AI, Jobs, and the Infrastructure Behind It
Global Economics Research Team
By: Patrick Gorman
Overview
Aggregate U.S. employment data shows no clear sign of AI-driven job losses so far, but entry-level workers in AI-exposed fields are seeing measurably worse outcomes than their more experienced peers.
The four largest U.S. hyperscalers are on pace to spend roughly $700 to $725 billion on AI infrastructure in 2026, up about 77% from 2025.
The electricity demand from that buildout is already showing up in household utility bills, with regulators only beginning to settle who pays.
Both trends share a common shape: capital and hiring behavior are moving well ahead of confirmed, broad-based economic effects.
AI in Employment Numbers
The headline numbers do not show a labor market in crisis. The U.S. economy added 57,000 nonfarm payroll jobs in June 2026, with unemployment at 4.2%, according to the Bureau of Labor Statistics. That is soft compared to expectations, but it is not a collapse. Broader academic work supports this reading. A large-scale study published by Anthropic's economics team in March 2026 found no detectable increase in unemployment among AI-exposed workers relative to the rest of the workforce since the launch of ChatGPT in late 2022. A separate Danish study covering 25,000 employees and 7,000 firms reached a similar conclusion on wages and hours, and Stanford's 2026 AI Index reports that large-scale aggregate job losses have not materialized in the overall employment data.
Beneath the aggregate numbers, several independent sources point to a more specific and more severe effect among young workers in AI-exposed fields. Stanford's Digital Economy Lab found that employment for workers aged 22 to 25 in AI exposed occupations sits roughly 19% below trend as of August 2026, while more experienced workers in the same fields show no comparable decline. Goldman Sachs' April 2026 analysis estimates AI is reducing net U.S. employment by around 16,000 jobs a month, concentrated in specific functions, and Challenger, Gray & Christmas found AI was explicitly cited as a factor in about 13% of U.S. layoffs so far in 2026, up from roughly 4.5% in 2025.
Hiring Is Repricing Faster Than Firing Is Rising
Perhaps the clearest signal is on the hiring side rather than the firing side. AI related skills now appear in roughly 2.5% of all U.S. job postings, up 55% year over year, and mentions of agentic AI skills specifically are up nearly 280% in a single year, according to Stanford and Lightcast data. That suggests employers are changing what they screen for well ahead of any confirmed wave of aggregate job losses, consistent with an augmentation first pattern rather than a displacement first one.
The open analytical question is whether the muted aggregate signal is simply a lag before broader displacement shows up, or whether AI's labor market effect is structurally concentrated in a narrow set of roles and career stages, especially new graduates. The entry level data is the strongest evidence so far of a real, measurable effect, and whether it stays contained to that group is worth tracking closely over the next several jobs reports.
Data Center Buildout
The four largest U.S. hyperscalers, Amazon, Microsoft, Alphabet, and Meta, are guiding toward a combined $700 to $725 billion in capital expenditure for 2026, up roughly 77% from about $410 billion in 2025. Amazon alone is projecting around $200 billion, most of its digital infrastructure. Alphabet has raised its guidance to as much as $205 billion, Microsoft is tracking toward roughly $190 billion, and Meta has raised its own guidance twice, to a range of $115 to $145 billion. Wall Street analysts already expect the group to cross $1 trillion in combined annual spending by 2027, and Goldman Sachs projects roughly $5.3 trillion in cumulative hyperscaler capex from 2025 through 2030.
This spending is running into a hard physical constraint in electricity. Gartner projects global data center power demand will rise 27% in 2026 alone, reaching 132 gigawatts, on its way to roughly 290 gigawatts by 2030. In the U.S. specifically, total data center energy demand is expected to nearly double between 2025 and 2028, from about 80 to 150 gigawatts, according to Bloom Energy. AI optimized servers are expected to account for 31% of data center power consumption in 2026 and to overtake conventional servers' power draw by 2027.
Who Pays for the Buildout
The cost of this expansion is increasingly showing up on household electricity bills rather than staying contained to the tech companies driving it. The independent market monitor for PJM, the grid operator covering 14 mid-Atlantic and Midwest states, attributed roughly $23 billion in customer price increases through 2028 primarily to data center demand. Academic modeling published in May 2026 by researchers at NC State, Carnegie Mellon, Pittsburgh, and Toronto projects a 6% to 29% national increase in electricity costs by 2030 attributable to data centers and cryptocurrency mining, with the hardest hit regions seeing increases as high as 57%. Virginia and Texas currently carry the largest projected data center electricity loads in the country.
Several major tech companies have signed on to a White House sponsored pledge aimed at limiting data centers' effect on residential electric bills, but the mechanics of who actually absorbs the infrastructure cost are set state by state, by public utility commissions, and the allocation is far from settled. This is likely to be one of the more consequential state level regulatory fights of the next several years.
Outlook
Both threads describe the same underlying dynamic from different angles, capital is moving into AI with speed and conviction that is running well ahead of confirmed, broad based effects on employment, and well ahead of any settled answer to who bears the infrastructure costs. That gap between how fast capital is deploying and how slowly the downstream effects are showing up in measurable data is the thing to watch over the coming year. Natural follow-on pieces include entry level labor market outcomes in the fall jobs reports, state level electricity rate case decisions, and hyperscalers capex to revenue ratios as more data becomes available.
Sources
Bureau of Labor Statistics, “Employment Situation,” June 2026.
Anthropic, “Labor Market Impacts of AI: A New Measure and Early Evidence,” March 2026. https://www.anthropic.com/research/labor-market-impacts
Stanford HAI, AI Index Report 2026; Stanford Digital Economy Lab, August 2026.
Goldman Sachs Global Investment Research, April 2026.
Challenger, Gray & Christmas, Q1 to Q2 2026 job cuts reports.
Gartner, “Data Center Electricity Consumption to Grow 26% in 2026,” June 2026. https://www.gartner.com/en/newsroom/press-releases/2026-06-10-gartner-says-data-center-electricity-demand-to-grow-26-percent-in-2026
Bloom Energy U.S. data center energy demand report, cited in Consumer Reports, January 2026. https://www.consumerreports.org/data-centers/ai-data-centers-impact-on-electric-bills-water-and-more-a1040338678/
PJM Interconnection / Monitoring Analytics customer cost findings, cited in Fortune, July 2026. https://fortune.com/2026/07/14/data-centers-23-billion-electricity-bills/
NC State, Carnegie Mellon, University of Pittsburgh, and University of Toronto joint academic modeling, May 2026.
Hyperscaler 2026 capex guidance, cited in CNBC, February 2026. https://www.cnbc.com/2026/02/06/google-microsoft-meta-amazon-ai-cash.html
Goldman Sachs hyperscaler capex projections, 2026.
Comments