[@CityPrepping] America Is Not Ready For What’s Coming - pt 1

Link: https://youtu.be/Ru5zJ_eRtPg

Duration: 24 min

Transcript: Download plain text

Short Summary

The episode explores how surging electricity demand from AI data centers is straining the US power grid, with Lawrence Berkeley National Lab finding data centers consumed 4.4% of all US electricity in 2023 and projections showing 6.7% to 12% by 2028. Experts discuss how hyperscale AI facilities require power plant-scale demand (300-1,000 MW), creating concentrated round-the-clock demand that arrives faster than infrastructure can be expanded. The conversation examines how communities are pushing back against transmission lines, why utilities face challenging long-term planning, and what households can do to reduce their exposure to grid strain.

Key Quotes

1. "The cloud, it's not a cloud. It is buildings, land, water, transformers, substations, backup generators, power purchase agreements, transmission lines, workers, steel, concrete, and electricity." (00:01:56)
2. "Lawrence Berkeley National Laboratory found that data centers use about 4.4% of all electricity in the United States in 2023, and by 2028, that number could rise to somewhere between 6.7 and 12%." (00:01:06)
3. "hyperscale AI facilities requiring 300 to 1,000 MW or more, which means these are not small office buildings with servers in the back. They are power plant scale demand. One large AI campus can require as much electricity as a midsize city." (00:01:16)
4. "We're talking about whole new cities of demand popping up within cities that already have demand issues. Cities of demand on top of a city's demand." (00:01:22)
5. "The grid was built for one kind of America, and now it is being asked to power another." (00:01:11)

Detailed Summary

Overview

This episode examines the escalating tension between AI-driven electricity demand and the US power grid's capacity to meet it. The discussion draws on data from Lawrence Berkeley National Laboratory, the International Energy Agency, and Department of Energy advisors to paint a picture of infrastructure strain at a national scale.

Electricity Demand Growth

• Lawrence Berkeley National Laboratory found data centers used 4.4% of all US electricity in 2023, with projections reaching 6.7% to 12% by 2028
• The International Energy Agency projects global data center electricity demand could more than double by 2030
• For nearly two decades, electricity demand was relatively flat, allowing utilities to plan around slow growth and gradual upgrades
• US electricity use has been pushing into record territory with EIA projections showing total consumption continuing to rise

AI Infrastructure Scale

• Department of Energy advisors have discussed hyperscale AI facilities requiring 300 to 1,000 MW—described as power plant scale demand rather than small office buildings
• One large AI campus can require as much electricity as a midsize city, representing concentrated round-the-clock demand that can arrive faster than surrounding power infrastructure can be expanded
• AI is more deeply embedded into cloud computing, search, logistics, finance, military planning, software development, medicine, and manufacturing, making demand harder to unwind compared to crypto mining which can move quickly and chase cheap power

Competing Demand Drivers

• Manufacturing reshoring is adding demand through semiconductors, batteries, defense production, and advanced manufacturing—all power-intensive industries requiring reliable electricity on a large scale
• Recent heat events create long stretches of high temperatures that settle over regions and keep air conditioners running for days, turning cooling from a comfort issue into a life safety issue

Grid Infrastructure Challenges

• Power plants take years to permit and build; transmission lines take even longer due to crossing counties, states, private and public land, requiring environmental reviews
• The grid must keep supply and demand balanced every second, and one failure can force power onto other lines, spreading problems if already stressed
• Large transformers are custom built, expensive, heavy, and slow to replace—not available off the shelf when failures occur
• SDG&E is proposing 140 miles of 500 kV high voltage transmission towers running directly through Temecula, California, a city that does not receive power from the line and has officially opposed the project
• Northern Virginia became the center of the data center world due to fiber connections, land availability, tax policy, proximity to major customers, and an existing technology corridor
• Virginia state-level analysis has warned that ordinary residential customers could eventually see higher monthly bills if the cost of new grid infrastructure is spread across everyone instead of being isolated to the largest new users

Regional Grid Strain

• In the PJM grid region, which serves tens of millions of people across the Mid-Atlantic and Midwest, the market monitor has warned that data center demand is already contributing to sharp increases in capacity prices
• The core challenge is not whether America can produce more electricity in a general sense, but whether the right electricity can reach the right place at the exact moment most people need it
• Grid strain shows up when heat settles over a region for several days with millions of air conditioners running almost non-stop, or when a sudden cold snap forces homes, businesses, and utilities to pull hard at the same time
• A region can have enough annual energy and still lack enough deliverable power at a specific hour in a specific place if transmission is constrained

Geopolitical Context

• The United States and China both view AI as an economic, military, and technological priority, comparable to a new industrial arms race rather than a passing tech trend

Utilities and Planning

• Some utilities are buying themselves more room by managing demand, building capacity, and upgrading infrastructure early, while others will struggle because demand arrives faster than planning, equipment, or local politics can absorb
• Data center states may attract investment and jobs, then spend years arguing over who has to pay for power lines, substations, backup capacity, and water demand needed to support them
• The country can be building more power and still feel strain while trying to catch up

Impact on Households

• For households, the outcome may look less like one national crisis and more like a steady loss of margin, showing up through higher utility bills, peak demand pressure, local infrastructure fights, and outages from storms, wildfires, heat waves, equipment failures, or human error
• The common thread is margin—if demand keeps rising while infrastructure lags, that margin gets thinner and disruptions matter more because the system has less room to absorb them
• A 6-hour outage, an overnight outage, a 2-day outage, and a recurring outage are all different problems requiring different levels of preparation

Recommendations for Resilience

• A practical power plan does not require disconnecting from the grid or jumping straight into full energy independence; most households can reduce exposure one layer at a time
• Households can reduce exposure by identifying critical loads, protecting refrigeration, keeping communications running, planning for medical needs, and creating at least one way to function when the grid is under strain
• AI infrastructure may become more efficient, but building enough generation, transmission, storage, and backup capacity will take time, money, and political will