Data and compute

AI impacts gaming but not like you think it does

My four-year-old wore her Dutch soccer jersey last weekend. It’s from 2012, almost a decade before she was born, and initially belonged to my now 13-year-old. I know she’s almost too big for it, but she chose it without my prompting (she liked the color), and it somehow connects all the important dots in my world. I can’t think of a better reminder of the imminent World Cup.

Naturally, I have cleared my schedule to watch both the US and Dutch teams’ matches. Can’t miss it. And while watching a minimum of two matches a day will certainly have an opportunity cost, it’ll never be as expensive as actually going to the matches.

Officials in New York and New Jersey have launched an investigation into FIFA’s ticketing practices. I wrote previously about the enormous increase in the cost of attendance. Soccer, one of the more modest games enjoyed around the world, is pricing itself out.

The Dutch metaphor for this is that FIFA made soccer all about the marbles, not the game. It has adopted dynamic ticketing, a practice borrowed from concert venues that allows ticket prices to be adjusted according to demand, and it is creating a costly headache for fans. And that is in addition to the cost of getting to the stadiums. The most beautiful game in the world has never been closer and somehow feels further than ever.

FIFA can keep its $6,000 tickets. I’m watching it at home with the kids.

On to this week’s update.

---

BIG READ: Data and compute (Part 1 of 2)

Play often reflects the trends and tensions that take place in the larger world around it. We’ve already seen how tariffs and geopolitical conflict affect what the industry can produce. The biggest impact, however, comes from AI.

After Google’s Project Genie provided a free haircut for all major game companies’ share prices a few months ago, the jury is still out on the economic impact. Will AI make studios more efficient? (No, it won’t.) Will it make games more complex? (Also no.)

As AI marches toward becoming the center of the new economy, two macrotrends are making themselves felt: data and compute. This week, we’ll take a look at how data ownership will impact the games industry through the lens of Reddit’s recent policy changes. Following, in part two, we’ll take a closer look at the impact of AI on compute and device ownership.

All about Jasons

While the major AI companies are getting ready for their moment in the sun and filing for their inevitable IPO, the other half of this business is refining its strategies, too.

Specifically, once the dust settles from the reshaping of the planet’s skyline with data centers that provide access to computational power, the next infrastructural investments will focus on developing value-added services. That’s when access to a repository of unrefined materials will be key, and we’ll soon see a switch to a greater emphasis on data as the raw input.

Case in point, Reddit announced this week that it is changing its policies regarding data scraping. Specifically, it is “shutting down unauthenticated .json endpoints,” making it harder to download its data without permission.

While I’m a big advocate for the protection of users—your audience is the business—I’m also naturally wary when billion-dollar firms start wielding terms like ‘user protection’ because, frankly, that tends to be the number two reason.

First and foremost, social platforms and online communities have to resist the constant gravitational pull to monetize their user data. It is an important reason why I think Reddit is doing this, without having to say so. In its earnings report, it noted that it is “one of the internet’s largest corpuses of information with over two billion posts and 22 billion comments,” and that Reddit is “one of the most sourced and cited domains in AI.”

It may have sold some of its data a bit too soon.

According to its earnings report, Reddit generated an additional $100 million in 2024 after it started licensing its data to AI firms. (In its financial reporting, “Other Revenue” includes licensing plus income generated from Reddit Premium and its user economy.) Google reportedly pays Reddit about $60 million annually to use its data for training Gemini AI models. By 2025, this revenue line represented 6 percent of the total. That’s pocket change for perhaps the most-cited resource in AI, and precisely why Reddit is closing the gates.

Looking at its share price history, it is clear that things are going well, but that there’s also potential for more. Over the past 24 months, Reddit has had a wild ride, reaching nearly 5x its value, but ultimately trading 64 percent higher last week. A hefty infusion of revenue from its data feeds would go a long way toward sparking shareholder excitement.

