What the UE4 to UE5 adoption curve tells us about UE6

Everyone wants to know when to rebuild for the next engine

If you sell assets on Fab, the most expensive mistake you can make is mistiming an engine transition. Rebuild too early and you pour weeks into a version almost nobody is shipping yet. Rebuild too late and the buyers who would have paid have already bought from the seller who moved first. The question underneath every roadmap argument - when do I commit to the next engine - is really a question about a curve, and for once we have the curve.

We have four years of monthly data on how new Unreal asset releases migrated from UE4 to UE5. That single transition is the closest analogue we have to whatever comes next, so this report does three things. It tells the UE4 to UE5 story straight from the numbers. It extracts the shape those numbers trace - a classic S-curve - and pins down the parts of that shape that are robust. Then it projects that shape, explicitly and hypothetically, onto a future UE5 to UE6 move so you can reason about timing instead of guessing.

One ceiling to state plainly before any number appears, because it governs how every figure here should be read. This dataset is the MythicLemon Marketplace Index, which measures the share of new Fab asset releases each month that target each engine version. It is a proxy for marketplace supply - what sellers choose to publish - not a census of all Unreal developers, and not a count of shipped games. When we say UE5 reached 90 percent, we mean 90 percent of new listings, not 90 percent of the world's UE projects.

The UE4 to UE5 story, straight from the numbers

Before UE5.0 shipped as a stable release, its marketplace footprint was noise. Through the Early Access window UE5's share of new releases bounced around in the low single digits, roughly one to four percent - the signature of a handful of early adopters experimenting rather than a market moving. UE5 entered Early Access in May 2021, but Early Access is not where a marketplace commits.

The real clock starts at the UE5.0 stable launch on 5 April 2022. That month, UE5's share of new releases jumped to 14.1 percent - the first genuine break from the noise floor. By December 2022 it had more than doubled to 32.5 percent. The first month UE5 crossed half of all new releases was October 2023 at 53.1 percent, roughly eighteen months after the stable launch, and it then oscillated around the halfway mark into early 2024 before settling durably above 50 percent from January 2024 at 56.2 percent.

From there the curve climbed steadily and then flattened into a ceiling. January 2025 reached 75.0 percent; August 2025 hit 87.3 percent. By 2026 the transition is effectively complete on the supply side: 90.6 percent in January 2026, 92.8 percent in March, a peak of 94.3 percent in April, and 91.7 percent in May 2026, the latest data point. The published migration dashboard headlines a smoothed figure of 93.2 percent for UE5 in new releases; the latest single monthly point is 91.7 percent. Both are honest cuts of the same trend - read them together as roughly 90 percent and climbing, not as a contradiction.

Notice what did not happen. New releases did not flip to UE5 overnight when the stable build dropped, and the last stragglers did not vanish either. Four years on, the catalogue as a whole is still 46.7 percent UE4 by listing, even though new releases are overwhelmingly UE5. Supply of new work and the standing back-catalogue move on very different timescales, and conflating them is the single most common error in these debates.

The shape that matters is an S-curve, not a date

Plot those points and they trace a textbook S-curve: a slow noisy toe, a steep middle where the majority of the market converts in a compressed window, and a long flattening shoulder as the last laggards trickle over. This is the same diffusion shape that describes how most technologies get adopted, and recognising it is what lets you forecast responsibly. The dangerous temptation is to forecast a calendar date for the next engine; the defensible thing to forecast is a duration measured from a launch you can actually see happen.

Stated as durations from the UE5.0 stable launch, the milestones are clean and memorable. About 1.5 years to reach a majority of new releases. About 3 years to reach roughly 75 percent. About 4 years to reach roughly 90 percent. Those three numbers are the robust, transferable core of this entire report. They describe the rhythm of an Unreal marketplace transition independent of what month it started, which is exactly the property you want in a template.

It is worth being precise about what is robust and what is not. The shape is robust - diffusion curves are S-shaped almost regardless of the technology. The durations are reasonably robust because they come from a real, recent, same-marketplace transition rather than a textbook abstraction. What is not robust is any specific calendar prediction, because that requires knowing the start date, and the start date of the next transition does not yet exist.

Projecting the curve onto a hypothetical UE6 - read this as a scenario

Everything in this section is an explicit projection, not a forecast of fact. As of this writing UE6 is unannounced: it has no official name, no version number, and no release date, and the newest engine available is UE 5.8. Nothing below should be read as a claim that UE6 exists, ships on a given date, or includes a given feature. The exercise is narrow and useful: if a future engine transition followed the same diffusion curve UE4 to UE5 did, what would the supply share look like month by month after its hypothetical stable launch?

To make that concrete we anchor the curve to a hypothetical launch month - call it T0 - and read the projected new-release share off the UE4 to UE5 template at each interval. The scenario table below does exactly that. At T0 the new engine is a sliver and the prior engine still owns nearly everything; by roughly T0 plus 18 months the two are near parity; by around T0 plus 48 months the new engine dominates new releases and the prior one has become the back-catalogue. The shape is borrowed wholesale from the transition we measured.

