Predictions for AI in the Game Industry 2025 | AI and Games Newsletter 15/01/25
With the new year upon us, what do we think will cross our path?
The AI and Games Newsletter brings concise and informative discussion on artificial intelligence for video games each and every week. Plus summarising all of our content released across various channels, like our YouTube videos and in-person events like the AI and Games Conference.
You can subscribe to and support AI and Games on Substack, with weekly editions appearing in your inbox. The newsletter is also shared with our audience on LinkedIn. If you’d like to sponsor this newsletter, and get your name in front of our audience of over 6000 readers, please visit our sponsorship enquiries page.
Hello all, and welcome back to the first edition of the
newsletter for 2025. We’re a little late in getting back into the swing of things, and that’s largely because I was enjoying my time off! Remembering what my job is and how to do it took pretty much all of last week, and then some of this week too! Behind the scenes there’s a lot of stuff going on in and around all things AI and Games, so I figured it best to keep you waiting an extra week until we started writing the newsletter in earnest. I hope you all had a great holiday season (or non-holiday season) and are ready to face this new Gregorian age with sufficient caffeination!Already we’ve got news stories to discuss, and some announcements for current and upcoming projects and events. But I figured lets spend this inaugural issue of 2025 looking at the year ahead, and how I think things will pan out.
At the top of article, I’ve actually narrated the main story that is available below. Let me know if that was something you find useful or interesting!
Alrighty, let’s do this.
Follow AI and Games on: BlueSky | YouTube | LinkedIn | TikTok | Twitter | Threads
Announcements
Some quick announcements of upcoming events, but before we do that, a quick update on our content across all things AI and Games.
AI and Games for 2025
As discussed in the Wrapped 2024 livestream last month, and in even more detail in last month’s sponsor newsletter, the plan for 2025 is for AI and Games to focus on four key pillars: YouTube, Substack, Conference, and Courses (Goal State).
We’re in a fantastic position to start the year, with planning for the 2025 conference already underway, and Goal State being funded (but you can still pledge late right now for a limited time). So right now I’m focussed very much on writing. I have a lot of writing to do this year. Plus we have a lot of videos in the pipeline. With over 20 recordings from the 2024 conference to release, plus around 5 case studies in the edit right now. All of this on top of my public speaking engagements, and my actual work with games studios. Plenty to be getting on with, and we’re in a good position financially for me to just get on with this. That is, in large part, thanks to all of you out there who have supported one, several, or all of my endeavours this past year or so. Thank you all once again.
But as a result of this focus, we are making some smaller changes to the pipeline that ultimately don’t affect this newsletter all that much, but does shift a few things in me in the background:
AI and Games Plus videos will be less frequent, and the focus will be on vlogs, talk recordings, and tie-in material to the main channel. Originally it was providing videos that were the precursor to what is now the newsletter. But frankly we don’t have the time to edit up a full video every fortnight alongside writing the newsletter themselves.
The Branching Factor podcast will wind down for now as we focus on other content. It’s been fun, and there’s a good chance it will return in some form, but for now we’re letting it taking a break.
Other Announcements
I was invited to be a guest on The Artificial Human podcast for BBC Radio 4 late last year. I sat down with hosts Aleks Krotowski and Kevin Fong to discuss the current state of AI adoption within the video games industry. Critically, I haven’t heard the final version as it goes live today, January 15th, over on the BBC site. Fingers crossed it turned out pretty well!
I’ll be participating in a panel at the 2025 Guildford Games Festival on Friday February 14th (how romantic). Very excited to be attending this event, with Guildford being one of the most prolific regions of the UK for game development. More details on this as it comes!
I’m now opening up bookings for meetings at GDC 2025! We’re only two months out, and so if you’re keen to meet to discuss working together, our conference, Goal State training, or just for a blether over coffee, hit the button below!
AI (and Games) in the News
While I took a break, AI in the news most certainly did not! Oh boy. Let’s take a look.
