LAT 00.0000LON 000.0000NODE PLANETARY-001● LIVE / ILLUSTRATIVERUN SCENARIO-A · 2025
Electricity
Is Eating
Civilization.
From steel mills and transportation to AI superclusters and planetary-scale computation — humanity is rewriting civilization into electrical flow. The combustion era is being replaced by an electronic one, one substation at a time.
Global electricity generation
0
TWh/yr
2023, Ember/IEA — ~5× what it was in 1980
Global data-center demand
0
TWh/yr
2022 IEA; projected ~1,000 TWh by 2026
AI-attributable demand (US)
0
TWh/yr
2024 indicative; doubling every ~18 mo
EVs sold worldwide
0.0
M / yr
2024 BNEF — 22% of new car sales
Electricity share of final energy
0
%
Up from ~10% in 1973
Annual grid capex needed
0
B USD/yr
IEA — to keep pace with electrification
SCROLL TO DESCEND ALONG THE LOAD CURVE
●GLOBAL GEN 30.5K TWh ▲/●DATA CENTER 460 TWh ▲ +18%/●EV STOCK 44 M ▲/●GRID CAPEX 600 BN USD/●CU PRICE 9,420 USD/t ▲/●LI CARBONATE 13.5 K USD/t ▼/●URANIUM U3O8 78 USD/lb ▲/●NUCLEAR RESTARTS 18 announced/●AI CLUSTER WAITLIST 72 GW/●ARC FURNACE SHIFT +9%/●HVDC LINES IN BUILD 31/●HEAT WAVE PEAKS 12 GRIDS RED/●GLOBAL GEN 30.5K TWh ▲/●DATA CENTER 460 TWh ▲ +18%/●EV STOCK 44 M ▲/●GRID CAPEX 600 BN USD/●CU PRICE 9,420 USD/t ▲/●LI CARBONATE 13.5 K USD/t ▼/●URANIUM U3O8 78 USD/lb ▲/●NUCLEAR RESTARTS 18 announced/●AI CLUSTER WAITLIST 72 GW/●ARC FURNACE SHIFT +9%/●HVDC LINES IN BUILD 31/●HEAT WAVE PEAKS 12 GRIDS RED/
01 /DEFINITION
What electrification actually means.
Not 'using electric devices.' Electrification is the conversion of civilization from combustion-driven systems into electronically controlled ones — every step in production, transport, and computation routed through wires, switches, and code.
FROM ↘
○Steam
○Coal
○Mechanical linkages
○Combustion engines
○Hydraulic systems
○Human labor
TO ↗
●Electricity
●Semiconductors
●Power electronics
●Automation
●AI-controlled systems
●Digital infrastructure
CIVILIZATION-SCALE ENERGY ERAS · CONTROL REVOLUTIONS
EACH ERA EATS THE LAST
—~1700
Wood
Heat. Forest as battery.
1700–1900
Coal
Steam. Mechanical force.
1900–2000
Oil
Mobility. Liquid energy.
1880–now
Electrical
Light. Motor. Wire.
1947–now
Semiconductor
Logic. Switching at light-speed.
2017–
AI Compute
Intelligence as a load profile.
control · Manual
control · Mechanical
control · Hydraulic + thermal
control · Electromechanical
control · Programmable
control · Self-optimizing
Every leap in civilization has been an energy-control revolution — not the discovery of more energy, but the discovery of finer ways to command it. Wood, coal, oil, electrons, transistors, models: each new layer is a control surface stacked on the layer beneath. AI is the latest. The substrate is still the wire.
WHY ELECTRICITY BECAME THE UNIVERSAL FORM
PROP_01Controllable
Switched in nanoseconds; modulated in MHz.
PROP_02Scalable
Same physics from a watch battery to a 22 GW dam.
PROP_03Efficient
Motors ~95% vs ICE ~25–35% well-to-wheel.
PROP_04Transmittable
HVDC moves GW across 3,000+ km at ~3% loss per 1,000 km.
PROP_05Programmable
Power electronics shape current at will.
PROP_06Universal
Any energy source can become electricity; electricity drives any load.
02 /LOAD BANK · HEAVY INDUSTRY
The industries that eat the grid.
Eleven processes account for the majority of all industrial energy use on Earth. Some already run on electrons. Others run on burning carbon for chemistry that electrons cannot replace — yet.
