THE DIGITAL LOAD: Data Center Energy, Infrastructure & Policy Intelligence

A Production of Hedley & Company | Charleston, W.Va.
Vol. 1, No. 1 | Week Ending August 23, 2026 | Data: Week 34 (Aug. 17-23, 2026)

FAST FACTS

Note on data cadence: unlike EIA's weekly coal production series, no single federal agency publishes a weekly national data-center production or absorption report. Figures above are drawn from the most recently published data as of this issue: FERC docket filings dated as shown; corporate and market data (CBRE, JLL, Avison Young) reflect Q2 2026, the most recent full-quarter market data available; hyperscaler capex reflects Q2 2026 earnings-call guidance. Reporting lag ranges from same-week (FERC) to one full quarter (commercial real estate data). Where this issue cannot obtain a true weekly figure, it says so rather than estimating one.

EXECUTIVE SUMMARY

The interconnection reckoning FERC set in motion in June came due this week. Under Section 206 show-cause orders issued to all six FERC-jurisdictional grid operators — PJM, MISO, SPP, CAISO, ISO-NE and NYISO — August 17 was the deadline for each to either defend its current large-load tariff as just and reasonable or propose reforms. PJM, which carries the heaviest data-center load growth of the six and projects 32 gigawatts of peak demand growth by 2030, is the filing every other region will be measured against. A related governance docket, AD26-7, set August 21 as the deadline for post-technical-conference comments, with FERC's Alternative Dispute Resolution forum opening September 1. Separately, PJM's proposed $20 billion capacity backstop procurement (Docket ER26-3380) remains pending with no ruling yet, following a second consecutive Base Residual Auction clearing at the $325/MW-day price cap and a 6,831-megawatt reliability shortfall.

On the ground, the story split two ways this period. Developers kept announcing multi-gigawatt, multibillion-dollar campuses at a pace that shows no sign of slowing — OpenAI's $20 billion Project Camellia in Effingham County, Georgia and Meta's sixfold capex increase on its El Paso campus to $10 billion both illustrate how fast individual projects are scaling past anything the grid was designed for. At the same time, local governments from Texas to New Jersey kept passing moratoriums and bans that are now generating real litigation exposure: Hill County, Texas rescinded the state's first county-level moratorium after a developer sued for more than $100 million, and similar suits are now active in Michigan, Tennessee and New Jersey. More than 30 North Carolina jurisdictions alone have adopted or floated some version of a pause.

A Bloomberg analysis published August 18 quantified something the industry had downplayed for two years: 99 tracked data-center-dedicated gas plants could add roughly 318 million metric tons of CO2 annually — about 20 percent of current U.S. power-sector emissions — if run at typical utilization rates. And underneath all of it, the real ceiling on this industry's growth is neither chips nor capital. Demand for generator step-up transformers grew 274 percent between 2019 and 2025, lead times on large high-voltage units now run past three years, and industry analysts estimate 30 to 50 percent of the roughly 16 gigawatts of U.S. data center capacity announced for 2026 is at real risk of delay or cancellation — driven by equipment shortages, not financing or GPU supply.

Five things to know this week: FERC's national interconnection clock ran out on August 17; the local moratorium wave has become a litigation wave with real financial exposure on both sides; hyperscaler capex is tracking toward $630 billion for 2026 even as Wall Street questions the underlying GPU depreciation math; a Georgia mega-campus and a West Texas gas-plant permit both illustrate how fast individual projects are scaling past anything the grid was designed for; and the transformer and switchgear shortage — not GPU supply — is now the hard ceiling on how much of this actually gets energized on schedule.

SECTION 1 | NATIONAL DATA CENTER DEVELOPMENT PIPELINE

Source: Data Center Knowledge, Data Center Daily Notes (Strisker), Blackridge Research, Barchart, DediRock, DOE, TCEQ. Compiled from project disclosures dated within the past 30 days as of Aug. 23, 2026. N/D = not disclosed by developer or reporting source at time of publication.

Pipeline Analysis

Every project in the table above disclosed its megawatts and its capital commitment before it disclosed permanent headcount, and three of the eight did not disclose permanent headcount at all as of this issue. That pattern is not necessarily deceptive — permanent staffing on a hyperscale campus is genuinely small relative to its physical and capital footprint, typically 50 to 200 people regardless of size (see Section 5) — but it means the jobs figure most often cited in a county commission press conference is the construction-phase number, which is real payroll for two to four years, not a permanent economic base. Communities and reporters covering individual project announcements should request both figures separately and in writing before treating either as the headline number.

