DEV Community

dev game
dev game

Posted on

Why ARC Raiders Bet on Unreal Engine 5: The Ex-DICE Divorce Behind a 12M-Copy Extraction Hit

Short answer: ex-Battlefield devs quit their own engine to hire faster

ARC Raiders is a 2025 multiplayer extraction shooter developed and published by Embark Studios, released on October 30, 2025 for PlayStation 5, Windows and Xbox Series X/S, built on Unreal Engine 5. It can be played solo or in teams of two or three.

That engine choice looks boring until you remember who made it. Embark was founded in Stockholm on November 8, 2018 by former DICE staff including Patrick Soderlund, the former CEO of DICE and Chief Design Officer at Electronic Arts, alongside Magnus Nordin, Rob Runesson and others. The studio now employs about 350 people and has been a Nexon subsidiary since 2021.

In other words: the team that grew up inside DICE — the studio that built Frostbite, the proprietary engine first released in June 2008 for Battlefield: Bad Company and since used across EA studios — chose not to build another Frostbite. They chose Epic's UE5 instead, the engine revealed in May 2020 and formally launched for developers on April 5, 2022 with Nanite virtualized geometry and Lumen dynamic global illumination.

Most coverage says they did it for graphics. I think that misses the point. My read, labeled clearly as analysis throughout: Embark bought a labor market, not a renderer. UE5 was the fastest way for 350 ex-Battlefield developers to hire, outsource, test, and ship two live games without maintaining an engine company inside a game company.

That decision paid off commercially. ARC Raiders sold more than 12 million copies and reached 960,000 concurrent players, making it the biggest launch for a Nexon game. It also won Best Multiplayer Game at The Game Awards 2025.

What ARC Raiders actually is: 30 minutes to get rich or lose everything

You start in Speranza, a makeshift subterranean city under post-apocalyptic Italy. You go Topside during low ARC activity to scavenge old-world goods and downed-machine parts, then sell, dismantle or craft them to keep Speranza running.

A raid lasts up to 30 minutes, extended to 40 minutes on conditions like Hidden Bunker and Locked Gate. You face hostile NPC robots called ARC plus other human raiders. If you die you lose what you carried except loot in a safe pocket, and you return to base. You get out through extraction elevators, metro stations and air shafts, or safer raider hatches that require keys.

Back home you sell items, upgrade perks and skills, buy gear, and take vendor quests that usually ask you to find special items or visit locations while exposed. Matchmaking generally puts solo players against other solos, with a similar duos adjustment deployed on November 12, 2025, and a January 27, 2026 option letting solos join trios queues for higher rewards.

The setting is Earth in 2180. Decades earlier, machines identified by the shape of their radar signatures — ARC — fell from space and forced survivors underground. You are a raider from Speranza, one contrada of Toledo, and raiding is unregulated: cooperation is encouraged, theft is allowed.

Development history matters here. ARC Raiders was announced at The Game Awards 2021 as a free-to-play cooperative third-person shooter. In 2023 the genre changed to PvPvE extraction shooter. After a closed alpha under NDA, Embark ran a public Tech Test on October 24, 2024 and Tech Test 2 on April 30, 2025, then announced the game was dropping free-to-play for a $40 purchase. A free Server Slam ran October 17-20 before launch.

That $40 flip is usually framed as business. I read it as architecture, and I will defend that below.

Why Unreal Engine 5 fit an extraction shooter built by ex-Frostbite people

1. Nanite and Lumen as anxiety tools, not beauty tools

UE5's headline features are Nanite, which automatically manages level of detail by scaling meshes dynamically and partitioning them into 128-triangle clusters called meshlets, and Lumen, a dynamic global illumination and reflections system using software and hardware ray tracing.

Black Myth: Wukong used that stack to show off. ARC Raiders uses it to hide things.

Topside is junk-dense by design: rubble, rusted suburbs, broken Tenerife-inspired sprawl. Nanite lets a mid-size team import film-quality clutter without hand-optimizing every LOD. Lumen makes darkness directional — you hear a robot before you see it, then a headlamp sweep gives you half a second. Most outlets praise the lighting as pretty. My take: the prettiness is a side effect. The real win is readable clutter. In an extraction shooter you need 50 pieces of cover that all look different but behave the same. Nanite gives you visual variety for free; Lumen gives you threat direction without UI arrows.

Cost: that density is expensive to read. Reviews noted the UI feels built for mouse and keyboard, with tiny fonts and little direction on controller. When every pile could hide a spring trap or a player, new players freeze. That is a UE5 tax: you can render everything, so you do, and then you must teach restraint.

If you want to feel that corner-check pressure in a browser before risking a 30-minute raid, try a quick run in BodyCamera Shooter. It is a single-player browser FPS where you clear buildings through a shaking chest camera with fisheye distortion. It punishes standing still in the open and rewards short bursts, cover pops, and listening before peeking — the same habits that keep you alive Topside.

2. Quitting Frostbite was a hiring decision

Frostbite is a brilliant destruction engine developed by DICE. It powered Battlefield: Bad Company onward, with HDR Audio, Destruction 1.0 to 3.0, deferred rendering and real-time radiosity.

