Power Consumption in Satisfactory: Budgeting a Grid

Grids do not trip because your average demand exceeded average supply. They trip because demand exceeded supply at one instant. Budgeting for the spikes rather than the average is the whole discipline.

What machines actually draw

At 100% clock speed:

Building	Draw
Constructor	4 MW
Smelter	4 MW
Miner Mk.1	5 MW
Packager	10 MW
Assembler	15 MW
Miner Mk.2	15 MW
Foundry	16 MW
Water Extractor	20 MW
Refinery	30 MW
Oil Extractor	40 MW
Miner Mk.3	45 MW
Manufacturer	55 MW
Blender	75 MW
Drone Port	100 MW
Resource Well Pressurizer	150 MW

The spread is the useful part. A Manufacturer draws as much as fourteen Constructors, and a Blender nearly nineteen. When a grid gets tight, the buildings worth counting are the ones at the bottom of that table, not the rows of small machines.

The database on this site lists every building with its draw if you need one that is not here.

The costs that are easy to forget

Three categories that sit outside the production line and add up.

Logistics infrastructure runs constantly. A Train Station draws 50 MW whether or not a train is present, and each Freight Platform adds another 50. A station with two platforms is 150 MW before anything moves. Truck Stations are 20 MW each, Drone Ports 100 MW.

Vehicles draw from their own fuel, not your grid — but the stations they use do not.

Extraction is not free. A Miner Mk.3 at 45 MW plus an Oil Extractor at 40 MW is a meaningful load for an outpost that produces nothing on site.

The spikes that trip a grid

Average load is not what breaks things. Four sources of instantaneous demand:

Locomotives. An Electric Locomotive draws between 25 and 110 MW. Four trains accelerating simultaneously briefly want 440 MW where four idling want 100.

Startup surge. Connecting a large new load before generators reach output trips the grid. Build generation ahead of demand rather than alongside it.

Geothermal fluctuation. A pure geyser swings between 200 and 600 MW continuously. If you sized against its average, the dips will find you.

The Hoverpack. It draws 100 MW and that draw does not appear on the power chart. This is the single most confusing failure mode in the game: the grid trips, you check the chart, and the chart does not show what caused it. If your grid drops while you are flying and recovers when you land, this is why.

Overclocking costs more than it looks

Production buildings scale power by an exponent of 1.321929, not linearly. Pushing a building to 250% gives 2.5 times the output for about 3.36 times the power.

Underclocking works in your favour for the same reason. A machine at 50% draws about 40% of its base, so matching a machine to actual demand rather than running it flat out and backing up is a genuine power saving, not just a tidiness one.

Generators are the exception. They scale linearly, so overclocking a generator is a neutral trade — 2.5 times the power for 2.5 times the fuel.

Some buildings have variable draw

A caution about our own data: the Particle Accelerator, Quantum Encoder and Converter show zero in the dataset we publish, because their consumption varies by recipe and is not stored as a single figure.

They are not free to run. They are among the heaviest consumers in the late game. We show a blank rather than a wrong number, and if you are budgeting a grid around them you will need to measure in game.

Sizing generation

Two rules that between them prevent most trips.

Build for roughly 1.5 times steady-state draw. That absorbs normal peaks without wasting large amounts of fuel. Generators only consume when producing, so surplus capacity costs materials to build and nothing to keep.

Add Power Storage on any grid with fluctuating input. Storage charges from surplus and discharges into shortfalls, which is exactly what geothermal variation and train acceleration need. On a geyser-heavy grid this is not optional equipment.

A useful diagnostic: a storage bank that drains steadily is telling you generation is genuinely short, which is far easier to read than a grid that trips occasionally for no visible reason.

Reading the power chart

The chart shows production and consumption over time, and two habits make it useful.

Watch the peak, not the line’s average. A grid sitting at 90% capacity with visible spikes is a grid about to trip.

Remember what the chart does not show. The Hoverpack draw is invisible on it, which is the exception worth memorising.

Splitting grids, or not

A question that comes up once a factory spans several sites: one grid or several?

One unified grid is usually right. Surplus generation anywhere covers demand anywhere, so you size against total load rather than against each site’s peak separately. Peaks at different sites rarely coincide, which means a unified grid needs less total generation than the sum of isolated ones.

The cost is a power line between sites, which is trivial — cheap, maintenance-free, and it can follow the same route as your rail or road.

The argument against is that a unified grid fails together. One trip takes everything down, including the miners feeding the generators, which is how a grid ends up unable to restart itself.

Two mitigations make the unified approach safe:

Build generation with real headroom, since the consequence of a trip is larger.

Keep a small isolated grid for critical restart infrastructure — a biomass burner or a couple of geothermal generators powering the miner and belt that feed your main coal plant. If everything else goes dark, that island lets the main plant come back without manual intervention.

Fuse behaviour

When demand exceeds supply the grid trips entirely rather than browning out gradually. Everything stops, and it stays stopped until you reset it manually from a power switch or building.

This is why the restart problem matters: a coal plant whose coal belt is powered by the same grid cannot restart itself, because the belt is dead. A storage buffer on the coal line, or an isolated island grid, solves it.

The habit worth forming: after any trip, check what actually caused it before resetting, because resetting into the same overload just trips again.

Common mistakes

Sizing against average demand. Grids trip at the instantaneous peak.

Forgetting station power. 50 MW per train station plus 50 per platform, constantly.

Blaming the grid for a Hoverpack brownout. 100 MW that never appears on the chart.

Overclocking to save space without checking power. 2.5x output costs 3.36x power on production buildings.

Reading a zero in a data table as free. Variable-draw buildings are among the heaviest consumers in the game.