The clearest structural signal in this period's pipeline is the widening gap between projects funding their own generation and transmission (SB Energy/AEP Ohio, Meta's El Paso pledge) versus those relying on private, off-grid generation entirely (GW Ranch). The former model, however imperfect, at least creates the possibility of new capacity benefiting the broader grid; the latter removes the project from public interconnection queues but also removes any obligation to share the resulting generation. Regulators and communities negotiating with developers should treat this distinction as a primary term of any agreement, not an afterthought.

SECTION 2 | POWER PROCUREMENT & GENERATION ECONOMICS

Source: American Action Forum; U.S. DOE; PJM Interconnection; company disclosures. Dates as shown per deal.

Power Market Commentary

Gas remains the default answer to the industry's power problem, and the West Texas GW Ranch permit — at up to 7.7 gigawatts, the largest approved gas-fired power project in U.S. history — was built explicitly to serve a private grid rather than the public system. NextEra's separate 10 gigawatts of combined new gas capacity in Texas and Pennsylvania connects to the public regional transmission grids instead, a meaningfully different cost-and-benefit structure that policy discussions frequently blur together under the single heading of "new gas capacity." The distinction matters directly for ratepayers: capacity that interconnects to the public grid is at least theoretically available to non-data-center customers; behind-the-meter capacity is not.

PJM's capacity market result is the dominant power-price signal of the period. A second consecutive Base Residual Auction clearing at the $325-per-megawatt-day price cap, with the entire RTO falling 6,831 megawatts short of its reliability requirement, is not a one-time event — it is the clearest available evidence that PJM's capacity base has not kept pace with load growth driven overwhelmingly by data centers. PJM's pending $20 billion backstop procurement (ER26-3380) proposes a one-time reliability auction at a $555-per-megawatt-day cap to cover the shortfall; whoever pays for that backstop, developer or ratepayer, is the single largest open cost-allocation question in the entire national data-center policy debate right now.

SECTION 3 | GRID RELIABILITY, FERC DOCKETS & LARGE-LOAD INTERCONNECTION

Source: FERC docket filings (RM26-4-000; EL26-67 through EL26-72 series; AD26-7-000; ER26-3380); PJM Interconnection public filings.

Grid Analysis

FERC's June 18 show-cause orders directed all six FERC-jurisdictional RTOs and ISOs to either justify their existing large-load interconnection tariffs as just and reasonable or propose reforms addressing five common categories the Commission identified. The orders followed a Department of Energy directive last October under Section 403 of the DOE Organization Act, calling for standardized rules on how loads above 20 megawatts — including co-located generation arrangements — connect to the transmission system. August 17 marked the 60-day deadline for each RTO and its transmission owners to file. PJM's response is the one every other state commission and developer nationally will be measuring their own region against, given PJM's outsized share of national data-center load growth and its two consecutive capacity-auction shortfalls.

The core policy question FERC is resolving is cost causation: whether the load that triggers a network upgrade pays for it directly, or whether that cost gets socialized across every ratepayer in the footprint. Whatever gets filed in response to the show-cause orders will set that answer region by region. This publication will carry a full docket-by-docket breakdown of the PJM and MISO filings once they are available in the public record.

SECTION 4 | REAL ESTATE MARKET: LEASING, VACANCY & ABSORPTION

Source: CBRE Global Data Center Trends 2026; Avison Young Q2 2026 U.S. Data Center Market Overview; JLL North America Data Center Report, Midyear 2026.

Real Estate Analysis

National data center vacancy fell below 1 percent in the second quarter of 2026 — the lowest ever recorded — even as U.S. inventory grew nearly 80 percent year-over-year. That combination should be structurally impossible in a normal commercial real estate market; it is possible here because almost nothing coming online is speculative. Roughly 80 percent of capacity currently under construction nationally is already pre-leased, and JLL reports most tenants signing today are locking in space for 2028 delivery. Northern Virginia, still the largest single market in the country, posted 0.3 percent vacancy. Atlanta posted the steepest vacancy compression of any major market, falling from 3.6 percent to 1.0 percent.