Treat the table as a planning lens, not a timetable. Its value is the relative spacing of the milestones - majority near month 18, three-quarters near month 36, ninety percent near month 48 - because that spacing is what the real data supports. Slide the whole curve forward or back to whatever T0 eventually turns out to be. The next section is about why the real curve might run faster or slower than this template, which is where the judgement actually lives.

Why the next curve could run faster - or slower

Several forces argue for a faster transition than UE4 to UE5. The marketplace has institutional memory now: sellers who were slow to move to UE5 watched faster rivals capture the early demand, and that lesson tends to compress the toe of the next curve. The store itself is more consolidated - Fab replaced the old Marketplace and unified the storefront, so there is a single place where version expectations propagate quickly. And AI-assisted migration is real and improving: agent tooling that can mechanically chase down API breaks, fix up redirectors, and re-validate a project lowers the cost of moving, and lower migration cost steepens the middle of the curve.

Pulling the other way are forces that argue for a slower transition. The UE4 to UE5 move had an unusually strong technological pull - Nanite and Lumen were generational, headline features that gave buyers a concrete reason to demand UE5 content, which dragged the curve forward. An engine transition without an equally visible pull would migrate on inertia alone, and inertia is slow. A new scripting or runtime layer carries its own learning curve; Verse and AutoRTFM, which Epic has signalled as part of Unreal's direction, are genuinely new models to learn, and learning curves flatten adoption before they accelerate it.

And then there is the sheer mass of the existing UE5 catalogue. The reason 46.7 percent of listings are still UE4 four years after the UE5 launch is not that those sellers are lazy - it is that a large, proven, revenue-generating back-catalogue exerts real gravity. Every successful UE5 product is a small future drag on the next migration, because rebuilding something that already sells is a cost with uncertain upside. The net of these forces is genuinely uncertain, which is the honest answer; the toe and the middle of the next curve could plausibly be faster, while the long shoulder is likely to be just as stubborn.

What the curve means for when to rebuild

Translate the shape into seller decisions and a clear playbook falls out. The expensive errors cluster at the two ends of the curve. Rebuilding during the noisy toe - the first few months after a launch, when new-engine share is in the low teens - means building for a market that has not arrived; you carry the maintenance cost of two versions while almost no one is buying the new one. Waiting until the shoulder - past the 75 percent mark - means arriving after the early-mover demand has already been captured by someone else.

The defensible window is the steep middle, roughly the run from majority to three-quarters, which in the UE4 to UE5 data was the stretch from about 18 to about 36 months after launch. That is when demand for new-engine content is rising fastest and competition has not yet saturated. The practical move is to watch your own category's monthly new-release share rather than the whole-market average, because categories convert at different speeds, and to have a port ready to publish as your category crosses the 40 to 50 percent line rather than starting the port then.

There is also a back-catalogue play the curve makes obvious. Because the standing catalogue lags new releases by years - 46.7 percent UE4 even now - a proven UE4 product cleanly rebuilt and properly documented for the current engine often beats inventing something new. The demand already exists and is visible; the buyers are the ones still on the older version asking whether anyone will support theirs. The lowest-risk transition strategy is usually to migrate your existing winners early in the steep middle, not to chase a brand-new concept on a brand-new engine.

Where agent-assisted migration fits

The single accelerant a seller actually controls is migration cost. The market timing, the feature pull, the consolidation of the store - those are external. How fast and how cheaply you can move a project from one engine to the next is the one lever in your hands, and it is the lever that decides whether you can hit the steep middle of the curve instead of trailing into the shoulder.

This is where agent-native editor tooling earns its place. MythicDevAssist (MDA) is an MCP bridge that lets an AI agent observe and act inside the Unreal editor - read the live editor state, run editor actions, and check its own work against what the editor actually reports rather than guessing. For a migration that means an agent can drive the repetitive, error-prone parts of a port - chasing deprecated calls, re-validating assets, confirming a rebuild opened clean - with a human approving the changes. Lower migration friction is exactly what lets a small seller treat the next engine transition as a window to exploit rather than a tax to dread.

To be precise about what shipped where: UE 5.8 itself contains a first-party agent stack - an experimental, editor-only AI Assistant and a ToolsetRegistry that exposes engine functions to a language model as tools - but that built-in assistant is off by default, marked NoRedist, and gated as Epic-internal in the shipped binary, so it is not a flip-a-switch public feature today. UE 5.8 does not ship MCP support; the in-editor agent transport is Epic's own runtime, not the open Model Context Protocol. MCP remains the open, client-agnostic route, available through third-party bridges like MDA. None of that changes the migration argument - it just means the open, available tooling is where the practical leverage is right now.

UE4 to UE5 - share of new Fab releases on UE5, month by month

MythicLemon Marketplace Index. Share of new Fab asset listings each month targeting UE5. This is marketplace supply, not a census of all Unreal developers or shipped games. The 93.2% smoothed dashboard figure and the 91.7% latest single point are two honest cuts of the same trend.

Hypothetical UE5 to UE6 scenario - projection only, not a forecast

Illustrative only. UE6 is unannounced, unnamed, unnumbered and undated; the newest engine is UE 5.8. This table slides the measured UE4 to UE5 S-curve onto a hypothetical launch month (T0). Shares are projected supply (new-release share), not facts. Anchor T0 to whatever a real launch turns out to be.

FAQ