UK Government’s Open Consultation on AI and Copyright
As mentioned briefly at the end of last year, the UK government opened up a consultation on ‘Copyright and Artificial Intelligence’ a week before Christmas. This consultation, which closes in late February, is focussed on three key elements. To quote the website:
Boosting trust and transparency between sectors, by ensuring AI developers provide right holders with greater clarity about how they use their material.
Enhancing right holders’ control over whether or not their works are used to train AI models, and their ability to be paid for its use where they so wish.
Ensuring AI developers have access to high-quality material to train leading AI models in the UK and support innovation across the UK AI sector.
One of the suggestions proposed in this consultation, is essentially an opt-out system where by default any copyrightable material produced in the UK is by default permitted to be used in AI training. Essentially meaning that anything published in the UK’s creative industries is immediately okay for generative AI companies to exploit in the pursuit of their own model training.
As you can imagine, this has not been received well at all. I was so happy to end my year with all of my social media platforms filled with people annoyed at this. Thanks UK government! I really wanted just one week without worrying about another AI-laden nightmare for the games industry.
Over on Forbes, Virginie Berger wrote a short but succinct piece breaking down the inherent flaws of the proposal and the issues this could lead towards in the UK creative industries. Perhaps critically, that opt-out approaches are bullshit, and a lazy approach to proper regulation of these issues.
This situation has since been exacerbated by the publication of an independent AI Opportunities Action Plan that was commissioned by the UK Government, that suggests the results of this consultation may well be ignored.
As mentioned, the consultation is open until February 25th, and I will be addressing all of this in much more detail later this month.
Music Lyric Protections for Large Language Models
The music industry is no stranger to the wave of generative AI systems out there. With everything from generative audio tools, to the fake AI musicians on Spotify, for much like gaming for every possibly useful application, there is some egregious nonsense that is disrupting the sector in far from positive way.
But of course one area that is more readily exploited is song lyrics, given they are of course just words and there’s plenty of work being scraped and stolen online in the pursuit of training large-scale LLMs like GPT. Song lyrics, like books, and a variety of other text-driven artifacts, are protecting by copyright and licensing. You can’t simply copy and past the lyrics from one song into another without making an agreement (even most legit websites that host song lyrics do so under license). So what happens when an LLM starts spitting out song lyrics?
As detailed in ArsTechnica, AI company Anthropic - creators of the Claude LLM - have ultimately lost a court battle against a number of music labels who had raised concerns about the how Claude was generating song lyrics without consent. The original lawsuit was to stop Anthropic for training their models against this copyrighted material, and while that appears to be an ongoing battle, for now they have forced Anthropic into an agreement to ensure they have sufficient safeguards in place to prevent those song lyrics appearing in its outputs.
NVIDIA Announces New RTX 50-Series GPUs, Plus More AI Rendering Tech
At CES last week, NVIDIA announced a whole suite of new software and hardware that re-affirms their commitment to AI. Including Projects Digits, the most powerful, small-scale, AI supercomputer in the world (and yours for only $3000), and a new suite of RTX AI PCs to optimise specific generative AI workflows. More work in their exploration of autonomous vehicle technology and human robotics. But for games the big things were the announcement of new RTX-50 series GPUs, as well as their new AI rendering technologies, including an updated version of DLSS, neural texture compression technologies (which make in-game textures less expensive to store/render), and generative AI based hair and face rendering tech (the latter of which we’ll talk about later).
PCGamer has done a good job of breaking down the technical specifications of the new RTX-50 series. These are impressive, to the point that the base 50 series GPU is now as powerful, if not more so, than the top of the range GPUs from the 40 series. But all of this comes at a cost. Sure, it’s cheaper than many expected, with the RTX 5080 coming in at $999, which is cheaper than the RTX 4080 was at launch by $200. But still to get the most out of this hardware you’re looking $1900 for the RTX-5090. This is of course a great boost for people who work in AI and want to use these GPUs for research and development. But yeah, as a ‘gamer’, it’s not as straightforward a bargain.
Predictions for 2025
With the year only having just started, I figured why not take the time to list some of my expectations for how things will go down in 2025. If anything, this will make for some interesting content later in the year, as I reflect on these predictions and how accurate they were.