Financing structures are shifting to match. Data-center-backed securitizations — asset-backed and commercial mortgage-backed structures collateralized by leases and equipment — reached $17 billion in the first half of 2026, up 29 percent from a year earlier. That is institutional capital underwriting the assumption that current demand is durable rather than speculative, a useful counterpoint for any argument that this is a bubble on the verge of correcting.

SECTION 5 | EMPLOYMENT & ECONOMIC FOOTPRINT

Source: Associated General Contractors of America; Brookings Institution; NetChoice/Commonwealth Economics Partners (Kentucky); Northern Virginia Technology Council; U.S. Census Bureau construction spending data.

Employment Analysis

The honest read on data-center employment cuts against both sides of the public debate simultaneously. Critics are correct that permanent operating employment at a hyperscale campus is genuinely modest — typically 50 to 200 people regardless of the facility's capital scale — and that press materials citing large "jobs created" figures are almost always blending in the temporary construction-phase workforce. But proponents are also correct that the construction-phase impact is real, not manufactured: national data center construction spending hit $59.3 billion on a seasonally adjusted annual basis in May 2026, up 23 percent year-over-year, and now accounts for roughly 8.0 percent of all private nonresidential construction spending nationally. Construction wages in this sector are rising faster than the private-sector average and already run about 20 percent above it.

Tax revenue is frequently the more durable and better-documented local benefit than either jobs figure. Loudoun County, Virginia's data center corridor generates more than $600 million a year in local tax revenue on a cumulative, multi-decade basis. A single Ohio project has been modeled to produce roughly $84 million in annual peak state and local tax collections. A typical 400-megawatt Kentucky project is projected to generate $290 million to $435 million in local tax revenue over a 12-year window. Communities negotiating incentive packages should treat enforceable, clawback-backed tax and job commitments as the more reliable metric than either headline jobs number.

Modeled National Construction Labor Footprint — Methodology Note

Using AGC's cited Ohio case study as a per-worker output benchmark ($2.4 billion total project economic output supported by approximately 9,700 peak construction jobs, or roughly $247,400 in output per construction worker), and applying that ratio to the $59.3 billion national annualized data-center construction spending run rate reported for May 2026, yields a modeled national estimate of approximately 240,000 FTE-equivalent construction workers supported industry-wide at the current build pace. This is an extrapolation from a single named project's output-per-worker ratio applied nationally, not a direct headcount measurement, and it should be read as a directional order-of-magnitude estimate rather than a precise employment count. A materially different capital-to-labor mix in other states' projects would change this estimate meaningfully; readers using this figure for planning purposes should treat it as illustrative rather than authoritative and verify against state-specific labor data where available.

SECTION 6 | SUPPLY CHAIN: THE TRANSFORMER AND SWITCHGEAR BOTTLENECK

Source: Wood Mackenzie transformer survey data as cited in Inflect and The Geopolitical Economist; Sightline Climate/Bloomberg pipeline-delay estimates; company capacity disclosures (Cummins).

Supply Chain Analysis

Every generation deal and every hyperscaler capex commitment in this issue is contingent on physical equipment that does not yet exist in sufficient quantity. Demand for generator step-up transformers grew 274 percent between 2019 and 2025, and demand for substation transformers rose 116 percent over the same period — both figures far outpacing domestic manufacturing capacity. Lead times that ran 24 to 30 months before 2020 now stretch past three years for large high-voltage units, with some reporting sources citing ranges as wide as 80 to 210 weeks depending on unit size and manufacturer backlog. Cummins is reportedly sold out of high-horsepower gensets through 2028.

The consequence is quantifiable at the pipeline level: of roughly 16 gigawatts of U.S. data center capacity announced for 2026, only about 5 gigawatts was actually under construction as of mid-2026, and analysts at Sightline Climate estimate 30 to 50 percent of the broader announced pipeline is at real risk of delay or cancellation — driven by transformer and switchgear shortages rather than financing or chip availability. A $2 billion campus can sit dark waiting on a $40 million transformer order. This is the single most important fact in this entire report for anyone modeling how much of the announced pipeline actually gets energized on the timeline developers are promising.

