Pipeline + Energy + Cybersecurity Layers
Routes — color by demand (default)
Data Center / AI
LNG export
Power generation
LDC
Industrial
Storage
Other
Unspecified

Routes — color by basis (toggle on): destination citygate − Henry Hub
discount (≤ $0)
+$0 to +$1
+$1 to +$3
+$3 to +$6
+$6 and above

Data layers
Planned NG plant
Operating NG plant ≥200 MW
PeeringDB data center
Meta hyperscaler campus

Toggle "State NG-generation share" layer to see what % of each state's power comes from gas. Toggle "Routes by BASIS" to switch coloring from demand-served to live basis spread.
VoidLink → Countdown How Bad Is It? Shodan Exposures Public Comments Eminent Domain Power Map Critical Minerals About
Color routes by:
Built with extensive assistance from Claude Code—via both the command-line interface and the Visual Studio Code extension—The Big Drop is an experimental project aimed at identifying and leveraging data to map a macro-level view of U.S. pipeline infrastructure and its rapid expansion to power hyperscale data centers spreading across the country. That infrastructure runs on industrial control systems that open and close valves and hold pressure and temperature within safe limits. On a public network, reaching one of these controls means commanding it, and many run software with known flaws already under attack. The daily newsletter and ongoing map updates are produced by publisher and editor-in-chief Jayne Lytel. If you spot any inaccuracies or have ideas to improve the map, please send them her way at jayne [at] runfromsunday [dot] com. The Big Drop is AI-generated and not fact checked—only some items are spot-checked—which means you need to verify any information if you plan to act upon it.
The Big Drop — Pipeline Infrastructure Daily
The Worst News of the Day (Or How Bad Is It?)
The Big Boom Not Yours Anymore Under Development
BLUF

Bottom Line Up Front

2026-07-17 06:19 PDT

WHOUS gas-pipeline operators (Williams, KMI, Enbridge, TC Energy), hyperscalers (Meta, Google, Microsoft, AWS, xAI, OpenAI/Stargate), and the AI labs leasing their compute (Anthropic on AWS+GCP).

WHATThe largest US natural-gas pipeline buildout since 2008—~18 Bcf/d new capacity in 2026—feeding gas-fired power plants that feed AI data centers and LNG exports.

WHENMost capacity online 2026–2028. Gulf Coast LNG and Northern Virginia are the biggest demand centers; PJM peak demand is up +4.4% YoY—the data-center signal.

SO WHATOperators segment their own networks well (TC Energy, Enbridge: 0 exposed devices). The risk lives where gas plants and data centers share neighborhoods—40,897 ICS endpoints within 25 km of Ashburn, 1,934 near Meta's New Albany (Project Socrates). A parallel “shadow grid” is being built for hyperscalers, often via eminent-domain seizure (Georgia Power Wansley, NextEra MidAtlantic Resiliency Link, Williams NESE). State legislatures are pushing back.

Cyber watch29 pipeline-relevant KEVs tracked (13 high-priority on SCADA platforms with US exposure; 1 ransomware-campaign flagged = 3.4%). 0 added in the last 7 days; 0 in the last 30. VulnCheck flags 26 additional CVEs as actively exploited that CISA has not yet added.
Layers

Pipeline · Energy · Cybersecurity Layers

Map legend
Routes — color by demand
Data Center / AI
LNG export
Power generation
LDC
Industrial
Storage
Routes — color by basis
Discount (≤ $0)
+$0 to +$1
+$1 to +$3
+$3 to +$6
+$6 and above
Data layers
Planned NG plant
Operating NG plant ≥ 200 MW
PeeringDB data center
Meta hyperscaler campus
Toggle “Routes by basis” to switch coloring from demand-served to live basis spread.
Hyperscalers

AI / Hyperscaler Layers

Compute footprint
Operators
Meta — 28 US campuses
Google — 33 US sites
xAI — Memphis Colossus
OpenAI / Stargate — 5 sites
Cloud regions
AWS region
Azure region
GCP region
Use the layer control (top-right) to toggle each region layer on the map.
Anthropic

About Anthropic

Where Claude runs

Anthropic doesn't operate its own data centers.

Claude runs on AWS (primary partner; AWS Trainium / Bedrock) and Google Cloud (per the Oct 2025 expanded agreement).

To see where Anthropic's compute physically lives, enable the AWS / Azure / GCP regions layers and look at AWS us-east-1 (Virginia), us-west-2 (Oregon), and GCP us-east4 / us-central1.