Insight

Data center facts vs myths (public-number edition)

The loud claims are usually national percentages, queue totals, or job promises. The public numbers are ranges, process clocks, and local constraints. Sources sit under each pair. We do not invent a seventh figure to make the chart smoother.

Myth

Data centers already use most U.S. electricity.

Fact

Recently published shares are in the mid-single digits, not a majority. The DOE/LBNL 2024 report framed 2023 use at 176 TWh, about 4.4% of U.S. electricity. The 2025 update — summarized by CRS report R49326 — cites about 192 TWh / about 4.7% in 2024. Growth scenarios for 2030 are wide: LBNL discusses about 9.5–15.3% of U.S. electricity (521–843 TWh band); EPRI discusses about 9–17% / roughly 380–790 TWh. Those are scenario bands, not a point forecast.

Source: LBNL 2024 U.S. Data Center Energy Usage Report (2023: 176 TWh / ~4.4%); CRS R49326 summarizing the 2025 DOE/LBNL update (~192 TWh / ~4.7% in 2024; 2030 share 9.5–15.3% / 521–843 TWh); LBNL 2025 update publication page; EPRI Powering Intelligence 2026 (2030: ~9–17% / ~380–790 TWh). Footnotes 1–4.

U.S. data-center share — published bands only

Source: Recent band is 2023 ~4.4% (LBNL 2024 report) through 2024 ~4.7% (LBNL 2025 update / CRS R49326) — two published years, not a reconstructed series. 2030 LBNL scenario share ~9.5–15.3%; 2030 EPRI ~9–17%. See also the TWh bands on the public arc of U.S. data-center power. Footnotes 1–4.

2030 U.S. data-center electricity — two published TWh bands

Source: 2024 177–192 TWh from EPRI 2026 (LBNL/CRS cite ~192 TWh at the top of that band). 2030 EPRI ~380–790 TWh; 2030 LBNL compounded-uncertainty band 521–843 TWh (CRS R49326). Two labs, two scenario families — not a single forecast. Footnotes 1–4.

Myth

Every proposed MW in a queue will energize.

Fact

Queues are claims. Eligibility, security, audits, and county votes cull them. ERCOT’s mid-2026 public contrast is the teaching example: more than 438,000 MW of proposed large-load requests against an 85,508 MW peak, with Batch Zero verification and a December 10 audit-report clock in August 2026 PUCT materials. A request total is not a peak and not an energized campus.

Source: ERCOT Trending Topics (June 18, 2026); Batch Zero Verification and Audit PUCT presentation (Aug. 20, 2026). Process context on who’s building now. Footnotes 5–6.

Myth

Data centers = crypto.

Fact

Different use classes. Some public processes separate them on purpose: ERCOT’s August 2026 verification materials discuss approvals to energize data centers or virtual-currency mining facilities as related but named classes, and LBNL’s energy-usage reports exclude cryptocurrency mining from their data-center TWh totals. Still verify the caption. A “computational load” row is whatever the issuing page says it is.

Source: ERCOT Aug. 20, 2026 PUCT presentation (data-center and virtual-currency mining facilities named separately); DOE/LBNL 2025 update (crypto mining excluded from the report’s data-center electricity totals). Footnotes 2 and 6.

Myth

They always create hundreds of permanent local jobs.

Fact

Construction and the supply chain are the heavy labor. Ongoing operations jobs are modest. Brookings research on first large facilities in a county suggests on the order of 100–200 jobs in a typical treated county over the first decade, depending on facility type — not a promise that every campus hires a factory-scale workforce. Confirm the fiscal-impact memo in the local packet rather than a national talking point.

Source: Brookings, “New evidence on data center employment effects” — research suggests roughly 100–200 jobs in a typical treated county. Not a Watt Brief census and not a finding for any one campus. Footnote 7.

Myth

Water use is always catastrophic — or always zero.

Fact

Cooling tech varies: evaporative, closed-loop, dry. Local capacity is the constraint, not a national slogan. Mid-2026 ERCOT community-impact work is the process example: public materials describe RFIs to large and medium loads and a Community Impact Review Report, including questions communities actually ask about water and related impacts. The fact is that the RFI exists. The withdrawal number lives on the local copy.

Source: ERCOT Batch Zero verification / community-impact presentation (Aug. 20, 2026) — RFIs and a December 10 Community Impact Review Report. Cooling technology is site-specific; do not flatten it to always-catastrophic or always-zero. Footnote 6.

Myth

Trade press is the first signal.

Fact

Permits and dockets often move 30–180 days before a polished recap. That lag is a teaching band, not a measurement of every campus. Watt Brief keys the row when the packet or the list moves. Estimate a heuristic window with the record → press lead-time estimator, and see from permit stack to press cycle.

Source: Watt Brief educational heuristic used on the lead-time tool and the permit-to-press insight — labeled as such, not a measured sample. Confirm the posted agenda.

Charts and facts on this page use cited public sources or labeled educational heuristics. They are not legal, engineering, or investment advice, and not a comps book. Confirm every number on the issuing agency or laboratory copy. Contact hello@wattbrief.com.

Sources

Free charts and calculators. The weekday brief is still the product.

See pricing All insights Tools Articles Sources