Even if it’s a decades-old industry, video games remain an incredibly young business that stands at the forefront of new technologies including, obviously, AI.
Since the beginning of the year, AI has been all about multimodal models and notably text-to-video, with OpenAI revealing its generative model Sora. But Google pushed the envelope of what’s possible, with its DeepMind unit showing in recent weeks its text-to-video games model (pictured above), called Genie (for “GENerative Interactive Environment”).
Genie, which was trained on 30,000 hours of video of hundreds of 2D platform games, is indeed able to build playable games in the style of classic 2D platformers like Super Mario based on a prompt that can take the form of a short description, a hand-drawn sketch, or a photo.
As one would expect, the technology is still early stage and has some limitations. Notably, the games built by Genie are slow, running at one frame per second vs. 20-30fps for basic playable games and 60-120fps for current PC/console games. That said, as AI inference speed is expected to improve materially, game speeds around 30fps should be within reach over the next couple of years.
The implications for the gaming/Metaverse industry are massive. Generative AI indeed takes assistive technology to a new level, materially reducing development times not only for gaming assets creation but also for music generation and animation, hence materially speeding up the creation of games and virtual worlds. It is also a game changer in terms of development costs as it is expected to make creative and development teams much more efficient by letting them focus on complex tasks (design of key avatars/settings, scenario, virtual world mechanics…) and get rid of basic artwork and time-consuming processes (simple code snippets, software testing and documentation…).
While we are still in the early days of generative AI, some software developers estimate that they can be 25% more productive when AI automates manual processes such as code documentation and testing. Even taking a more cautious assumption (10% productivity gain), we estimate that the impact for companies building video games and virtual worlds would stand at least at +300/400bps on operating margins, pointing to a 10-15% EPS upside potential.
While we see margin upside for many players and notably the biggest game publishers (e.g. EA, Take-Two, Tencent…), we also see some downside risks for other players, notably mobile and casual game publishers (e.g. Playtika), as AI is expected to significantly lower the barriers to entry in mobile games and as anyone will soon have the ability to develop a casual game with the help of AI tools.
Talking about AI tools, design software firms including Unity and privately-held Epic have been fast to integrate AI technology in their software products with the stated objective to speed up professional developers’ work and/or to bolster user-generated content, notably from users with no- or low-coding skills.
While these initiatives should spark some topline upside in the near term as game developers demand more and more AI tools, we are concerned that the competitive environment of these design software firms might intensify in the future as Google DeepMind has made no secret that Genie, that is an in-house research project for now, could one day be turned into a game-making tool. At some point, it’s also likely that some of the most powerful AI models will provide similar capabilities and will compete head to head with design software firms.
In conclusion, even if AI is likely to bring disruption to some gaming segments (mobile, software tools), its impact on large studios that have a competitive moat (= franchise strength) is expected to be largely positive at the margin/EPS levels. Virtual worlds such as Roblox should also benefit as AI should give a boost to platform content (and then user engagement/traffic) by opening content creation to non-professional users.