SECTION 7 | DOMESTIC DATA CENTER INDUSTRY NEWS

FERC's National Show-Cause Deadline Passes for All Six RTOs/ISOs

Federal Energy Regulatory Commission, Docket Nos. EL26-67 through EL26-72 | Aug. 17, 2026

WHY IT MATTERS: This is the largest single regulatory event of the period. Whatever PJM, MISO and the other four grid operators filed in response determines, region by region, whether data center developers or the broader ratepayer base pays for new transmission and generation triggered by large-load interconnection. THE DIGITAL LOAD is pulling the actual filings now and will carry a full breakdown once available in the public docket.

Bloomberg: Data-Center Gas Plants Could Boost U.S. Power Emissions 20%

Bloomberg News, citing BloombergNEF data | Aug. 18, 2026

WHY IT MATTERS: This converts a two-year-old activist estimate into a quantified, citable industry number — 99 tracked plants, roughly 318 million metric tons of CO2 annually at typical utilization. Expect this figure cited in every state air-permitting fight and hyperscaler sustainability-report footnote from this point forward.

Texas: Hill County's $100 Million Moratorium Lesson

MultiState Insider | Aug. 19, 2026

WHY IT MATTERS: Hill County rescinded the state's first county-level data center moratorium after a developer sued for more than $100 million in damages, prompting neighboring Tom Green County to abandon a similar plan. This is now the cautionary case every municipal attorney nationally cites when a client proposes a fast, reactive moratorium.

Texas: Gov. Abbott Orders Statewide Grid-Connection Audit

MultiState Insider | Aug. 2026

WHY IT MATTERS: The audit of every data center connecting to the Texas grid creates a de facto statewide pause pending review — a significant development for the nation's fastest-growing data-center interconnection market and a wildcard for every pending ERCOT-territory project.

Michigan: Gibraltar and Wixom Sued Over Targeted Moratoriums

Govtech; Click On Detroit | Aug. 19, 2026

WHY IT MATTERS: Developers argue both moratoriums were written specifically to stop already-pending applications. Courts have a long history of striking down ordinances targeted at a single named applicant while upholding generally applicable rules adopted in advance — the distinction is the whole case for both suits.

Tennessee: DC BLOX Sues Nashville Over Moratorium and Buffer Ordinance

Brushwood Media Network / The Center Square | Aug. 12, 2026

WHY IT MATTERS: The federal suit tests whether a proximity ordinance passed after a specific application was filed can survive constitutional scrutiny. The dispute drew bipartisan local opposition to the specific site, including public criticism from Republican gubernatorial nominee Sen. Marsha Blackburn.

Louisiana PSC Votes to Keep Meta's Richland Parish Facility Details Confidential

Data Center Moratoriums Weekly Tracker | Aug. 13, 2026

WHY IT MATTERS: A 3-1 vote shielded project details despite an administrative law judge's own subpoena — a live transparency fight over how much ratepayers funding nearby infrastructure ever get to see about the project driving the bill.

SECTION 8 | INTERNATIONAL DATA CENTER INTELLIGENCE

Global Vacancy Diverges Sharply by Region

CBRE Global Data Center Trends 2026; JLL EMEA Data Centre Mid-Year 2026 Report

WHY IT MATTERS: European colocation vacancy across FLAP-D markets held at 6.4% in Q2 2026, roughly six times the U.S. national rate, while APAC vacancy varies enormously by market — Hong Kong at 18% (down from 28% a year earlier) against Singapore at 2%. U.S. developers and capital increasingly view the domestic power-constrained market as tighter, not looser, than most international alternatives.

Meta's First Canadian Campus Signals Siting Diversification

Data Center Knowledge, New Data Center Developments: August 2026

WHY IT MATTERS: The $9 billion-plus Sturgeon County, Alberta project is Meta's first Canadian data center. Hyperscalers actively diversifying siting outside the U.S. interconnection queue is a pressure valve worth watching as domestic permitting and transformer delays stretch on.

AWS Commits $48 Billion to India Through 2030

Data Center Knowledge, New Data Center Developments: August 2026

WHY IT MATTERS: The commitment includes $21 billion allocated specifically to cloud and AI infrastructure from 2026-2030, alongside workforce development programs — a signal that hyperscaler capital deployment is not solely a U.S. grid-constraint story but a global capacity race.

