What Changed In Washington

The United States has been moving from AI policy as software policy to AI policy as industrial policy. The difference matters. A software policy debate focuses on model safety, privacy, competition, bias, export controls, and government adoption. An industrial policy debate asks a harder physical question: who will build the factories, substations, power plants, fiber routes, switchyards, transformers, cooling systems, maintenance regimes, and 24-hour staffing models that make AI infrastructure possible?

Executive Order 14318 made that shift explicit. It says AI data centers and the infrastructure that powers them, including high-voltage transmission lines and related equipment, are part of a rapid buildout priority. It sets up a path for qualifying projects to receive financial support, categorical exclusions where legally available, FAST-41 transparency and covered-project treatment, streamlined environmental reviews, brownfield and Superfund-site reuse guidance, biological and water permitting efficiencies, and potential use of federal lands.

That is a permitting story, but it is also a sequencing story. If a data center project needs more than 100 MW of new load, the people problem begins long before the facility is energized. Engineering studies, environmental work, site preparation, utility coordination, civil construction, high-voltage electrical work, cooling installation, commissioning, cybersecurity, facilities operations, and emergency coverage all have to move in sequence. A shortage in one trade can become a delay for the entire project.

FERC's June 18, 2026 action adds the grid-market layer. The Commission issued tailored show-cause orders under Section 206 of the Federal Power Act to PJM, MISO, SPP, CAISO, ISO-NE, and NYISO. These operators cover much of the organized wholesale power market footprint in the United States. FERC asked them to address five categories of reform: more efficient transmission service applications and studies, protection against cost shifting, better transparency into transmission costs, accommodation of co-location and behind-the-meter generation, new transmission services for flexible large loads, and study processes for generating facilities that serve electrically proximate or co-located loads.

The order does not say that every AI data center should jump the line regardless of impact. FERC emphasized consumer safeguards, regional differences, state authority over retail electricity rates and generation siting, and a need to prevent cost shifting. But the regulatory message is unmistakable: the old pace for large-load interconnection is not good enough for the scale of demand now arriving.

The Energy Demand Behind AI

The policy urgency is rooted in electricity demand. AI is unusually power-intensive because the work is concentrated in large, specialized compute facilities that run high-density servers, advanced chips, cooling equipment, backup systems, and network infrastructure. Training frontier models can require enormous compute bursts; inference spreads the demand across constant production use as customers use AI in search, software, customer service, manufacturing, medicine, coding, design, defense, and everyday business operations.

The 2024 Lawrence Berkeley National Laboratory report prepared for the U.S. Department of Energy estimated that U.S. data centers consumed about 176 TWh in 2023, or about 4.4% of total U.S. electricity use. The same report projected U.S. data center consumption could reach 325 to 580 TWh by 2028, equal to roughly 6.7% to 12% of national electricity consumption. Even the low end of that range is a major infrastructure planning problem. The high end is a structural change in how the country allocates electricity.

The International Energy Agency's 2025 Energy and AI report puts the issue in global perspective. It estimated that data centers used around 415 TWh worldwide in 2024 and projected that consumption would more than double to about 945 TWh by 2030. The IEA also found that the United States accounts for the largest share of global data center electricity consumption and that data centers account for nearly half of U.S. electricity demand growth through 2030 in its base analysis.

Those numbers help explain why a data center is no longer just a real estate asset with a server hall inside. At 100 MW, a single project can be comparable to the load of a small city. Larger campuses and multi-phase buildouts can require hundreds of megawatts or more. That kind of load forces difficult choices about generation, transmission, distribution equipment, rate design, standby capacity, cooling water, local land use, noise, emissions, reliability, and who pays for network upgrades.

There is another timing mismatch. The IEA notes that building new transmission lines in advanced economies can take four to eight years and that wait times for critical grid components such as transformers and cables have doubled in recent years. AI infrastructure developers often want power in months or a few years. Apprenticeships for electricians can take four or five years. Transmission projects can take longer than the facility they serve. That mismatch is why "speed to power" cannot be solved by policy language alone.

The Infrastructure Stack

Fast-tracking AI infrastructure onto the grid is sometimes described as if it were a single connection request. It is not. The order and FERC action both point to a broader stack. A data center project can depend on power plants, pipelines, substations, switchyards, transformers, switchgear, backup power, water systems, fiber, roads, site security, environmental reviews, interconnection studies, tariff design, and workforce availability.

Employers that want to participate in this buildout need to understand where their labor sits in the stack. A general contractor may focus on site prep and building shell. An electrical contractor may own medium-voltage and low-voltage work inside the facility. A utility may own transmission studies, line upgrades, substation interconnection, and system protection. A mechanical contractor may handle cooling systems. An operations contractor may supply 24-hour facilities teams. A training provider may be responsible for turning entry-level workers into credentialed electricians, HVAC technicians, lineworkers, safety professionals, or data center technicians.

The workforce problem is not simply "hire more people." It is "hire the right people at the right phase, document the right credentials, schedule them without burning them out, allocate labor to the right project, and produce clean time and payroll records even when work is mobile, remote, unionized, prevailing-wage, multi-state, or tied to grants and public incentives."

The Workforce Gap

The workforce bottleneck is already visible in federal labor projections. The Bureau of Labor Statistics projects electrician employment to grow 9% from 2024 to 2034, much faster than average, with about 81,000 electrician openings each year. It also notes that most electricians learn through apprenticeships and that many apprenticeships run four or five years with about 2,000 hours of paid on-the-job training per year.

