AIC FACTORY OPTIMIZATION: THROUGHPUT

TL;DR - Key Points

• Port space is your hard ceiling — depot loaders waste output bandwidth; every input port costs you potential exports
• Feed machines at 100% utilization — underfeeding a moulding machine at 50% utilization halves your entire line throughput
• Belt speed is fixed at 0.5 unit/s — fast machines need dedicated lanes; slow machines should merge via convergers
• Never store throughput multipliers — process sandleaf plant to powder and feed directly; sending it to depot annihilates the 3x advantage
• Purple battery optimization doubles efficiency — $/Port/s jumps from 0.29 (naive) to 0.70 (optimized with on-site processing)
• JIT processing slashes port count — on-site sandleaf grinding reduces purple battery line from 15 ports to 10 at same output
• Splitters divide, convergers add — use this mental model to plan belt layouts and wean taxes for personal crafting

If you have been diving into the factory side of Endfield as obsessively as I have, you have probably hit that moment where your production lines look like a tangled bowl of noodles. Your batteries trickle out at a pace that makes watching paint dry feel thrilling. Maybe you have copied a blueprint or two, crossed your fingers, and hoped for the best.

But here is the thing: the AIC system is not just a cute add-on for crafting. It is the beating heart of late-game progression. Once you understand the math behind it, you can turn a sluggish, space-hogging factory into a sleek, profit-printing machine.

After more all-nighters than I care to admit — pushing past level 40 and deep into the manufacturing endgame — I have reverse-engineered what actually makes an Endfield production line tick. This is not a rehash of the in-game tutorial. It is the hard-won insight you earn after tearing down your own inefficient factories a dozen times.

If you want to stop guessing and start maximizing throughput — whether you are selling batteries for credits or powering your entire outpost network — this deep-dive guide is for you.

Related read: AIC Converger Priority: Belt & Pipe Logic goes deeper on factory guide.

The Real Endgame: Why Throughput Optimization Matters

Endfield hides its complexity behind a simple interface. Belts move at a fixed speed. Machines have fixed cycle times. Ports into and out of your AIC hub are the only way to send items to your depot or power grid.

Once you accept that every limitation is baked in, a single truth emerges: your maximum possible output is determined by how you use your ports and how you feed your machines. That is it. You cannot make a belt go faster. You cannot overclock a port. The only dial you can turn is efficiency.

Most players spend the early game happily building 1-to-1 bus lines — one refinery feeding one crafter — and they get by. But as soon as you unlock purple-tier batteries and stare down the resource demands of thermal banks, those naive setups become anchor chains.

I have seen factories where a high-tier battery line was using over a dozen ports to produce a laughable trickle of units per minute. They performed worse per port than a properly built green battery line. That is the kind of inefficiency that will have you grinding for hours just to keep the lights on.

Let us fix that. We will start by defining the language we will be speaking, then move into the three core insights that will reshape how you think about every conveyor belt you place.

See also: Arknights: Endfield PC Setup & Optimization: April 2026 for more on pc setup.

Endfield Automation Jargon: The Terms You Need to Know

Before we dive into the math, let us align on a handful of terms you will see me use. Understanding these is half the battle.

• Throughput: The rate at which finished products exit your system. Measured in units per second (unit/s). This is the ultimate number you care about.
• Flow Rate: The game’s fixed belt speed. Everything — belts, ports, machine inputs and outputs — runs at 0.5 unit/s. This is your inviolable speed limit.
• Bottleneck: The single component that is choking your entire line. Find it, fix it, and your throughput jumps until you hit the next bottleneck.
• Utilization: The percentage of a machine’s maximum output you are actually achieving. A machine that could output 0.5 unit/s but only manages 0.25 unit/s is running at 50% utilization. You want 100% wherever it matters.
• Just In Time (JIT): Inputs arrive exactly when the machine needs them, with no buffer buildup. Think of it as a perfectly timed conveyor dance.
• Saturation: A belt or buffer is completely full. Once saturated, the travel time of an item does not matter — there is always product ready to be consumed.
• Intermediate: Any item that is not a final product. Refined powders, molded plates, strands. They exist only to feed another machine.
• Port: The input/output connection between your AIC hub and the depot/power system. Each port has a hard throughput cap of 0.5 unit/s. No exceptions.
• Backpressure: When a machine’s output is so full that it cannot unload, causing the machine to halt and push the jam upstream.
• Buffer: Any storage that absorbs variance. Protocol Stashes act as explicit buffers; a long stretch of belt can also serve as a tiny, passive buffer.
• Bus: A single conveyor line carrying one type of item.