INDUSTRYEJ/yrELEC %HARDNESS
DETAIL · STEEL
Steel
SCALE
1.9 Gt/yr
PROCESS FLOW
Iron ore▶Coke▶BF▶BOF▶Mill
ENERGY USE
24
EJ/yr
CO₂ SHARE
7% global CO₂
ALREADY ELECTRIC
16
%
ELEC HARDNESS
88
/ 100
Blast furnace burns coke as both heat & reductant — electrons can replace heat but not chemistry.
DOCTRINE
Some processes don't resist electrification — they resist decarbonization. Coke in a blast furnace isn't fuel, it's the chemistry. Calcining limestone splits CO₂ out of the rock no matter what fires the kiln. For these, electrons alone are not enough; we need new molecules and new processes — green hydrogen, electrolytic iron, calcined-clay binders, e-crackers.
03 /LOAD CATALOG · MACHINES THAT EAT POWER
Electricity black holes.
Where do the electrons actually go? From a single LED bulb to a 1-GW compute campus, the spread is nine orders of magnitude — and the upper end is being built right now.
HOME · APPLIANCES
LED bulb
DRAW · 8 W
ANNUAL · 12 kWh/yr
Per fixture
Refrigerator
DRAW · 150 W avg
ANNUAL · 400 kWh/yr
Always on
Air conditioner
DRAW · 1.5 kW
ANNUAL · 1,500 kWh/yr
Per unit, summer
Electric heater
DRAW · 2 kW
ANNUAL · 4,000 kWh/yr
Cold-climate home
EV charger (L2)
DRAW · 7–11 kW
ANNUAL · 4,000 kWh/yr
12,000 mi/yr EV
Induction stove
DRAW · 3.5 kW
ANNUAL · 500 kWh/yr
Per home
Heat pump (whole house)
DRAW · 3 kW
ANNUAL · 6,000 kWh/yr
Northern climate
COMMERCIAL · BUILDINGS
Office HVAC
DRAW · 0.1–0.3 kW/m²
ANNUAL · 100–300 kWh/m²·yr
Largest line in most office bills
Supermarket cold-chain
DRAW · 300 kW typ.
ANNUAL · 2.5 GWh/yr
Per store, refrigeration alone
Elevator (high-rise)
DRAW · 20 kW peak
ANNUAL · 60 MWh/yr
Per shaft
Subway line
DRAW · 10–40 MW
ANNUAL · 350 GWh/yr
Single metro line, mid-density city
Urban street lighting
DRAW · —
ANNUAL · —
~1% of urban demand; LED retrofit cuts 60%
INDUSTRIAL · MACHINES
Electric arc furnace
DRAW · 80 MW
ANNUAL · 450 GWh/yr
Recycles scrap steel
Hall–Héroult cell line
DRAW · 350 kA
ANNUAL · 14 MWh/tonne Al
Continuous DC; 24/7
Hydrogen electrolyzer
DRAW · —
ANNUAL · 50–55 kWh/kg H₂
Industrial scale 100 MW+
Industrial boiler
DRAW · varies
ANNUAL · —
Replacement target #1
Compressor station (gas pipeline)
DRAW · 30 MW
ANNUAL · 260 GWh/yr
Per station
Pumping station (city water)
DRAW · 5–20 MW
ANNUAL · —
Always on, big chunk of municipal load
EUV lithography (ASML)
DRAW · 1.4 MW
ANNUAL · 12 GWh/yr
Per scanner; 1+ per advanced fab
AI · COMPUTE INFRA
NVIDIA H100
DRAW · 700 W TDP
ANNUAL · —
Per GPU; B200 ~1,000 W
DGX SuperPOD (256-GPU)
DRAW · 200 kW
ANNUAL · 1.7 GWh/yr
Reference design
100k-GPU cluster
DRAW · 150 MW
ANNUAL · 1.3 TWh/yr
xAI Colossus class
Hyperscale DC campus
DRAW · 500 MW – 1 GW
ANNUAL · 4.4–8.8 TWh/yr
New Texas/Virginia builds
Liquid cooling loop
DRAW · —
ANNUAL · —
Cuts PUE from 1.5 → 1.1; required >50 kW/rack
Frontier model train
DRAW · —
ANNUAL · ~50 GWh/run
GPT-4 class, single training run
SCALE LADDER · POWER DRAW (LOG)
8W
LED
150W
Fridge
1.5kW
AC
11.0kW
EV charge
700W
H100 GPU
80MW
EAF
40MW
Subway
150MW
Cluster
1.0GW
DC campus
22.5GW
Three Gorges
04 /THE SECOND GREAT ELECTRIFICATION
AI converts electricity into intelligence.
