Megawatt to Terawatt

In the US, roughly 750–1,000 homes (average consumption ~1.2 kW per home). In Europe, where usage is lower, 1 MW can serve 1,500–2,000 homes. But this is average — on a hot summer afternoon when everyone cranks AC, that number can drop to 300–400 homes. Grid planners must size for peak demand, not averages, which is why installed capacity far exceeds average load.

A small data center uses 1–5 MW; a large hyperscale facility (Google, AWS, Microsoft) draws 50–200 MW — some exceeding 300 MW. The entire US data center industry consumed about 17 GW in 2023, roughly 4% of national electricity. AI training clusters are pushing demand higher: a single large GPU cluster can draw 50–100 MW, and planned AI-focused campuses target 1 GW or more.

Onshore turbines typically rate 2–6 MW; the latest offshore monsters reach 14–16 MW per turbine. Vestas' V236-15.0 MW turbine has a rotor diameter of 236 meters — wider than two football fields. A single sweep of its blades can generate enough electricity for a UK household for two days. Capacity factors run 25–45% onshore and 40–55% offshore, so actual average output is roughly half the nameplate rating.

Most operating reactors produce 500–1,400 MW of electrical power. The world's largest, at France's Gravelines plant, has six reactors totalling 5,460 MW. Small Modular Reactors (SMRs) being developed target 50–300 MW each. Nuclear plants run at 85–95% capacity factor — far higher than wind (~35%) or solar (~25%) — meaning a 1,000 MW reactor actually delivers about 900 MW on average.

MW tells you the maximum instantaneous power the battery can deliver (how fast it can discharge), while MWh tells you total stored energy (how long it can sustain that output). A 100 MW / 400 MWh battery can deliver 100 MW for 4 hours, or 50 MW for 8 hours. Grid operators care about both: MW for handling sudden demand spikes, MWh for sustained backup during extended outages or evening solar fade.

The Sun delivers about 173,000 TW to Earth's surface. Human civilisation uses roughly 18 TW total. So we'd only need to capture 0.01% of incoming solar energy to power everything — an area of solar panels roughly 400 km × 400 km, about the size of Montana. The challenge isn't total energy availability; it's cost, storage, transmission, and the fact that sunlight is spread thin and intermittent.

Imagine 18 trillion light bulbs burning continuously, or 9 billion people each running a 2 kW heater non-stop. That 18 TW figure includes everything — electricity, transport fuel, industrial heat, cooking, heating. About 40% comes from oil, 27% from coal, 24% from gas, and the rest from nuclear and renewables. The US alone accounts for about 3 TW despite having only 4% of world population.

Replacing all 18 TW of human energy with clean sources would require roughly 60–75 TW of installed solar capacity (accounting for ~25% average capacity factor). That's about 40 times current installed solar. At 2023 installation rates of ~0.4 TW/year, it would take 150 years — but installation rates are doubling every 2–3 years. If that exponential trend holds, we could theoretically reach 60 TW of solar within 15–20 years.

Earth radiates about 47 TW of geothermal heat from its interior, driven by radioactive decay and residual primordial heat. That's 2.5× human energy consumption, but it's spread across the entire surface at extremely low density (~0.09 W/m²). Iceland, sitting atop a mantle plume, exploits geothermal for 90% of its heating. Globally, geothermal electricity capacity is only about 16 GW — a tiny fraction of what's theoretically available.

No — the terawatt scale is a very recent phenomenon. In 1800, global human power consumption was about 0.5 TW (mostly biomass burning). By 1900 it reached 1 TW with coal industrialisation. We crossed 10 TW around 1985. The jump from 1 to 18 TW in just 120 years tracks almost perfectly with global population growth times rising per-capita energy use. Pre-industrial humans used about 0.1 kW each; Americans now average 10 kW per person.