That is the first pressure point. The United States can change permitting timelines much faster than it can produce fully licensed electricians. If a regional AI infrastructure wave needs high-voltage electrical crews in 2027 or 2028, the apprenticeship pipeline needed to start years earlier. Pulling experienced electricians from housing, manufacturing, hospitals, schools, utilities, and industrial maintenance may solve one project while creating shortages elsewhere.

HVAC and cooling are another pressure point. BLS projects heating, air conditioning, and refrigeration mechanics and installers to grow 8% from 2024 to 2034, with about 40,100 annual openings. AI facilities may use sophisticated air cooling, liquid cooling, chillers, pumps, controls, heat exchangers, and high-density thermal management. These are not optional comfort systems. Cooling is part of the compute supply chain.

Power-line installers and repairers are a third pressure point. BLS projects this occupation to grow 7% from 2024 to 2034, with about 10,700 annual openings. These workers install and maintain the power grid itself. They handle serious high-voltage hazards, storm response, maintenance, distribution and transmission equipment, and work that often cannot be done remotely or cheaply replaced.

When those three categories are combined, the economy already needs more than 131,800 openings per year on average across electricians, HVAC technicians, and power-line installers before counting engineers, welders, equipment operators, cyber analysts, safety managers, plant operators, control-room staff, project managers, environmental specialists, security officers, fiber technicians, and data center operations staff.

The Hiring Clock Is Already Running

AI infrastructure schedules are unforgiving because the construction clock, grid clock, and training clock run at different speeds. Developers may want to energize a facility quickly. Grid operators may need studies, tariff changes, and network upgrade plans. Utilities may face equipment queues for transformers, cables, turbines, and other hardware. Training providers may need years to move a new worker from interest to competency.

The practical lesson for employers is to plan backwards from energization. If a project needs fully credentialed crews in 2028, waiting until 2028 to recruit is not a plan. The plan has to include immediate hiring for project managers and supervisors, near-term pre-apprenticeship and apprenticeship intake, mid-term cross-training for incumbent workers, and long-term retention for critical roles that cannot be replaced quickly.

Employers also need to prepare for wage pressure and schedule pressure. When data centers, utilities, manufacturing plants, housing projects, transportation agencies, and industrial contractors all chase electricians and technicians in the same region, pay rates rise, overtime rises, and poaching risk rises. Without disciplined scheduling and time tracking, the result can be fatigue, safety incidents, cost overruns, payroll disputes, and weak project cost data.

The Training Model The U.S. Needs

The AI Action Plan calls for a national initiative to identify high-priority occupations essential to AI infrastructure, create skill frameworks and competency models, support industry-driven training, expand pre-apprenticeships, modernize CTE pathways, and expand Registered Apprenticeships. That is the right frame because the labor need is too broad for one employer, one school, or one state to solve alone.

A serious training model should be regional, employer-led, and evidence-based. Regional matters because the grid is regional, data center clusters are regional, and commuting distance matters for trades. Employer-led matters because training that does not connect to real hiring demand becomes a classroom exercise. Evidence-based matters because public and private funding should go to programs that can show completions, placements, retention, wage progression, safety outcomes, and project readiness.

The best model is not to wait for workers to appear. It is to build a pipeline with clear stages: awareness, pre-apprenticeship, apprenticeship, supervised field experience, specialty credentialing, journey-level progression, supervisor development, and ongoing upskilling. Each stage should have a named employer outcome. If a student finishes a pre-apprenticeship, what job can they interview for? If an incumbent HVAC technician completes liquid-cooling training, what new assignment can they take? If a journey electrician completes data center commissioning training, what premium work can they perform?

Employer Risks In The AI Infrastructure Rush

When federal policy accelerates infrastructure, employers often feel the opportunity first and the administrative burden second. That is risky. The faster projects move, the more important it becomes to manage labor records, safety documentation, shift rules, training status, payroll compliance, contractor visibility, and project cost allocation from the beginning.

Energy infrastructure work can involve remote sites, mobile crews, multiple employers, union dispatch, apprenticeship ratios, public funding, grants, incentives, environmental obligations, security requirements, and overtime-heavy schedules. A company that uses informal spreadsheets, manual timesheets, or disconnected payroll systems may be able to survive during a small project. It will struggle when multiple crews are spread across substations, data center shells, cooling equipment, and commissioning windows.

Where TimeTrex Fits

TimeTrex is not an energy permitting platform and it does not build substations. Its role is more practical: helping employers manage the people who make fast infrastructure schedules possible. Contractors, utilities, manufacturers, cooling specialists, security providers, and data center operators need workforce systems that can keep up with mobile crews, changing shifts, cost centers, overtime, training, attendance, scheduling, payroll, and reporting.

For AI infrastructure employers, the operational problem is often the gap between the project schedule and payroll reality. A crew moves from a switchyard to a data hall. A technician splits time between maintenance and commissioning. An apprentice needs hours documented toward progression. A supervisor needs to approve time before payroll closes. A project manager needs labor cost by phase. HR needs to know which workers are trained for which site. Finance needs clean data. The workforce system has to connect those dots without turning every pay period into detective work.

A Practical Employer Playbook

The policy environment is moving toward faster infrastructure approvals and faster large-load integration. Employers do not control every regulatory or grid decision, but they do control their workforce readiness. The following playbook is designed for organizations that may hire, train, schedule, subcontract, or operate crews tied to AI data centers and the energy infrastructure around them.

FAQ

Sources

This article uses official policy, energy, and labor-market sources wherever possible. Source links open in a new tab.