A frontier model training run consumes more electricity than a small city does in a year. Inference scales the bill into the trillions of tokens per day. For the first time in industrial history, the limiting factor on a software company is the grid.
PIPELINE · ELECTRON → TOKEN → INTELLIGENCE
[00]
Electron
Coal, gas, sun, wind, fission
1 MWh
▶
[01]
Grid
Transmission, transform, switch
−6% loss
▶
[02]
Datacenter
Substation, UPS, cooling
PUE ≈ 1.2
▶
[03]
GPU
Tensor cores, HBM, NVLink
~700 W/die
▶
[04]
Transformer
Attention × layers × tokens
~1 J / token
▶
[05]
Tokens
Inference output
—
▶
[06]
Intelligence
Cognition on demand
USD ?
CORE STATEMENT
The next global competition is Power Generation × Chips × Grid Capacity.
Three factors, all hard, all multiplied. Lose any one and the other two are stranded.
ANNOUNCED POWER DEALS · HYPERSCALERS & FRONTIER LABS
BUYER
DEAL
CAPACITY
ONLINE
Microsoft
20-yr PPA, Three Mile Island restart
835 MW
2028
Amazon
Talen Energy nuclear-adjacent DC
960 MW
2024
Google
Kairos Power SMR offtake (multi-site)
500 MW
2030
Meta
Entergy gas + nuclear PPAs, Louisiana
2,200 MW
2027
xAI
Memphis 'Colossus' on-site gas turbines (interim)
150 MW
2024
OpenAI
Stargate JV — multi-site, multi-GW
10,000+ MW
2025–29
Sources: public filings, IEA & press, 2024–25. Capacities indicative, often shared across multiple sites.
Nuclear restart
TMI-1, Palisades, Duane Arnold under discussion — first US restarts in a generation, all on AI demand.
Behind-the-meter gas
xAI's Memphis 'Colossus' fires interim natural-gas turbines on-site because grid interconnect queues are 4–7 years.
Power → location
New compute siting follows electricity, not talent — north Virginia, central Texas, Quebec, Riyadh, the Gulf.
05 /LIMITS · WHY THE ENERGY TRANSITION IS HARD
Why humanity cannot easily achieve 100% electrification.
The end-state — fully electric, fully decarbonized civilization — is physically possible. Getting there is constrained by five hard layers: grid, storage, transmission, minerals, and the industries chemistry refuses to release.
BOTTLENECK 01
Grid capacity
The wires were sized for the 1990s. AI compute + EV charging + heat-pump retrofits + reshored manufacturing are arriving simultaneously, and a substation takes 4–7 years to permit and build. Interconnect queues now hold 2,600+ GW of projects in the US alone.
US interconnect queue2,600 GW
Wait time for new connection5 yr
Transformer lead time120 wk
Annual US grid capex130 B$
BOTTLENECK 02
Energy storage
Solar and wind produce when the weather decides; civilization runs on schedules. We can buffer hours with lithium. We cannot buffer winters. Seasonal storage is the unsolved engineering problem of the century.
CHEMISTRY
Wh/kg
DURATION
$/kWh
Lithium-ion (LFP/NMC)
220
1–4 h
$140/kWh
Sodium-ion
140
1–6 h
$80/kWh
Pumped hydro
0.3
8–24 h
$5/kWh
Green hydrogen
33,000
weeks+
—
Thermal (molten salt)
250
6–15 h
$30/kWh
CAES (compressed air)
5
4–24 h
$50/kWh
Hydrogen looks great on energy-density. Round-trip efficiency is ~35%, against ~90% for batteries.
BOTTLENECK 03
Transmission
The Sahara has the sun. The Gobi has the wind. The cities are in Tokyo, Mumbai, and Frankfurt. Moving electrons across continents requires HVDC corridors that cost tens of billions and cross sovereign borders.
FLAGSHIP HVDC PROJECTS
Changji – Guquan, CN
±1100 kV
3,293 km
12 GW
Belo Monte, BR
±800 kV
2,539 km
4 GW
Rio Madeira, BR
±600 kV
2,375 km
6.3 GW
Xinjiang–Anhui, CN
±800 kV
3,324 km
8 GW
NordLink, NO↔DE
±525 kV
623 km
1.4 GW
SunCable AAPL, AU↔SG
±525 kV (planned)
4,300 km
4 GW
Loss budget for HVDC: ~3% per 1,000 km of line, ~1% per converter station.