It is also a company-town engine. If you build on Frostbite outside EA, you must train every hire, maintain every integration, and explain your crash reporter to every outsourcer.

Embark's founders knew that cost intimately. Soderlund left DICE in August 2018 after nearly two decades on Battlefield, Star Wars Battlefront and Mirror's Edge. By 2019 Embark was already over 50 people, explicitly focused on procedural content, player collaboration, machine learning and cloud. Nexon took a 66.1% stake in 2019 and completed acquisition in 2021.

Choosing UE5 gave that team instant hiring, instant documentation, instant Quixel assets, and instant console support for PS5 and Xbox Series X/S. The Finals proved the pipeline first: a free-to-play first-person shooter with fully destructible environments, also on UE5, that attracted more than 10 million players in its first two weeks after its December 7, 2023 launch.

Common wisdom says ex-DICE devs would want their own tech to control destruction. My contrarian read: they had already paid that tax once. Frostbite's destruction is a full-time job. UE5's Chaos plus replication let them buy 80% of the feel for 20% of the maintenance, and spend the savings on extraction economy — safe pockets, keys, vendor quests, elevator timing — which is where ARC Raiders actually lives or dies.

For the team-play side of that economy, Bit Gun.io is a useful warm-up. It is a 5v5 browser FPS with loadouts, recoil patterns, ranks and small maps. Use it to practice the unglamorous ARC skill: calling positions, timing pushes, and not chasing kills when holding map control matters more.

3. The $40 price is server architecture in disguise

Dropping free-to-play for $40 at the last minute looked like cold feet. I think it was honest engineering.

Free-to-play extraction at 960,000 concurrent players means free cheating, free server bills, and free content hunger. A $40 gate does three technical jobs: it filters day-one cheaters better than any kernel driver, it funds the bi-annual major update cadence Embark announced in May 2026 starting with Frozen Trail in October 2026, and it lets designers balance around loot scarcity instead of battle-pass engagement.

The Finals could afford free-to-play because a 10-minute arena match is disposable. A 30-minute ARC raid is not. When death costs 25 minutes of loot, every lag spike, exploiter, and server crash is personal. Charging upfront buys slower, more deliberate patching focused on bug fixes, exploit closures and balance — exactly what post-launch patches did through 1.26.0 on April 28, 2026.

No other UE5 extraction hit made this trade so visibly. That is why ARC Raiders, not The Finals, became Nexon's biggest launch.

4. Less destruction than The Finals — on purpose

Here is my sharpest disagreement with the usual take. People expected ARC Raiders to out-destroy The Finals. It does not, and that is correct.

The Finals is about remaking the arena every 90 seconds. ARC Raiders is about remembering the arena for 30 minutes. If every landmark were flattenable, extraction elevators, metro stations and air shafts would lose meaning. You need persistent geography so players can learn: this hatch needs a key, that metro is a death trap at minute 28, this ridge covers the elevator.

Embark throttled its signature tech to protect memory. Destruction still exists, but it is bounded so routes stay learnable across raids. In engine terms: they used UE5's replication and streaming to keep a shared mental map stable, not to show off physics. Less demolition, more dread.

That clutch-or-extract tension is hard to practice without risking loot. Block Contra: Clutch Strike captures a slice of it in the browser: quick competitive rounds where you must decide to hold an angle, swing, reload now or gamble the last bullets. Treat each round as an elevator decision — take the safe angle and live, or push for the highlight and lose it all.

What indie devs should actually steal

  1. Rent the labor market. If you have under 50 people, UE5's Nanite, Lumen and console support matter less than its hiring pool, tutorials, and outsourcer familiarity. Embark's 350 people still did not want to run an engine team. You do not either.

  2. Use density as gameplay, not wallpaper. Nanite lets small teams scatter 10,000 props. Only ARC Raiders makes 9,990 of them matter as cover, noise, or loot camouflage. If a prop does not change a decision, delete it.

  3. Charge to fix your servers. If your loop punishes death with real time loss, free-to-play multiplies support cost. A $40 box is a cheat filter and a patching budget. Do not copy Fortnite's free model for an extraction game without Fortnite's moderation army.

  4. Constrain your best tech. The Finals proved Embark can destroy everything. ARC Raiders proves they know when not to. For extraction, persistence beats spectacle. Lock your landmarks, then let players break everything else.

Limits and open questions

Sales and concurrency figures move fast and come from secondary summaries of Nexon and Embark announcements; treat 12 million and 960,000 as launch-window peaks, not permanent population. Patch notes through April 2026 focus on fixes and balance, but long-term Frozen Trail plans can slip. Performance on low-end PCs and controller UI readability remain the most cited costs of the UE5 density bet.

Still, the core lesson holds: ARC Raiders did not pick UE5 to look next-gen. It picked UE5 to escape its own past, ship extraction without running an engine company, and make 30 minutes of fear technically affordable. That is a rarer and more useful bet than prettier rocks.

Sources in text

Engine, release, loop and history from ARC Raiders and Embark pages; UE5 features from Unreal Engine 5 page; Frostbite history from Frostbite page; Finals destruction and player numbers from The Finals and Embark pages; browser trainers from EasyHub pages verified November 12, 2026.

Top comments (0)