All the math I will share assumes your input belts are saturated. When the upstream supply meets or exceeds downstream demand, the actual crafting time of those components becomes invisible to your throughput calculation — the belt’s fullness does the heavy lifting.

Key Insight I: Port Space Is Your Hard Ceiling (And Depot Loaders Are Traps)

Here is the first principle that reshaped every factory I have built since: the total number of output ports you can fit into your AIC slot layout is your absolute maximum throughput. Since all ports and conveyors run at the same 0.5 unit/s, you cannot push more items per second out of your system than the number of output ports multiplied by 0.5.

This means every tile you dedicate to an input-only structure has an opportunity cost that directly lowers your ceiling.

Depot loaders are the perfect example of a noob trap. Let us do some simple port-per-tile math. A single PAC output port occupies about 3 tiles of width (counting the belt connection). That gives you 0.33 ports per tile. The dedicated AIC input port side lets you cram 7 ports along a 9-tile edge, yielding 0.77 ports per tile.

If you plop down a depot loader to pull sandleaf powder from your depot and feed it back into a line, you are wasting that space with a low-density input. Worse, you are paying the opportunity cost of not being able to use that tile for a depot unloader that would export finished goods.

A Protocol Stash, by contrast, gives you three depot inputs on a 9-tile footprint (0.33 ports per tile) without killing your output ceiling. You get storage exactly where you need it, without sacrificing export bandwidth.

The takeaway: every depot loader you use reduces your maximum possible throughput. In a game where long-term sustainability and profit scaling depend entirely on throughput, depot loaders are a self-inflicted wound.

A related rule: never squander a built-in throughput multiplier. Right now, the only such multiplier I have found involves grinding a sandleaf plant into 3 sandleaf powder, or processing seeds to turn 1 plant into 2 seeds.

If you produce sandleaf powder on-site and feed it directly into the next machine, you are tripling the output of a single input port. The moment you send that powder back to your depot and then pull it out again with a loader, you annihilate that 3x advantage. You spend two port trips (one in, one out) for the same material that could have gone straight to manufacturing on one port.

In other words: never store a throughput multiplier if you can avoid it. Always process sandleaf powder and seeds right where they are made.

Key Insight II: Pay Obsessive Attention to Ratios (Stop Underfeeding Your Machines)

Scroll through community blueprints for long enough and you will see a classic design that I have dubbed the “Layman’s Filling Machine”: one ferrium or amethyst refinery feeding a moulding machine that then feeds a filler. It looks neat, symmetrical, and it works. Barely.

The trap: it takes two units of ferrium bottles (the intermediate) to make one bottle, not one. If you are only feeding one unit of intermediate from a single refinery line, your moulding machine is starved to exactly 50% of its potential. And because the filler needs 10 bottles every 10 seconds, your whole assembly grinds to a crawl.

Let us crunch the numbers. A filler producing ferrium-based bottles requires 10 bottles and 10 ground fruit powder per 10-second cycle. A belt delivers at most 5 units every 10 seconds.

If your bottle supply belt is only getting 2.5 bottles in that window (because the moulding machine is half-starved), you are meeting just 25% of the filler’s demand. Your filler’s utilization — and thus your entire line’s throughput — sits at a miserable 25%. For amethyst bottles (which only need 5 per cycle), the same underfed setup gives you 50% utilization. Still terrible.

The correct fix is brutally simple: feed your machines fully. A moulding machine consuming ferrium bottles needs two dedicated refineries upstream just to keep it saturated. That means a 2-to-1 bus design for that stage. Suddenly your filler runs at 100% and your output per port skyrockets.

This principle extends beyond just the bottle example. In Endfield’s hard-capped world, there is no free lunch:

• Splitting one saturated input port into two machines so that each runs at 50% does not create more total throughput; it just wastes space and gives you the exact same final output as one fully fed machine would.
• Overfeeding is equally wasteful — if you send more than 0.5 unit/s to a machine that can only eat 0.5 unit/s, the excess backs up and represents a wasted port that could have been used for something else.

The golden rule: every machine in your critical production chain should be running at 100% utilization, and you calculate your bus widths to match exactly what the next machine demands.

Key Insight III: Use Conveyors Wisely — Fast Lanes Stay Separate, Slow Lanes Merge

All belts are 0.5 unit/s. This is the unchangeable heartbeat of your factory. This fact dictates a deceptively simple belt discipline: fast machines get their own dedicated bus, and slow machines should share.

Any machine that operates at the belt’s full speed — every refinery input and output, for instance — must have its own unshared belt lane. Merging the outputs of two such fast machines onto a single belt is trying to shove 1.0 unit/s into a 0.5 unit/s pipe.