BOTTLENECK 04
Resource bottlenecks
Full electrification means a generational expansion of mining. Every motor needs copper, every battery needs lithium, every magnet needs rare earths, every reactor needs uranium. The mines do not exist yet.
CuCopper
USE
Wire, motor, transformer
SUPPLY → DEMAND
26 Mt/yr → Doubles by 2035
RISK
Pit-grade decline
LiLithium
USE
Cathode, electrolyte
SUPPLY → DEMAND
180 kt LCE/yr → 3× by 2030
RISK
Chile/Australia/China concentrated
NiNickel
USE
NMC cathode, stainless
SUPPLY → DEMAND
3.6 Mt/yr → +60% by 2030
RISK
Indonesia processing
CoCobalt
USE
NMC cathode, superalloys
SUPPLY → DEMAND
230 kt/yr → +40%
RISK
DRC ~70% of supply
REERare earths
USE
Magnets (Nd, Dy, Pr)
SUPPLY → DEMAND
350 kt REO/yr → +150%
RISK
China ~85% refining
UUranium
USE
Reactor fuel
SUPPLY → DEMAND
60 kt U/yr → +50% by 2040
RISK
Kazakhstan 40%, conversion bottleneck
BOTTLENECK 05
Industries chemistry won't release
Some processes are not just hard to electrify — they are non-electric by their nature. Either the molecule is the fuel, or the temperature is too extreme, or the energy density of a battery cannot compete with a hydrocarbon.
Aviation
WHY
Jet-A ~12,000 Wh/kg vs Li-ion ~250 Wh/kg — energy density gap 50×.
PATH
SAF, hydrogen for short-haul, no clear long-haul electric path.
Cargo shipping
WHY
Voyages 10,000 km, hold occupied by cargo not batteries.
PATH
Ammonia, methanol, on-board nuclear (concept).
Steelmaking
WHY
Carbon is both fuel and reductant of iron ore.
PATH
H₂-DRI + electric arc; SSAB/HYBRIT, Boston Metal MOE.
Cement
WHY
Calcination of limestone emits CO₂ regardless of fuel.
PATH
Carbon capture; novel binders (calcined clay, geopolymer).
Heavy chemicals
WHY
Crackers + reformers operate at 800–1100°C, molecule-shaping reactions.
PATH
Electric crackers, green H₂ feedstock.
Long-haul trucks
WHY
Payload + range = battery dominates GVW.
PATH
Megawatt charging, fuel cells; partly possible.
06 /GEOPOLITICS · CIVILIZATION-SCALE RESOURCE WARFARE
The new map is drawn in electrons, atoms, and amperes.
Every node on this map is a chokepoint. Some control the inputs (cobalt, uranium, lithium), some the conversion (EUV, refineries, smelters), some the consumption (AI campuses, megacities). Whoever wires them together first defines the next decade.
GLOBAL NODE GRAPH · ELECTRIC POWER STACKLEGEND · ● grid/compute · ● power/fuel · ● mineral
China UHV
US AI clusters
Gulf states
DRC cobalt belt
Chile lithium
Pilbara iron
Canadian hydro
Kazakh uranium
TSMC fabs
Samsung/SK
ASML EUV
Russian gas
GRID
China UHV · ±1100 kV
World's only operational ultra-high-voltage backbone — moves Xinjiang renewables 3,300 km east to coastal demand.
COMPUTE
US AI campuses
Northern Virginia, central Texas, central Iowa, central Ohio — every new mega-DC follows the substation, not the talent.
MINE
DRC · Chile · Australia
Cobalt belt, Atacama lithium triangle, Pilbara iron — primary upstream of the entire transition.
CHIP
ASML · TSMC · Samsung
Single supplier (EUV) feeding two foundry hubs feeding the entire AI infrastructure stack.
07 / SYSTEM / CIVILIZATION-SCALE
Modern civilization is a giant electrical system.
Who controls energy → controls industry.
Who controls the grid → controls AI.
Who sustains electricity → controls the future.
For 300,000 years, every human innovation was an energy-control revolution — fire, the plow, steam, the internal combustion engine. The current one wraps the planet in copper and silicon, and asks intelligence itself to ride on the wire.
A control panel for planetary civilization