Background on power consumption and cost
Electricity consumption is one of the more poorly understood topics in everyday life, even though the underlying math is trivial. Three quantities are enough: power in watts, time in hours, price per kilowatt-hour. Multiply power and time, divide by 1000, and you get kilowatt-hours. Multiply kilowatt-hours by tariff and you get cost. This calculator does exactly that, with automatic extrapolation to typical comparison periods.
Why bother looking at single-device consumption? Because the cost drivers are often invisible. A server idling permanently at 50 watts consumes 438 kilowatt-hours per year, costing roughly 144 EUR at 0.33 EUR per kWh (as of 2026). A leaky freezer averaging 120 watts pushes past 1000 kilowatt-hours annually. An old 60-watt incandescent bulb, four hours a day, eats almost 88 kilowatt-hours, while a modern 7-watt LED delivers the same brightness for around 10 kilowatt-hours. Over ten years of operation the difference adds up to several hundred euros, for a single bulb.
For IT infrastructure the effect is even more pronounced. A home NAS running 24 hours a day pulls between 15 and 80 watts depending on the model. A gaming PC at idle draws 60 to 120 watts and four to six times that under load. If you self-host a server, you should honestly factor electricity into your total cost of ownership: the monthly bill often exceeds the depreciated hardware cost. KernelHost optimizes for low PUE in Frankfurt FRA01 with modern hardware and free cooling, which keeps that line item small.
A common misconception is the difference between watts and kilowatt-hours. Watts is power, the instantaneous rate of energy use. Kilowatt-hours is energy, power times time. A 60-watt bulb does not consume 60 kilowatt-hours, it consumes 60 watt-hours per hour, which equals 0.06 kilowatt-hours. That is also why electricity bills are charged in kWh, not in watts: the meter integrates power over time.
When entering values, be realistic. Devices are often labelled with peak power, not steady-state. A clothes dryer rated at 2500 watts pulls that only during the heat-up phase; the average over a full cycle is around 1500 watts. For exact numbers, buy a plug-in energy meter (15 to 30 EUR) and measure over 24 hours; the kWh reading is then the exact basis for your calculation. With that result you can also evaluate the value of energy-efficient replacement appliances or smart-home switching outlets in a serious way.
Electric mobility and household electricity
An electric vehicle changes a household's load profile dramatically. Where a typical four-person household without an EV consumes 3500 to 4500 kWh per year, a Tesla Model Y driven 15,000 km annually at 17 kWh per 100 km adds another 2550 kWh, closer to 2800 kWh once charging losses are included. A 1300 EUR yearly electricity bill quickly becomes a 2200 EUR bill (as of 2026), depending on your tariff.
Even so, electric driving wins against combustion in most setups. One kWh of home electricity costs around 0.33 EUR, one liter of premium gasoline 1.80 EUR. The Tesla Model Y consumes 17 kWh per 100 km, a comparable petrol car 7.5 liters. That is 5.60 EUR vs 13.50 EUR per 100 km, a 58 percent saving on energy alone. Charging at HPC public stations (0.55 to 0.79 EUR/kWh) drops the saving to 30 to 40 percent. With a private PV system charging the car midday, the cost falls to 1.50 to 2.50 EUR per 100 km, almost free of marginal cost.
In practice you charge at home with 11 kW (3-phase, 16 A, CEE plug). That delivers a full charge in 7 to 8 hours, a perfect overnight fit. A 22 kW wallbox only pays off if the car can actually accept it (many cannot) or if multiple EVs charge in parallel. Important: 22 kW requires utility approval, 11 kW is only a notification.
Photovoltaics and self-generated electricity
A 10 kWp PV system in Central Europe (as of 2026) generates between 8500 and 10,500 kWh per year, depending on location, orientation and shading. South-facing roofs in southern Germany outperform north-west roofs in northern Germany by up to 25 percent. Investment cost is 12,000 to 16,000 EUR without battery, 18,000 to 25,000 EUR with a 5 to 10 kWh battery.
Economically the self-consumption portion is what matters. Self-used solar electricity replaces grid imports at 0.33 EUR/kWh, while exported electricity earns just 7 to 8 cents in feed-in tariff. A 10 kWp system at 30 percent self-consumption saves 990 EUR per year and earns another 470 EUR in feed-in payments, so 1460 EUR yearly return. With a battery, self-consumption rises to 70 percent and the combined number lands around 2400 EUR.
