For a studio earning between $200,000 and $1 million a year, Unity is the most expensive of the three big engines by license cost alone. Almost nobody budgets for that.

The arithmetic is short. Once revenue and funding pass $200,000, Unity wants a Pro seat for every person who opens the editor, and Pro has cost $2,200 per seat per year since January 2025. Five people is $11,000 every year, whether the game sells or not. Unreal asks for nothing until a single product has grossed $1 million over its lifetime, and after that it takes 5 percent of what sits above the line. Godot is MIT licensed, so it asks for nothing, ever. Take a studio at $400,000 a year, and the three bills are $11,000, $0, and $0.

I am not telling you to drop Unity. I have shipped on Unity, Unreal, and Godot since 2015, and Unity is still the one I open most often. What I am telling you is that the license line is one of seven places where an engine pick quietly costs money, and it is not even the largest. Most advice on this topic reads like a checklist with seven equal boxes. Mine will not, because in my experience a wrong engine never hurts in equal portions. One mistake costs you an afternoon, another costs you a launch window, and you rarely know which is which until the invoice arrives. Seven mistakes follow. They are uneven in length because they were uneven in damage.

Do not pick the trailer’s engine

A roguelike deckbuilder started in Unreal. A mobile idle game started in Unreal in 2022, because the team had just watched a Nanite demo. Neither game needed anything the engine is famous for, and both lost months to editor weight and package size. Unreal likes third-person 3D with real lighting. Unity is comfortable with 2D, mobile, and odd ideas. Godot is excellent at 2D and still growing into large 3D scenes. Before anyone opens an editor, write one page: genre, camera, platforms ranked by revenue, and the three systems you dread building. The trailer does not get a vote. Neither does the engine your favorite streamer uses, and neither does whatever won the last big game jam. I have watched all three sway a studio meeting, and not one of them has ever paid a salary.

Do not choose an engine you cannot hire for

In the summer of 2022, my studio tried to hire a mid-level Godot gameplay programmer for a six-month contract. It took four months. For those four months I wrote the gameplay code myself, at night, and badly. The candidates we did find wanted triple the rate or were already booked through the winter, and the two who were free had never shipped anything on console.

Two years later, a Unity port outgrew our two programmers, and we had three qualified contractors within a week. Agencies that sell unity game development services exist because that pool is wide, and renting a few engineer-months for a feature you cannot staff beats a four-month search. Try getting the Godot equivalent on two weeks’ notice.

The GDC 2026 State of the Game Industry survey makes the same point in numbers: 42 percent of respondents said they work in Unreal, 30 percent in Unity, and 5 percent in Godot. That sample leans toward larger studios, so I would guess the indie split is friendlier to Unity and Godot. That guess is mine, not the survey’s. The ranking at hiring time, though, tends to look like the survey.

I also keep a rough rule for team size. Below three programmers, a thin hiring pool hurts less, because you can train a good generalist in a month. Above three, every open seat becomes a schedule risk. That rule is mine and I have no survey behind it.

Post an imaginary job ad before you commit. Count the credible replies you would get in a week. If the honest answer is “one, maybe,” the engine has to be dramatically better for your game to justify the wait.

Read the license again every year, and model it

On September 12, 2023, Unity announced a Runtime Fee of $0.20 per install, charged once a Personal or Plus game passed $200,000 in revenue and 200,000 lifetime installs. We had a title on Steam. I spent that week with a spreadsheet, a lawyer friend, and a headache.

Twelve months later, on September 12, 2024, Unity canceled the fee, doubled the Personal ceiling from $100,000 to $200,000, and raised Pro 8 percent to $2,200 per seat. Fine outcome. I never got the week back, and I stopped assuming terms stay put. Since then I keep a one-page file for every engine we use, with the date I last read its terms and the date the terms were last changed. It takes ten minutes a year, and it has already caught two quiet edits to tier names.

Unreal is simpler on paper. Epic charges 5 percent of lifetime gross revenue above $1 million per product, and sales through the Epic Games Store are royalty-free. Two details bite. The base is gross, so Valve’s cut does not reduce it. And the threshold is lifetime, so a game that earns $400,000 a year crosses the line in its third year, and from then on every dollar carries the 5 percent. Godot owes you nothing and you owe it nothing.

Here is the table I now build for every new project. Assumptions: five seats, one product, first year of sales, all revenue gross through Steam rather than the Epic store, and Unity Pro required above $200,000 in revenue and funding.

Annual revenue Unity Unreal Engine 5

Godot 4

 
$100,000 $0 (Personal tier) $0 $0
$400,000 $11,000 (five Pro seats) $0 $0
$1,500,000 $11,000 $25,000 (5% of the $500,000 above $1 million) $0

Look at the last row, because it surprised me the first time I ran it. Once a game becomes a real hit, Unreal costs more than Unity, and in year two of that same hit the gap grows, since the royalty applies to the whole year’s gross. Which engine is cheapest depends on where you land, and that is the whole point. If you expect to stay between $200,000 and $1 million, Unity is the expensive one. If you expect a breakout, Unreal can be. If you expect neither, you are choosing on features and people, which is how it should be. Notice what the first row hides, too: a studio at $100,000 pays nothing anywhere, which is exactly why teams stop looking. The bill arrives the year revenue does. A model you can rerun in an afternoon is the only protection I know against terms changing under you.

