Research
The 200-gigawatt queue, and the much smaller industry already on the books
A queue the size of two grids
On August 15, Texas grid officials told regulators that Gov. Greg Abbott’s audit would cover about 250 to 300 projects, mostly data centers, representing some 200 gigawatts of claimed future demand—more than twice ERCOT’s peak record from the prior month. The broader interconnection queue already held more than 1,800 projects, about 90 percent of them data centers, totaling more than 474 gigawatts of potential load. Officials and developers alike said many of those filings would never be financed or leased. The queue was still large enough that ERCOT had stopped issuing demand forecasts.
Abbott’s line, repeated as Amazon, Anthropic, and others said they would comply, was that the centers must “pay their own way.” Wisconsin, where more than three-quarters of poll respondents said data-center costs outweigh benefits, turned the same fight into a governor’s race. North Dakota lawmakers heading into a September special session had drafts ranging from quarterly reporting to a temporary ban on new permits.
The political argument is about power, water, and who pays for transmission. A quieter question sits underneath it: how large is the industry that already exists? The Bureau of Economic Analysis tracks output in “data processing, hosting, and other information services,” a NAICS grouping that includes internet access, search, hosting, and related services—not data centers alone. Combined with state payroll counts, it is the closest official picture of whether the boom is already in the books or still mostly a claim on the grid.
Output has exploded. It is still a sliver of the state.
The output boom is not imaginary. Texas real GDP in data processing and hosting rose from $12.4 billion in 2019 to $29.7 billion in 2024, in chained 2017 dollars—a 140 percent increase, faster than the 111 percent national gain and a shade ahead of Virginia’s 134 percent. From 2015, Texas more than tripled. Wisconsin nearly doubled after 2019. Ohio rose 87 percent. California, already enormous, rose 91 percent and still dominates the map.
Size, though, is not the same as growth. In current dollars, California’s 2024 industry was $197 billion, about 37 percent of the U.S. total. Texas was $25.8 billion, or about 5 percent of the national industry and 0.93 percent of Texas’s own $2.77 trillion economy. Virginia’s $13.5 billion was 1.78 percent of state GDP—more specialized, still not a pillar. Wisconsin matched Texas on share (0.94 percent) with only $4.3 billion of output. North Dakota’s entire industry was $117 million, 0.15 percent of state GDP. If Bismarck is arguing about data centers, it is not because they already run the state.
That California figure is a warning label as much as a ranking. The BEA category folds in Silicon Valley cloud, search, and hosting work that has nothing to do with a shed full of GPUs on the prairie. Northern Virginia’s reputation as the world’s data-center capital does not show up as a GDP crown in this series; Texas has not “overtaken” Virginia in the official accounts, and California has not been in the race these governors are running. The useful comparison is among the states now legislating and campaigning over new halls: they have a fast-growing industry that is still, in every case, a rounding error in the state total.
Data-processing output has more than tripled since 2015
View exact chart values
| year | Texas | Virginia | Wisconsin | Ohio | North_Dakota | United_States | California |
|---|---|---|---|---|---|---|---|
| 2015 | 100 | 100 | 100 | 100 | 100 | 100 | 100 |
| 2016 | 101.8 | 99.4 | 96.6 | 105.4 | 125.2 | 113.2 | 121 |
| 2017 | 97.9 | 92.2 | 91.4 | 106.7 | 121.2 | 123.5 | 148 |
| 2018 | 102.1 | 91.3 | 100.8 | 111.1 | 95.2 | 137.2 | 164.7 |
| 2019 | 133.3 | 128.9 | 133.7 | 134.7 | 111.3 | 166.5 | 190.9 |
| 2020 | 154.2 | 143.4 | 139.2 | 137.6 | 147.4 | 193.5 | 223.5 |
| 2021 | 194.6 | 167.7 | 155.8 | 161.3 | 134.1 | 231.6 | 271.5 |
| 2022 | 255.9 | 224.5 | 230.8 | 204.7 | 187 | 269.5 | 277.3 |
| 2023 | 308.6 | 266.9 | 246.8 | 232.7 | 176.6 | 310.6 | 312.6 |
| 2024 | 319.4 | 301 | 265.6 | 251.6 | 162.5 | 350.9 | 365 |
California still dwarfs the data-center battlegrounds
View exact chart values
| state | gdp_billions | share_pct |
|---|---|---|
| California | 197.36 | — |
| Texas | 25.79 | 0.93 |
| Virginia | 13.55 | 1.78 |
| Ohio | 5.46 | — |
| Wisconsin | 4.28 | 0.94 |
| North Dakota | 0.12 | 0.15 |
Texas’s data-processing output grew faster than Virginia’s after 2019
View exact chart values
| state | real_pct_2019_2024 |
|---|---|
| Texas | 139.7 |
| Virginia | 133.6 |
| Wisconsin | 98.7 |
| California | 91.2 |
| Ohio | 86.8 |
| North Dakota | 45.9 |
The jobs arrived early, then stopped
Payrolls tell a less heroic story. Nationally, seasonally adjusted employment in computing infrastructure, data processing, web hosting, and related services rose from 345,100 in July 2019 to a peak of 488,600 in July 2023, then slipped to 463,500 by July 2026—a 5 percent decline from the peak, even as politicians talked about an accelerating buildout.
Texas looks like the national pattern with a cowboy hat on. Data-processing and hosting jobs, not seasonally adjusted, climbed from 34,100 in July 2015 to 36,300 in July 2019, then jumped to 48,100 by July 2022. They have barely moved since: 48,400 in July 2026, a 0.6 percent change over four years. Georgia, a smaller market, kept adding—from 14,500 in July 2019 to 25,000 in July 2026. The broader Texas information supersector, which includes publishing, telecom, and motion pictures as well as hosting, was almost unchanged over the same span (209,000 to 213,500) and down from a 2022 peak.