Sadly, for reasons that will be obvious in a moment, several of my predictions were proven accurate as I was writing the issue! I’m not going to shy away from that or hide it. But, rather let’s just put it front and centre and acknowledge that some readers may think I stole the headlines when in truth it was sadly the vicissitudinous of our beloved sector.
A Make or Break Year for AAA Game Development
The past couple of years has been a precarious space for big budget videogames for a number of reasons. It’s been discussed several times here and elsewhere that large-scale AAA games and their studios are struggling to handling a volatile market, with increasing more fickle consumers, at a time where costs are escalating both courtesy of economic factors and the expectations they have set for themselves.
Last year was a relatively quiet year, but it was also one with several high-profile failures or missteps. Many a big budget title came and went without much fanfare. Titles such as Star Wars: Outlaws, and Dragon Age: The Veilguard were criticised for lacking in innovation, while Suicide Squad: Kill the Justice League and especially Concord suffered from being late entries to largely saturated markets. Meanwhile even beloved, award-winning titles such as AstroBot, Metaphor: ReFantazio, Black Myth: Wukong and Final Fantasy VII Rebirth are not the sales juggernauts people would expect - with none of these games breaking the top 10 best selling games of 2024 in the US (and the UK). All-in-all, releasing a big budget AAA title seems like a riskier proposition now in games than it ever has.
And so it’s exciting, and concerning to see that we’re going to have one of the most bumper February’s in gaming history. March will see the delayed Assassin’s Creed Shadows get released, while I’d argue Capcom’s Monster Hunter Wilds is the big game of February. That’s already two massive titles. But alongside Monster Hunter we can expect Kingdom Come: Deliverance 2, Avowed, Civilisation 7, and the latest (albeit spin-off) entry of the increasingly popular Like A Dragon series - Pirate Yakuza in Hawaii. In amongst all this between now and March we have a slate of smaller fare, plus a whole bunch of PC ports of games like Final Fantasy VII Rebirth, and Marvel’s Spider-Man 2.
Then pushing it even farther out, we have a lot of big games still dated for 2025. Off the top of my head that includes Borderlands 4, Metroid Prime 4, DOOM: The Dark Ages, Death Stranding 2, Pokémon Legends: Z-A, Ghost of Yotei, Fable, Elden Ring: NightReign, and of course a little game called Grand Theft Auto VI. Plus, we know the new Nintendo Switch console is coming to market in the 2025 fiscal year, and that will no doubt bring several high-profile first-party titles for Nintendo.
At this point there’s a reason a lot of the games mentioned above aren’t dated, and that’s because they’re worried about the impact of both Nintendo’s new hardware drop, and perhaps more critically whether Rockstar keep their original 2025 release date. At a time where AAA games are struggling to find audiences (largely because they’re sticking to older free-to-play games), we have an incredibly busy release schedule. And that’s before I mention any annual releases like EA Sports FC or Call of Duty.
I think we’re going to see more high-profile ‘failures’ or lacklustre responses, and that in turn is going to have a negative affect on studios (see below), alongside the argument by many executives of the need for AI (because it will somehow fix all of these problems).
Funnily enough, Chris Tapsell over at Eurogamer had a similar thought.
Layoffs will Continue Into 2026
While I’d argue that the economics surrounding games are improving, we’re still a long way off from it being close to safe, or stable. My previous point regarding a big year for AAA means that some of these companies are going to struggle as their games once again fall flat; either through market saturation, outdated ideas finally coming to market, or simply that there are just too many games that will need to come to start making money for their respective studios and publishers.
Sadly we’ve already seen this happen in the first week or so of 2025, with two of the UK’s biggest developers, Rocksteady (creators of the ill-fated Suicide Squad) and Splash Damage - who cancelled their project based on the Transformers IP. Best of luck to everyone at those studios. That’s a rough start to the year.