Do not trust “exports to Switch”

A founder asked me last spring whether his engine exports to Switch. I said yes. He said, “So we’re covered for console.” I said no, and he asked what I meant.

What I meant was the paperwork. Unity and Unreal both ship console modules behind NDA, and you need registered developer status with Nintendo, Sony, or Microsoft to download them. For a new studio I have seen that take two to four months. He said that was fine, he would start the application now. Good. I added that the application wants a business address, a security review of the office, and sometimes a publisher’s letter, so a home-office studio should start earlier than he thought. Then he asked about cost.

Godot is where the bill changes shape. There is no public console export, so you buy a toolchain or hire a porting house. W4 Games, a company founded by Godot’s creators, announced console pricing from $800 to $10,000 per team per year in December 2023, depending on tier and number of platforms. Check the current page before you quote it to an investor, since I am reporting a 2023 announcement and prices move.

He asked whether that settled it. Not quite. Our own Unreal game failed Nintendo certification in October 2021 over suspend and resume handling, a bug in our save system that assumed the app would never freeze mid-write. Six weeks and one embarrassed resubmission later, we passed. The engine was fine. Nobody had budgeted for those six weeks, and the port is its own project. Add a line for it in the budget, a line for certification retries, and a buffer of at least two months before your planned launch date.

Build the scary part twice

In early 2024, a two-person team asked me to help them pick between Unity and Godot for a rope physics puzzle game. I told them to stop debating and build the rope in both. This is how the five days went, as they told it to me afterward.

Day 1 went to setup. Same rope spec in both engines: twenty segments, a hanging weight, a hook the player drags. Nothing else. No menus, no art, no sound. They agreed on one test before writing a line: drag the weight hard across a wall corner and see whether the rope survives.

By Day 2, the Unity rope worked. It swung, it snagged on geometry, and it stayed stable when they dragged hard. They called that a pleasant surprise and kept going, adding a second rope to see what would happen. Still stable. They wrote “boring” in their notes, which is the best word a physics prototype can earn.

Day 3 was Godot’s turn, and the first version looked fine at rest. Under a hard drag it stretched like taffy. They spent the afternoon tuning solver iterations and joint settings, and the fix held for about ten minutes before the next odd case broke it. A forum thread from a year earlier described the same symptom and ended without an answer.

Come Day 4 they had a rope that held together, mostly. With two ropes and a weight, it jittered under load in ways that looked like they would need a custom solver.

On Day 5 they wrote down what they had learned, with a table of hours spent per task, and chose Unity. The Godot column was not bad, only longer. Their game shipped in 2025, and the rope is still the part reviewers mention first. In the retro, they told me the real surprise was how little the prototype had to prove. Nothing in it was pretty, yet it answered the one question that mattered, and it answered it before they had spent a single dollar on art. By their own estimate, ten days of prototyping saved three to four months of fighting the physics later.

Here is my inference, and the team never claimed it: the rope was the hardest system, so the gap on the rope would have shown up on every system that touched it. A feature chart cannot tell you that. A week of building can. One more detail from that week. They did not polish anything, and they threw both prototypes away afterward, which they said felt wasteful for about an hour. Prototype the thing that scares you, not the movement controller, which will work anywhere.

Watch the clock on the Play button

Iteration time is a design budget, and the engine sets it. Our 2021 Unreal project began with cooks of about 90 seconds. By the final month a full cook took 38 minutes, and a shader recompile after one material tweak could stall an artist for ten. A shared derived data cache on a NAS cut the pain in half, two months late. Unity has its own version, where entering play mode creeps toward 25 seconds as the codebase grows. Godot is nearly instant on small projects and loses that edge only at scale. On day one every engine is fast, so test at month twelve size.

Audit the people behind the engine

Years ago a networking plugin underneath our multiplayer layer lost its single maintainer. Replacing it took five weeks. An engine is the same kind of dependency, a thousand times bigger, and it deserves the same audit.

Epic funds Unreal with Fortnite money, which means a lot of money tied to one company’s priorities. Unity is a public company that has changed CEOs and pricing more than once in three years. Godot runs on a nonprofit foundation, which reported €49,678 in monthly recurring income from 1,518 individual donors and 18 corporate sponsors at the end of 2024. Healthy for open source, small beside a studio-backed engine. When a Unity error hits at 2 a.m., somebody on a forum solved it in 2017. With Godot 4 I have sometimes been the second person to see the error.

Three questions people ask at every GDC

Someone always asks whether it is too late to switch engines mid-project. My answer is that the answer turns on one thing, which is a finished vertical slice. Before that, switching is cheap and a week of prototyping tells you whether to. After it, you are paying for a second game, and I would only do it if the first engine is physically blocking a launch platform.

The second question is whether Godot is a safe bet for a commercial game. For 2D and small 3D, yes, and I have watched it carry real revenue. The risks are console ports, hiring, and large 3D scenes, all of which you can price before you start. The one thing I would not do is pick Godot only because it is free. Free licensing saves you five figures a year at best, and a slow hire or a custom physics solver can eat that in a month.

Last comes whether one person can ship on Unreal. Yes, if the game fits what Unreal does well and you accept longer builds and a heavier editor. I have seen solo developers do it. I have also seen more of them quit around month eight than on any other engine, usually after a cook time they never planned for. That is a pattern from the projects I know, not a statistic, so weigh it accordingly.