SECTION 9 | LEGISLATIVE & REGULATORY DEVELOPMENTS

Industry Action key: MONITOR = track for developments; COMMENT = formal comment period open; FILE = filing deadline applies directly.

SECTION 10 | SWOT ANALYSIS & STRATEGIC OUTLOOK

Strategic Outlook

This period's data draws a sharper line than usual between the data-center industry's two live fault lines, and developers, communities and policymakers should read them separately rather than blending them into a single national narrative. On the regulatory-cost side, the question FERC is resolving through its six show-cause dockets is fundamentally about who pays: whether the load that triggers a network upgrade bears that cost directly, or whether it gets socialized across every ratepayer in the footprint. PJM's back-to-back capacity-auction shortfalls and its pending $20 billion backstop procurement make this the single most consequential open question in the entire sector heading into the fourth quarter.

On the local-political side, the moratorium-to-lawsuit pipeline is accelerating faster than most industry participants have priced in. More than 30 North Carolina jurisdictions alone have adopted or floated some version of a pause, and the financial exposure on the government side is no longer theoretical — Hill County's nine-figure damages exposure is a fact pattern now being cited nationally by municipal attorneys advising clients against fast, reactive ordinances. States writing enforceable, litigation-resistant tariff structures (Ohio, Kentucky, Colorado, Minnesota) are proving to be a more durable policy tool than outright bans, which continue to lose in court when adopted in a panic rather than through orderly process.

The single most important physical-world constraint in this issue is not a market or legal signal at all — it is a piece of hardware. The transformer and switchgear shortage detailed in Section 6 is the hard ceiling on how much of the current pipeline actually gets energized on schedule, and it will remain the binding constraint on this industry regardless of how favorably the FERC dockets or state legislatures resolve. Developers, utilities and communities negotiating project timelines should treat announced in-service dates with real skepticism until equipment procurement status is confirmed directly.

Week's Bottom Line

The data-center buildout remains structurally under-supplied by almost every real estate measure — sub-1% national vacancy, 80% pre-leasing, and record capex — even as its two biggest headwinds intensified this period: a national regulatory reckoning over who pays for the grid upgrades this growth requires, and a fast-metastasizing local moratorium-and-lawsuit pattern that is proving expensive for governments that move in haste. The transformer and switchgear shortage detailed in Section 6, not GPU supply or capital availability, is now the most reliable predictor of which announced gigawatts actually get built on schedule and which stay speculative interconnection requests.

SECTION 11 | PUBLISHER’S VIEW

The Fleet We Already Built

Every conversation about closing PJM's capacity gap starts in the wrong place. It starts with a new gas turbine that needs a transformer nobody can get for three years, or a small modular reactor that won't produce a kilowatt-hour until the next decade, or an interconnection queue that treats a shovel-ready project like it just showed up yesterday. Nobody starts where the fastest, cheapest, already-permitted capacity actually sits: idling.

PJM's coal fleet, as of the most recent full accounting on December 31, 2025, carries 38,366 megawatts of installed nameplate capacity. That is not a rounding error next to PJM's own reliability shortfall. It is more than five and a half times the 6,831-megawatt gap the RTO's 2028/2029 Base Residual Auction just failed to fill for the second consecutive cycle. The plants are permitted, interconnected, staffed, and — this is the part the "coal is dying" narrative keeps skipping — largely still standing. What they are not doing is running anywhere close to what they are capable of.

The most recent EIA figure on PJM coal utilization, drawn from 2023 generator data, put the fleet's capacity factor at 34 percent, down from 56 percent as recently as 2013. Run the arithmetic on what that gap means in megawatt-hours, and the number should stop every FERC commissioner and every state utility regulator cold. At 38,366 megawatts of nameplate capacity running 8,760 hours a year, the fleet's full theoretical output is about 336 million megawatt-hours annually. At a 34 percent capacity factor, actual dispatched output comes to roughly 114 million megawatt-hours. A well-maintained coal steam unit can sustain a capacity factor in the 80-to-85 percent range without heroics — the range baseload coal ran at for decades before gas got cheap. Call 85 percent a realistic, sustained ceiling, and the fleet's achievable output is closer to 286 million megawatt-hours a year. That leaves roughly 171 million megawatt-hours sitting on the table annually — call it 3.3 million megawatt-hours a week, or an average of 19,600 megawatts of continuous generating capacity that exists, is paid for, and simply is not being asked to work.

