, access, and rethinking readiness

Labor policy must include infrastructure policy, as readiness is constrained by access to reliable electricity and computing resources. The most immediate issue is electricity. Data centers already account for a large share of electricity use in advanced economies. Estimates show that US data center consumption was around 183 terawatt-hours in 2024 (about 4% of the country's electricity). Projections show a sharp increase under the current situation (DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers, 2024). Without careful planning, this demand will determine where the top jobs and artificial intelligence investments are being made. The scale of the infrastructure challenge is evident in the growth of electricity demand.

Policy should be adjusted in three specific ways. First, base community backing and approval of large computing projects on observable investments in the grid and community. Land deals or tax help should be tied to responsibilities for funding grid improvements, ensuring resilience, and creating local job pipelines. That will transfer parts of the infrastructure costs back to the projects that create the demand and ensure that those who pay electricity bills are not the default supporters of private ability.

Second, provide basic computing access to everyone equally through shared hubs for public benefit. Small and mid-size companies are unable to purchase large-scale stacks. Publicly funded regional hubs can give safe access to pre-trained models and abilities. These hubs can be situated with development banks and community colleges. The hubs will be designed to lower the costs to begin, give local companies places to adopt artificial intelligence responsibly, and protect information, something that private companies often overlook.

Third, change labor support beyond short courses. Support needs to be available for wages, hiring, and credentials to lower re-employment costs. Public support should focus on areas less likely to be replaced by computerized systems, such as elderly care, community health, green retrofits, and local production. This will make stable local positions and career opportunities. These procedures need fast implementation. Government workers need to enforce transparency in infrastructure, leverage buying power, and supervise competition to ensure that opportunities powered by artificial intelligence are not centralized. They need to treat computing and power as vital infrastructure for the job market, not just private property. This will create conditions that enable broad, equitable participation in the artificial intelligence economy.

Common anticipated arguments state that new technologies will create new jobs and markets will reabsorb workers. Although this is partly correct, the current wave is different in three ways: (1) payouts heavily depend on the concentration of computing and information; (2) investments are large and mainly for regions of computing and power; and (3) evidence shows less entry for young workers in jobs that are commonly exposed. If not addressed, these differences can lead to persistent generational and spatial divides rather than rapid overall integration. According to a strategic plan from the U.S. Nuclear Regulatory Commission, as artificial intelligence becomes more prevalent in regulated industries, focusing solely on training may limit effective policymaking.

The danger of the artificial intelligence readiness idea is that policy is lowered to only training. This ignores the more significant factors that will determine who succeeds in the age of artificial intelligence. A better way to think about this is to treat computing, reliable electricity, and information as shared infrastructure for proper work. This thought process brings different choices, such as basing motivations for large projects on improvements to grids and the community, creating shared regional hubs for computing for small companies and colleges, changing labor supports to protect transitions and entry positions, and using competition policy and purchasing to limit companies from controlling basic inputs behind closed doors. These actions support opportunity and are not anti-innovation. Other options are a job market that rewards a few excellent workers and locations, while others lack the resources necessary to progress. The time to impact that outcome is now.

References

Briggs, J. & Kodnani, D., 2023. AI may start to boost U.S. GDP in 2027. Goldman Sachs.
Congruence Foundation Team, 2025. Artificial intelligence impact on the global job market (2025–2030). Congruence Foundation.
Federal Reserve Bank of St. Louis, 2025. The impact of generative AI on work productivity. Federal Reserve Bank of St. Louis.
Ghosal, S., 2025. Generative AI reshapes U.S. job market, Stanford study shows. CNBC.
Ghosal, S. & Butts, D., 2025. Generative AI reshapes U.S. job market, Stanford study shows. CNBC.
Goldman Sachs Global Investment Research, 2023. The potentially large effects of artificial intelligence on economic growth. Goldman Sachs.
International Energy Agency, 2024. Energy and AI: Energy demand from artificial intelligence. International Energy Agency.
International Energy Agency, 2025. Energy and AI. International Energy Agency.
Leppert, R., 2025. What we know about energy use at U.S. data centers amid the AI boom. Pew Research Center.
McKinsey Global Institute, 2023. The economic potential of generative AI: The next productivity frontier. McKinsey & Company.
Powell, L., 2025. China’s data centres: Watts behind the bytes. Observer Research Foundation.
U.S. Department of Energy, 2024. DOE releases new report evaluating increase in electricity demand from data centers. U.S. Department of Energy.