I think we still have a handful of big companies that are looking especially risky. My eyes are on Ubisoft, Square Enix and pretty much everything the Embracer group has their hands on. But perhaps the most disappointing thing that will happen, is while AAA layoffs will dominate the conversation, more and more indie studios will quietly shutter as they struggle find the financing, or the sales, needed to survive.
NVIDIA’s Push Into Full-Fledged AI Company
On my first day back to work, January 6th, I wrote this header. My original argument was that NVIDIA has been slowly migrating away from gaming as its core means through which to build its revenue and overall company value. We saw this years ago with their GPU gradually migrating from being a traditional device for rendering, and instead being built to more readily enable for things such as AI and Blockchain. Over the past two years we’ve seen NVIDIA continue to push further into both AI hardware, with the likes of their RTX AI workstations, as well as cloud and edge computing solutions. Plus of course, software tools and platforms such as ACE, Omniverse for 3D workflows, and even things like Clara which is their AI-powered healthcare systems.
Ultimately, it was a fairly low-stakes bet to suggest that while NVIDIA maintains some involvement and intersection with game development, this was more as a result of it being a viable application area for their interests in AI, rather than it being the core of their business. Something that is evident by how their stock price has tripled in the past 12 months courtesy of the AI boom.
Well, we can scratch this one off as proven accurate already. As NVIDIA showed off a range of AI applications at CES 2025 last week as mentioned in the news segment above. At this point games, and more specifically graphics rendering, is now seen by NVIDIA as an AI problem. I mean NVIDIA CEO Jensen Huang literally said as such in a conference talk last year. Which given the complexities of the problem, and the fairly pervasive adoption of upscaling technologies (many of which use AI), it’s increasingly harder to argue with that assessment.
More AI Demos, More Start-ups, More Local Inferencing
In my penultimate newsletter issue of 2024, I summarised the state of AI in the games industry as a stand-off, in which everyone is talking about it, but nobody is committing anything substantial. I do think this will largely continue into the new year, given of course that AAA game development takes so long, we’re still going to be waiting a couple of years before big budget games emerge truly commit to generative AI technologies as a core component of the experience. 2026 at the absolute earliest I think, though I think 2027 is a more realistic launch date.
But in the meantime we’re going to see demos aplenty! More and more demos of potential technologies. You’ll see more news headlines of AI start-ups still emerging that seek to find ways to adopt (generative) AI in games. I have mentioned in the newsletter and in talks last year we’re now entering the second wave of generative AI as things will begin to cool off. The hype is being seen through more readily by savvy folks in business fields, and will start to identify more relevant and applicable technologies.
This is also in part because the technology is getting better. Rushing to adopt any new AI tech out the gate always strikes me as naïve as it’s getting faster and more efficient over time. Perhaps the one area I expect to see the most this year is demos that carry local inferencing. What does that mean? Well, rather than having to rely on a server hosted LLM or similar generative model (e.g. like plugging a game into something like GPT), you’ll start to see more and more projects where they’ve reduced the model down to something that can run locally without huge implications for the end user. We saw this already late last year with AI People now permitting local execution, and I expect this trend to continue.
Why? Well it isn’t just that online costs are prohibitive. Ultimately, games where a crucial gameplay component is always online, but the game itself isn’t an online multiplayer, will always come under scrutiny. It means a game that uses an LLM being played on mobile devices is useless if you have a poor signal. Your NPCs become braindead because you’re in a tunnel.
Like any new technology, you have to remove as many limitations or impediments as possible. It’s why the Quest 2 is the most successful VR headset of all time - it’s got decent power, cable-free tracking, and doesn’t cost a stupid amount of money. The more you can prove that these technologies don’t have all these additional trappings, the more convincing it becomes to game developers - who let me remind you, are incredibly risk-averse for a number of reasons.
AI-Native Games Will Start to Ship, But Mostly From Indie Studios
I often hear that the AI-powered future of games is right around the corner, and while I won’t argue that AI tooling exists and is being used, the idea of games that utilise contemporary AI techniques is not going to emerge across AAA games in 2025.