The optimal setup combines PV plus EV plus heat pump plus dynamic tariff. By day the car charges directly from PV (lifetime levelized cost of solar 8 to 12 cents per kWh), at night the dynamic tariff secures cheap grid electricity during off-peak hours. With an energy management system (Solarwatt, Sonnen, OpenEMS) self-consumption can be pushed to 80 percent and overall household self-sufficiency to 60 to 70 percent.
Dynamic electricity tariffs and smart meters
Classic electricity tariffs charge a flat price per kWh regardless of when you consume. Dynamic tariffs (Tibber, aWATTar, Octopus, Rabot Charge) pass the hourly EPEX SPOT exchange price directly to the customer, plus a small service fee. On a typical 2026 day the price oscillates between 5 and 25 cents, and on windy and sunny days some hours hit 2 cents or even go negative.
Shifting consumption into cheap hours saves 15 to 30 percent. Wallbox example: without control, the EV charges when the driver gets home (6 to 10 PM, often a peak phase). With a dynamic tariff and Tibber Pulse integration the wallbox starts automatically between midnight and 5 AM (cheap hours). A 75 kWh full charge costs 7 to 12 EUR instead of 20 to 25 EUR. Across a year the savings reach 400 to 600 EUR per EV.
Heat pumps benefit just as much: thermal storage volume allows pre-heating during low-price phases. With an SG-Ready (Smart Grid Ready) device and a controller like Home Assistant or evcc, 20 to 30 percent of yearly heating costs disappear without any comfort loss.
Fuses, wallbox connections and the 80 percent rule
A European Schuko outlet sits on a 16 A breaker and delivers a theoretical 3680 W at 230 V. For sustained loads the 80 percent rule applies: maximum 2900 W over multiple hours, otherwise the cabling overheats. That is exactly why Schuko emergency EV charging cables are limited to 2.3 kW (10 A) or at most 3.7 kW (16 A); an 11 kW charge over Schuko is physically impossible.
Wallboxes mandate 3-phase power. An 11 kW wallbox runs on a CEE 32 A 3-phase circuit, with each phase at 16 A: 3 times 230 V times 16 A equals 11,040 W. 22 kW wallboxes double the per-phase current to 32 A, requiring at least 6 mm² copper cabling. Both variants need a Type B residual current device (RCD) or built-in DC fault current detection in the wallbox.
In the household, high-load appliances belong on dedicated circuits. Kettle (2000 W) and vacuum (1200 W) on the same circuit add up to 3200 W and trip the breaker as soon as a motor stalls. Ovens and induction cooktops therefore have their own circuits, and a wallbox always gets a separate 3-phase circuit straight from the meter cabinet.
Typical household appliances at a glance
energy-calculator.about.appliances_introValues are guidelines; for exact measurement use a plug-in energy meter. As of 2026 at 0.33 EUR/kWh.
Datacenters, AI workloads and KernelHost FRA01
A classic hosting datacenter like KernelHost in Frankfurt am Main (Maincubes FRA01, Tier III) typically runs at 16 to 25 MW IT load per hall. With a PUE (Power Usage Effectiveness) of 1.2 to 1.3, total intake including cooling, UPS losses and lighting reaches 20 to 32 MW. Over a year that is around 175 GWh per hall. We use 100 percent renewable electricity, drastically lowering the carbon footprint per server hour.
AI datacenters dwarf this scale. Where classic DCs land between 5 and 30 MW, hyperscale AI datacenters (as of 2026) are far larger: xAI Colossus in Memphis hits roughly 200 MW, Microsoft Stargate is planned for several gigawatts. The reason: GPU clusters with tens of thousands of NVIDIA H100, H200 or Blackwell accelerators reach 50 to 130 kW per rack, versus 5 to 15 kW for classic CPU hardware.
Globally, datacenter electricity consumption was around 460 TWh in 2024, about 1.5 percent of world electricity use. IEA forecasts for 2030 land between 800 and 1000 TWh, primarily driven by AI training and inference. For comparison: Germany's total annual electricity demand sits around 500 TWh.
In practical terms: every ChatGPT query costs roughly 2 to 5 Wh of inference electricity, about ten times a Google search. A training run for a GPT-4 class model consumes 50 GWh or more, equivalent to the yearly electricity demand of a town of 15,000 people. KernelHost itself runs classic web hosting and VPS infrastructure, no AI training. Our hardware physically lives in Frankfurt FRA01, Tier III, with redundant network uplinks and diesel backup.