Nineteen point six gigawatts. That number alone is larger than the entire announced data-center capacity currently under construction nationwide. It would not require a single new interconnection agreement, a single new transformer order, a single new mile of transmission right-of-way. It requires PJM's own dispatch stack to call on plants that are already sitting there, fueled and staffed, waiting for a phone to ring.

Here is why the phone mostly does not ring, and it is worth explaining plainly because most of the public debate over this gets it backwards. PJM runs two separate markets. The capacity market pays a generator for being available — for existing, passing its performance tests, and standing ready to answer a dispatch call during the delivery year. That is what cleared at the $325-per-megawatt-day price cap in the most recent Base Residual Auction. The energy market is a different transaction entirely: it pays a generator only when it is actually dispatched, and dispatch order runs on marginal cost, cheapest megawatt-hour first. Most days of the year, a gas combined-cycle unit beats a coal steam unit on marginal cost, so gas gets called and coal sits. The coal plant still collects its capacity payment for being the insurance policy. It just does not get asked to actually do the job the insurance policy exists for, except on the handful of days a year — a January cold snap, a July heat dome — when gas prices spike or gas supply tightens and coal suddenly becomes the cheaper choice again.

That is a rational market design for an ordinary year. It stops being rational the moment the grid operator itself is telling FERC, in writing, that it is 6,831 megawatts short of its own reliability target for the second straight auction, while data centers alone are projected to help drive 32 gigawatts of new peak demand onto this same system by 2030. Ratepayers are now being asked to fund a $20 billion emergency backstop auction, at a $555-per-megawatt-day cap, to solve a shortfall that a fleet already sitting idle at 34 percent utilization could substantially close if the market simply called on it more often. That is not a hypothetical. It is 171 million megawatt-hours a year of existing, permitted, interconnected capacity that ratepayers are already paying to keep on standby and are barely using.

The honest counterargument deserves a straight answer, not a dodge. Coal's low capacity factor is not purely regulatory timidity; a real share of it reflects gas simply being cheaper on marginal cost most days, plus real environmental compliance costs, plus the wear-and-tear economics of cycling a baseload steam unit up and down instead of running it flat. Forcing dispatch preferences onto price formation has real costs too, including higher energy prices in the months coal beats gas on merit order and gets called anyway. That is a legitimate point, and any serious policy fix has to price it honestly rather than pretend the fleet can simply be switched to 85 percent utilization for free.

But "some of the gap is legitimate economics" is not the same claim as "all fifty-one points of the gap between 34 and 85 percent capacity factor is legitimate economics," and nobody arguing for premature retirement schedules has shown their work on that distinction. DOE's Section 202(c) emergency orders already established the legal and operational precedent that specific coal units can be directed to remain available and dispatch economically when reliability requires it. The next logical step — and the one this publication will keep pressing FERC and PJM to answer directly in the show-cause dockets now working through the system — is whether price formation and capacity accreditation rules should be doing more of that work automatically, rewarding the fleet for the reliability it already provides on paper by actually calling on the megawatt-hours it is capable of delivering. The data centers are coming whether the grid is ready or not. The fastest capacity PJM has access to is not the one still waiting on a transformer. It is the one already sitting on the interconnection, drawing a capacity payment, and waiting to be told to run.

— The Publisher

Methodology note: Capacity-factor headroom calculated as (installed nameplate MW × 8,760 hrs) at a realistic sustained 85% capacity factor, minus actual output at the most recently published EIA capacity factor (34%, 2023 generator data — EIA has not published a more recent PJM-specific coal capacity factor as of this issue). Installed coal capacity of 38,366.4 MW per Monitoring Analytics' 2025 State of the Market Report for PJM (as of Dec. 31, 2025). The 85% ceiling is an editorial assumption representing sustained baseload operation with normal planned maintenance and forced-outage allowance, not a PJM- or EIA-published target; actual achievable utilization would vary by unit condition, environmental permit limits, and fuel logistics, and should be verified against unit-specific EFORd and GADS data before use in any dispatch-planning or policy context.