It’s the kind of decision all of these firms have to make at some point or another. Meta, on one extreme, has already committed both its user and employee data to train its models. Discord, in contrast, has historically shied away from the practice, but since the departure of its founding CEO, Jason Citron, that may yet change, especially as it prepares to go public. Data, like regular oil, is a commodity tightly controlled for maximum value capture.

In turn, this creates a strategic problem for market research and intelligence firms. Access to audience data to reduce demand uncertainty and build more effective marketing models in cluttered digital entertainment markets is increasingly critical. (And, coincidentally, precisely the model we built for ALDORA, which measures the presence and performance of S&P 500 brands across the gaming eco-system.)

Just last week, I wrote about the data monopoly that has been materializing in mobile gaming. This is where the games industry is most vulnerable. The richest behavioral data in entertainment belongs to the platforms. Valve knows what every Steam user buys, refunds, and abandons; the app stores log every session; Roblox watches its players in real time. None of it belongs to the developers who generate it. The same logic walling off Reddit’s corpus applies to them in spades. Once your data is the raw material AI models are starving for, why keep handing it out for free? The day Valve or Apple decides to close or reprice that spigot, the analytics firms built on top of it, and the studios that lean on them to find an audience, find out they were renting, not owning. I have no doubt that Reddit is truthful when it says it’s changing the rules of engagement because of so many bad agents and automated aggregators that are up to no good. It would help, though, if there were some lawsuits or other evidence that made it more tangible than what management tells us.

The long view: Data will be both abundant and, therefore, inaccessible. We’ve mostly accepted that the record and output of our online activity no longer belong to us. After leveraging our sociographic information for almost two decades, social media firms have been targeting us with ads to the point that we can’t tell anymore whether anything we come across online is serendipitous or whether we are slowly moving through someone’s conversion funnel.

Making sense of all that information, however, requires centralized data depots capable of structuring and supplying specific data sets. Firms that excel at this will supply the proverbial ‘data is oil’ resource. Firms like Reddit provide a valuable service, certainly, and will increasingly represent aggregators that refine data so it can be licensed by others, specifically to make AI models smarter.

Data aggregators will become even more valuable, especially when the price of compute collapses (again). That’s a topic for next week.

---

NEWS

Valve raises prices on hardware and then some

“Due to rising memory and storage costs,” the existing Steam Decks are now substantially more expensive. The 512GB model will now cost $789 USD (up $240), and the 1TB model will cost $949 USD (up $300). It’s got gamers in a tizzy who now fear that the upcoming Steam Machine will be similarly expensive. I had previously updated the anticipated cost, but the extra 40 to 50 percent announced for the Steam Deck suggests the new hardware may launch closer to $800 USD on the low end, losing its competitive pricing versus the existing PlayStation 5 and Xbox.

And speaking of Valve, Bloomberg’s Cecilia D’Anastasio penned an excellent update on Valve’s monopoly moment. It is facing a class action suit that has produced some notable insights, including Valve employees arguing that consumers benefit from ecosystem policies, because it means users get the best available content for the same cost as anywhere else. I have a few thoughts on this, too, of course, but I will yield my time beyond pointing out that there really aren’t any meaningful competitors. Insisting on parity when you control the market means you’re preventing other platforms from competing.

Finally, Lord Gaben is apparently selling the Rocinante, one of his yachts. If any of you want to go in on it, hit me up. It’s currently docked at Canary Wharf in London, so let me know if you’re game, and I’ll ping the private jet so we can zip over there for a quick test run. (h/t Joost Rietveld)

---

PLAY/PASS

Play. I thought Lewis Packwood’s article on aging gamers was well-researched and timely. It offers an important strategic perspective (i.e., demographic changes) that goes beyond trying to guess the next hit game.

Play. The ESA issued its annual state-of-the-industry report.

---

NEXT UP

The growing momentum behind AI’s investment in compute is starting to take a material toll on gaming. Next week, we’ll take a look at how this may change how we play in the future.