The Biggest Bottleneck Wasn't My Conveyor Belt

I want to tell you about the week I lost to a Steel factory, because I made the same mistake five nights in a row and never once suspected I was solving the wrong problem. The factory kept stalling. Every instinct I had said belt. Every instinct I had was wrong. The bottleneck was real, it was severe, and it wasn't moving on any conveyor — it was measured in megawatts, and I couldn't see it from inside the factory at all.

A Steel line that wouldn't stay fed

Steel is where a lot of us first feel like real engineers. Iron Ore and Coal go into a Foundry, out comes Steel Ingot, and from there it forks into Steel Beams and Steel Pipes that feed half the mid-game. I'd built a tidy little Steel setup, and now I wanted more of it — Encased Industrial Beams were calling, and those want Steel Beams by the crate. So I did the obvious thing and scaled up. More Foundries. More Constructors behind them. More everything.

And the expanded factory would run beautifully for a minute or two, and then die. Not slow down — die. Belts empty, machines dark, the whole thing just stopped, like someone had unplugged it. I'd walk the line, find nothing obviously broken, and eventually it would come back on its own or with a nudge, run for another minute, and stop again.

I looked at the emptiest belt — the one feeding my new Foundries — and drew the natural conclusion. Not enough throughput. The belt couldn't carry the ore fast enough. Classic bottleneck. Easy fix.

Five nights of upgrading belts

So I upgraded the belt a tier. It ran, it died. Huh. Upgraded the belt again — surely now — it ran, it died. I started upgrading belts I hadn't even suspected, on the theory that the real limit was hiding one conveyor over. I re-routed. I split lines to spread the load. I spent one entire evening convinced the problem was a single stretch of belt between my miners and my smelters, and I rebuilt that stretch twice.

Here's the detail that should have stopped me on night one, the detail I kept looking straight past: the belts were never full. When the factory stalled and I checked the conveyor I'd just "fixed," it wasn't backed up and it wasn't saturated. It was empty. A belt that's the bottleneck is a belt running at capacity with a queue behind it. My belts were sitting there with nothing on them, which is not what a throughput problem looks like at all. It's what a factory looks like when nothing upstream is running — when the machines themselves have stopped. But I'd decided it was the belt, so that's the only thing I could see.

The fuse I forgot I had

It was the timing that finally gave it away. The stall always came a beat afterthe factory hit full tilt — after every new Foundry had spun up and started pulling its full draw at once. And a Foundry is not a cheap machine to run. Each one pulls sixteen megawatts, and I'd added a row of them, plus all the Constructors behind them, and I had not added a single generator. I'd spent the whole week thinking about items per minute and none of it about the number that actually matters when a factory dies all at once: total power draw against total power capacity.

What was happening wasn't a supply stall. It was a fuse trip. Every time the expanded factory reached full production, the combined draw crept past what my coal generators could supply, the grid tripped, and everything — every machine, on every belt, all at once — powered down. Which is exactly why it looked like the belts had emptied: they had, because the machines feeding them had gone dark in the same instant. A power trip and a supply starvation look identical from the factory floor. Both give you dead machines and empty belts. The only difference is the cause, and the cause was the one number I'd never once looked at.

I added generators. The factory has never stalled since. A week of belt upgrades, and the fix was megawatts.

Two different kinds of "not enough"

The lesson underneath this one is narrow but it's saved me a dozen times since: a factory can starve in two completely different ways, and they feel the same and demand opposite fixes. One is a throughput problem — a belt genuinely can't carry enough, and the tell is a belt running full with a backlog behind it. The other is a power problem — you have every resource you need and not enough electricity to run the machines that use them, and the tell is everything stopping at once, usually right when the factory reaches full load.

Upgrading a belt to fix a power deficit is like widening a road to fix an empty gas tank. The road was never the problem. And the reason this is so easy to get wrong is that power is invisible while you build. You watch belts fill. You watch machines whir. You do not watch megawatts — there's no conveyor for electricity, no satisfying stream of it to eyeball. It's just a number on a panel you have to go and check, and a number you have to remember to add up beforeyou flip the switch, because the game only tells you you've overdrawn by turning everything off.

The resource with no belt

I've come to think power is the sneakiest thing in the game precisely because it's the one resource that doesn't travel on a conveyor. Everything else you can watch. Iron moves, and you see it move. Plastic backs up, and you see the belt fill. You develop an eye for supply because supply is physical and right there in front of you, streaming past at a rate you can practically feel. Power isn't like that. It doesn't queue, it doesn't stream, it doesn't give you a satisfying backlog to point at. It's just a total on a panel, ticking toward a ceiling you can't see until you hit it.

Which is why the Steel week wasn't really about Steel. It was about the fact that I'd built an intuition for the resources I could see and none at all for the one I couldn't. Give me a starving belt and I'll read it in a second. Give me a factory that draws four megawatts more than I generate and I'll spend five nights blaming the conveyor, because nothing in the factory looks like "four megawatts short." It looks like a factory that stops.

The place this really paid off was aluminum, which is the part of the game most likely to punish you for exactly this mistake. The aluminum chain is a monster of Refineries and Foundries — Alumina Solution, then Aluminum Scrap, then Ingots — and Refineries are thirty megawatts each. String enough of them together and the power bill is genuinely enormous, far more than the machine count makes it feel. The old me would have built the whole gorgeous thing, flipped it on, and watched the grid fall over. Instead I added up the draw first, saw the ridiculous number it came to, and built the generators to match beforeI laid a single pipe. It powered on and stayed on. No drama, no blackout, no week. Just a factory that ran because I'd respected the invisible input this time.

Knowing the draw before the fuse does

After that week I changed one habit. Before I scale a factory now, I work out what the new machines will actually pull — not roughly, not "a lot," but the real number — and I make sure the generators are there to cover it before I place the first new Foundry. Sixteen megawatts a Foundry adds up frighteningly fast when you're building a row of them, and it's so much cheaper to learn that from a number than from a blackout.

That's really all the power planning I do: add up the draw of what I'm about to build, compare it to what I can generate, and close the gap on paper instead of on the factory floor. It's not clever. It's just the one check that turns "why does my factory keep dying" into a question I answer before it happens rather than after — the megawatts a build will pull, worked out while it's still a plan, so the fuse never has to teach me the hard way.

I still love a good conveyor. But I've learned to check the belt is actually full before I blame it — and to count the megawatts first, because half the time the belt was innocent and the power was the whole story.

See you in the factory.