Embark Studios’ game has been a success despite not having one of those massive AAA budgets.
Arc Raiders has been one of the biggest hits lately, and the more we hear about it, the better its story seems. Developers who previously worked at Electronic Arts and DICE joined forces to build the game they wanted to make. They didn’t rush development, and they had the patience to iterate, rework, and polish, even while finishing another project before Arc Raiders was ready for release in October.
After it emerged that Embark tested PvE, PvPvE, then PvE concepts during Arc Raiders’ testing, the studio’s founder and CEO, Patrick Söderlund, shared more in an interview with Gamesindustry about how Embark Studios ended up with one of 2025’s best-selling games. He also explained how the studio pulled it off on a fraction of the budget compared to other major AAA titles with similarly sized player bases and sales figures, but far larger development teams.
“It has nothing to do with how many you are. That’s not the point. Our ambition is to produce products with quality and depth similar to those of other teams and studios, but with a much smaller team. We’ve been part of this since the beginning, when seven or eight people could make a game. We were part of this spiraling exercise involving people, cost, and time. Now, with the available tools, pipelines, technology, and a different approach to building and structuring ourselves, we hypothesized that we could do this faster and better than before. We quickly realized that we could and that we could do much better. We can compete in the AAA space with maybe a quarter of the budget,” said Söderlund.
Söderlund isn’t referring to generative AI when he talks about the tools available today – at least not entirely. Embark’s use of generative AI is well known, primarily around NPC work in both Arc Raiders and The Finals, but Söderlund is explicit that it plays only a minor role in the studio’s output. He also addressed how the studio uses generative AI for voice work.
“We asked what other technologies are available to us. Could we use the topography from Google Maps? We use photogrammetry: we take photos of objects and use them to texture assets. Can we create a realistic landscape using procedural generation and pipelines? Very little of it is AI. Much of it involves reconfiguring what I consider to be outdated methods – outdated toolsets, pipelines, and engines – and recognizing that there must be a more efficient approach. We pay our actors for all the time they spend with us in the booth, and we continue to bring many of them back as we update the game. For select usage, we also pay them for approval to license their voices through text-to-speech for less essential lines, mostly ping system audio.
We re-recorded some lines after launching the game and used real voices for them. There is a quality difference. A professional actor’s performance is superior to AI, plain and simple. We primarily view AI as a production tool. We can test things internally. For example, we can test 15 different lines without recording them first, so we know what to record. It’s also a way for us to work alongside actors, not replace them. We don’t necessarily believe in replacing humans with AI all the time,” Söderlund added.
The question is whether they might take it too far at some point.
Source: WCCFTech, Gamesindustry