If you want the Texas payroll boom, look at the power plant. Electric power generation, transmission, and distribution jobs rose from 36,700 in July 2019 to 55,900 in July 2026, a 52 percent increase. Real GDP in Texas utilities rose 24 percent from 2019 to 2025; Wisconsin’s utilities barely budged (up 2.3 percent). Construction payrolls are up too—Texas 19 percent and Wisconsin 23 percent from July 2019 to July 2026, North Dakota only 6 percent—but construction is a large, mixed industry, and real construction output in Texas rose just 4 percent over 2019–2025. The data do not say data centers caused those extra hard hats. They do say the people getting hired in Texas’s digital-power story are more often line workers than rack technicians.
National data-processing jobs peaked in 2023 and have slipped
View exact chart values
| year | employees_thousands |
|---|---|
| 2010 | 242.6 |
| 2011 | 245.4 |
| 2012 | 254.5 |
| 2013 | 271.4 |
| 2014 | 279.2 |
| 2015 | 297.6 |
| 2016 | 303.7 |
| 2017 | 319.7 |
| 2018 | 331.5 |
| 2019 | 345.1 |
| 2020 | 365.7 |
| 2021 | 405.9 |
| 2022 | 468.8 |
| 2023 | 488.6 |
| 2024 | 480.3 |
| 2025 | 480 |
| 2026 | 463.5 |
In Texas, the payroll boom is at the power plant, not the server hall
View exact chart values
| year | data_processing_nsa_thousands | construction_sa_thousands | electric_power_nsa_thousands |
|---|---|---|---|
| 2015 | 34.1 | 685.5 | 35.7 |
| 2016 | 33.2 | 703.1 | 36.6 |
| 2017 | 33 | 710.4 | 36.6 |
| 2018 | 34.3 | 741.2 | 36.5 |
| 2019 | 36.3 | 776.7 | 36.7 |
| 2020 | 37.9 | 720.2 | 36.8 |
| 2021 | 41.1 | 739.4 | 37.6 |
| 2022 | 48.1 | 789.2 | 39.9 |
| 2023 | 48.3 | 824.5 | 44.2 |
| 2024 | 48 | 853.1 | 47.7 |
| 2025 | 48.6 | 904.2 | 51.5 |
| 2026 | 48.4 | 921.7 | 55.9 |
What the queue looks like beside the books
Put the 200-gigawatt audit next to the existing industry and the mismatch is the point. Texas already has a data-processing sector that, on BEA’s broader definition, is worth about $26 billion a year and employs about 48,000 people. It is growing in output much faster than in jobs, which is what you would expect if the incremental product is electricity-hungry capital rather than headcount. Compensation of employees still accounted for 48 percent of Texas’s 2024 industry GDP, and 37 percent in Wisconsin—this official category is not a pure server farm. The facilities in Abbott’s queue, if even a fraction of them connect, would be more capital-intensive than the average firm already in the series.
Wisconsin’s fight is even less about an incumbent industry. The state’s data-processing output is $4.3 billion, its information supersector has fewer jobs than in 2019, and construction hiring is the labor market that actually moved. North Dakota is arguing over permits for an industry that, in 2024, was smaller than a midsize factory. New Mexico, home to a proposed gas-fed AI campus, did post the fastest real-GDP gain in this set—up 285 percent from 2019 to 2024—from a base so small it was still under $1 billion.
None of this measures megawatts, water withdrawals, or who pays for 765-kilovolt lines. It does not prove that data centers raise household electric bills, or that they don’t. What the official accounts can say is narrower, and still useful: the boom governors are regulating is already visible in output, still tiny as a share of any state economy, no longer adding many data-processing jobs, and—in Texas—showing up more clearly on the utility payroll than in the server hall. The 200-gigawatt queue is a claim on a future grid. The books describe an industry that has already had its easy hiring years.
Sources and methods
This is retrospective enrichment using pinned catalogue snapshots, not an as-known-on-August-15-2026 reconstruction. Research extracts were retrieved on 2026-09-11.
BEA annual state GDP and income: release bea-regional-state-annual-2026-04-09, normalized snapshot created 2026-08-31. Detailed industry GDP for data processing, hosting, and other information services (NAICS 518-519) is observed for 2005–2024; 2025 cells exist but are missing. All-industry current-dollar GDP, utilities real GDP, and construction real GDP extend through 2025. Compensation of employees for the detailed industry ends in 2024. Shares of state GDP use current-dollar totals, not chained dollars. Real growth uses chained 2017 dollars. The industry grouping is broader than data centers; it includes internet access, search, hosting, and related services.
BLS Current Employment Statistics (national, seasonally adjusted) and State and Area Employment (Texas, Wisconsin, North Dakota, Georgia): snapshots dated 2026-08-24, monthly observations through July 2026. State data-processing employment is not seasonally adjusted; construction supersector employment is seasonally adjusted; Texas electric-power employment is not seasonally adjusted. July-to-July comparisons are used for NSA series. Duplicate period rows were not present in the extracts. Missing values were not treated as zero.
No megawatt, water, or interconnection-queue series are in the catalogue. Figures of 200 GW, 474 GW, and project counts come from August 15, 2026 reporting, not from these official accounts. Descriptive comparisons are not causal estimates.
Research completed 2026-09-11, for the August 15, 2026 news edition. This is retrospective analysis, not a reconstruction of information available that day.