A big reasons for that is many of the AAA games coming out this year - and as mentioned, there are a lot - all would have started production anywhere from 3-7 years ago, when most generative AI was nowhere near as ‘competent’ as it is now (I use that term generously). So the idea of seeing what are referred to as ‘AI Native’ games, i.e. titles that utilise modern AI approaches as core parts of the gameplay experience, is not something I’m expecting at this point. At best, expect more of what we saw in 2024 with games like EA Sports College Football 25, FoamStars (ugh), or Call of Duty: Black Ops VI where it’s used for incidental art or as a tool for user generated content.
However, I expect to see a lot more AI-native indie games come to market. Games that don’t just try to use generative art as a means to make a quick buck - after all we’ve sadly seen them crop up already. Rather, these games use generative AI as a critical part of the experience. We already mentioned Retail Mage launched late last year, and I’ve discussed games such as 1001 Nights and Dead Meat in previous issues as well. And while there have been other examples (see Suck Up!), these games are launching on Steam like traditional titles.
This is, both fortunately and unfortunately for these developers, a true litmus test for the viability of these technologies in building and launching an AI-Native game. So naturally I’m keeping an eye out for all of these games as they start to emerge. On that note, I will come back Retail Mage later in the year as I’d like to both play the game and evaluate how well it turned out.
Someone, Somewhere, Is Going to Put Their Foot in It
So while of course I don’t expect AI-Native AAA titles, I do expect another AAA studio to throw in generative AI as part of the game and its going to blow up in their face in spectacular fashion. I’m not going to hazard a guess as yet as to how it will blow up, but here’s some immediate thoughts:
A new generative AI tool will be added to a game that is a really poor fit for the technology, and the community will campaign for its removal.
A tool will drop in a game and people will immediately abuse it to levels beyond what the developers anticipated.
I mean to some extent we’re already seeing some backlash. NVIDIA’s ‘Neural Face Rendering’ technology that was announced at CES has gone down terribly on social media among game developers.
I’ve put a copy of the video in the newsletter above, rather than link to NVIDIA’s hosted version on YouTube. Why? Because the original video was taken down, and later re-uploaded with comments disabled. So I can’t trust it will stay online at this point. This demo has been lambasted for a number of issues, namely:
The neural face appears to have been ‘beautified’, in a way that is evocative of the much derided ‘Hire Fans (lol)’ social media post from 2021 that later became a meme.
There is an element of whitewashing as the neural face begins to exhibit features not reflective of the ethnicity of the original character.
It robs the character of looking the way that the character artist originally intended.
Frankly, it’s horribly inconsistent and doesn’t sit on her face correctly. It’s just not that good.
This is just the start. I am now waiting to see what emerges at the likes of GDC with great interest…
Platform Holders Will Begin to Dictate Policy
While AI regulation is continuing to take shape across the EU, the UK (ugh, see earlier) and the US still have little of substance behind it. As such, there’s very little clear guidance on what is deemed acceptable and viable applications of a number of generative AI technologies.
Valve have previously made steps to clarify their process on Steam. But I think this is the year that the likes of Sony, Microsoft, and Nintendo will have to start more formally declaring what they consider acceptable. While I suspect Nintendo will continue to be the strictest, even if they’ve been bad at catching stuff as it is submitted for lotcheck, the one I have my eye on is Microsoft.
Xbox is in a bit of a pickle with this one. Given Microsoft is all-in on generative AI across Windows, Office, Teams, and other products. They’ve heavily invested the likes of OpenAI, and even have struck up collaborations with companies specialising in generative AI for games like Inworld. So I suspect there may even be some internal pressures to start permitting and supporting generative AI games on their platforms, all the while knowing that’s probably going to be a net-negative for their ecosystem at this time.
Wrapping Up
A relatively safe return this week as I wanted to ease my way back into the swing of things - even though this still took me twice as long as usual to write. Thanks for reading, and welcome (back) to the AI and Games newsletter!
Next week will be our monthly sponsor update, but we’ll keep you up to speed on the news as it happens. Then we'll be discussing the UK’s AI plans and copyright consultation in the closing issue of the month! Catch you next week!