The result is immediate backpressure. One machine will stall, utilization tanks, and you have turned your second machine into dead weight. Never merge fast lanes. Keep them separated like a highway.

On the other end of the spectrum, your finishing machines — fillers, packagers — take a leisurely 10 seconds per cycle. Their output rate is a mere 0.1 unit/s. If you give each of these a dedicated bus, the belt will carry one lonely item every ten seconds, wasting 80% of its capacity.

The elegant solution is to merge these trickles. A converger adds the throughput of multiple slow lines together. You can merge up to five filler outputs onto one belt (5 x 0.1 = 0.5 unit/s) and saturate that lane perfectly. This saves a lot of belt real estate and condenses your output, so you use fewer output ports.

This leads to a handy mental model I use constantly: treat splitters as division and convergers as addition. A two-way splitter divides a belt’s throughput by two. A three-way splitter divides by three. Convergers add throughputs together. This arithmetic becomes your planning tool.

Need to siphon off a tiny fraction of your amethyst plates for personal crafting without hurting your main battery line? Chain a series of splitters: a 1/2 split followed by another 1/2 gives you a 1/4 branch. One more gives 1/8.

Return the unused fraction back to the main line with a converger, and you have “taxed” your production by only 1/8th. The line remains effectively saturated, yet you slowly build up a stash of plates. I call this a wean tax. It is a lifesaver for keeping your crafting depots topped up without tanking your economy.

Putting the Principles into Practice: Battery Production Math That Will Save Your Factory

Nothing illustrates the raw power of these insights better than comparing battery production line configurations side-by-side. I have compiled a quick reference table based on real builds and hard numbers.

It covers three battery tiers — Green, Blue, Purple — each shown at the sloppy utilization you see in naive builds, then at a fully optimized 100% state. I have included a metric I call $/Port/s, which measures how much money each output port is effectively printing per second.

(Calculations assume thermal banks consume 1 battery every 40 seconds, so power output scales directly with battery throughput.)

Study those numbers. A naive purple battery line is the worst item in the game to manufacture on a per-port basis, putting out 0.29 $/Port/s and using four ports to dribble out one battery per minute.

Compare that to a fully optimized purple line with on-site sand processing: 0.70 $/Port/s, ten ports, and 6 units per minute. That is more than double the efficiency.

Even a green battery line, with its humble output, will outperform a badly built purple line in terms of money per port. If you are in the late-game grind and wondering why your economy feels sluggish, this is your smoking gun.

The real magic shows when you bake the sandleaf powder multiplier directly into production. By refusing to store sandleaf powder and instead feeding it straight from the grinder into the crafting chain, you slash the port count for the purple line from 15 down to 10 while maintaining the exact same output. That is the power of JIT thinking and port discipline combined.

Building Your Own Optimized Factory: A Practical Methodology

Now that you have the theory, here is how I approach a fresh production line from scratch:

1. Start from the end. Decide what your final product is and how many units per minute you want. Work backward, machine by machine, calculating the exact input requirements.
2. Map your belts like an equation. For each machine, determine if it needs a dedicated 0.5 unit/s bus or if it can share. Fast machines (cycle under 2s) get their own lane. Slow machines (10s cycles) get merged.
3. Count your ports. Every time you introduce an input from the depot or send a product out, mark a port. Optimize the layout to maximize output ports on the 9-tile edge. Minimize loaders.
4. Respect the multiplier. If a process creates more than one unit per input (sandleaf grinding), keep it on the same local belt loop. Never let those bonus units take a round trip through the depot.
5. Test for bottlenecks. Run the line and watch the machines. If one is pausing with a full output, you have backpressure downstream. If it is idle with an empty input, you are underfeeding. Adjust splitters and belt counts accordingly.
6. Apply wean taxes carefully. Decide which intermediates you want to stockpile for manual crafting. Use fractional splitter chains to bleed off a tiny portion without starving the main line. I find 1/8 or 1/16 is usually safe.

Remember: inefficiency in the early game is forgiving. You will still churn out enough resources to limp forward.

But the moment you step into grinder hell — where purple batteries and complex intermediates dominate — every wasted port, every starved machine, every merged fast lane compounds into hours of lost productivity.

You can either fight the math or embrace it. I chose the latter, and my factory now runs so smoothly I can finally get some sleep. Well, almost. There is always another line to optimize.

Now, go forth and build something monstrous. Your thermal banks are waiting.

This guide reflects personal builds and community-sourced data as of May 2026. Game mechanics may change with future updates. Always verify ratios against your own AIC reports. For a step-by-step companion to power, batteries, and resource loops, see Factory Foundations: Stable Power & Resource Loops, or use the Industrial Planner to model your throughput before you place a single belt.