DATES & DEADLINES

• Aug. 21, 2026 — FERC Docket AD26-7 post-technical-conference comments due (PJM governance)

• Sept. 1, 2026 — FERC Alternative Dispute Resolution forum opens (PJM governance track)

• September 2026 — PJM target window for FERC-approved special capacity backstop procurement (ER26-3380)

• Ongoing — Ohio Supreme Court docket activity on the AEP Ohio data-center tariff appeal (no argument date set)

• Ongoing — Texas statewide data-center grid-connection audit; no stated completion date

• 2027 legislative sessions — Oregon and New Mexico statewide data-center moratorium bills

• Quarterly — CBRE, JLL, and Avison Young data center market reports (next: Q3 2026, expected mid-October)

PRIMARY SOURCES

• Federal Energy Regulatory Commission, Docket Nos. RM26-4-000, EL26-67 through EL26-72, AD26-7-000, ER26-3380 — large-load interconnection show-cause orders, PJM governance docket, and capacity backstop filing.

• PJM Interconnection, "PJM Capacity Auction Procures 138,318 MW of Generation Resources," July 14, 2026 — 2028/2029 Base Residual Auction results.

• Day Pitney, McGuireWoods, Orrick — client alerts on FERC's June 18, 2026 show-cause orders and large-load interconnection reform categories.

• CBRE, "Global Data Center Trends 2026," Q1/Q2 2026 data. https://www.cbre.com/insights/reports/global-data-center-trends-2026

• Avison Young, "Q2 2026 US Data Center Market Overview." https://www.avisonyoung.us/us-data-center-market-overview

• JLL, "North America Data Center Report, Midyear 2026." https://www.jll.com/en-us/insights/market-dynamics/north-america-data-centers

• Associated General Contractors of America, "Data Centers - Economic Impact." https://www.agc.org/understanding-economic-impact-data-center-development

• Brookings Institution, "New Evidence on Data Center Employment Effects," Aug. 2026. https://www.brookings.edu/articles/new-evidence-on-data-center-employment-effects/

• Data Center Knowledge, "New Data Center Developments: August 2026." https://www.datacenterknowledge.com/data-center-construction/new-data-center-developments-august-2026

• Bloomberg News, "Data Center Gas Plants to Boost US Power Emissions by 20%," Aug. 18, 2026. https://www.bloomberg.com/news/articles/2026-08-18/data-center-gas-plants-to-boost-us-power-emissions-by-20

• MultiState Insider, "The Local Fight Over Data Centers: A Texas Case Study," Aug. 19, 2026. https://www.multistate.us/insider/2026/8/19/the-local-fight-over-data-centers-a-texas-case-study

• Data Center Moratoriums Weekly Tracker. https://www.datacenterbans.com/

• The Geopolitical Economist, "AI Infrastructure Bottleneck: The Transformer Shortage Slowing Data Centers," Jun. 2026. https://medium.com/the-geopolitical-economist/everyone-is-watching-the-ai-boom-nobody-is-watching-the-transformer-that-cant-be-built-in-time-341c523e8e81

• Inflect, "Data Center Power Shortage 2026: Why Grid Capacity Is Now the Bigger Constraint Than GPUs." https://inflect.com/blog/data-center-power-shortage-2026-why-grid-capacity-is-now-the-bigger-constraint-than-gpus

• American Action Forum, "AI Data Center Power Surge: Shifting Trends Toward Natural Gas." https://www.americanactionforum.org/insight/ai-data-center-power-surge-shifting-trends-toward-natural-gas/

• U.S. Department of Energy, "Fact Sheet: Ensuring Affordable Energy Access in Ohio While Powering the Future of AI," Mar. 2026. https://www.energy.gov/articles/fact-sheet-department-energy-ensuring-affordable-energy-access-ohio-while-powering-future

• Futurum Group, "AI Capex 2026: The $690B Infrastructure Sprint," Feb. 2026. https://futurumgroup.com/insights/ai-capex-2026-the-690b-infrastructure-sprint/

AI ASSISTANCE DISCLOSURE: AI assistance was used in research compilation, data formatting, and drafting of this publication. All data sourced from the primary and secondary sources cited throughout; editorial judgments and strategic assessments should be reviewed and confirmed by the author before publication. Readers should verify data directly with cited sources, particularly figures drawn from secondary trade-press summaries of primary filings rather than the underlying dockets themselves.