Gaming · Sunday, 16 August 2026
01 · Briefing · what happened
Unity's game engine grew 2%. Its ad machine grew 69%
The best quarter in Unity's history came almost entirely from the part of the business that sells to players after they already own the game - and the rest of the week showed the same money moving the same way, from making games to running them.
2% vs 69%
Unity engine growth vs ad growth
same company, same quarter
$546m
Unity revenue for the quarter
up 24% on a year earlier
75
jobs at risk at Supermassive
third round of cuts in three years
$1.5bn
reported value of Alibaba's games arm
Trustar is said to be the likely buyer
At a glance
- Unity had its biggest quarter yet: $546m revenue, up 24%, with losses cut from $108m to $23m.
- The split is the story - the game engine grew 2%, the in-game advertising business grew 69%.
- Xsolla, Sega and Mattel all moved the same way, toward keeping players spending after the first purchase.
- Supermassive opened talks to cut up to 75 jobs, its third round in three years.
- Netflix shut Night School and Moonloot six weeks after praising Night School's last game on an earnings call.
- Amazon handed Lost Ark and Throne and Liberty back to their developers, leaving the big-multiplayer business it entered in 2016.
- Xbox union members planted one red flag for every developer laid off under the current boss.
- Marvel Tokon launched with almost no solo content - the game is other players, in casual and ranked queues.
Forces in play
Unity's ad business grew 69% while its engine grew 2%; Xsolla and Mattel are building for the same thing.
Supermassive is cutting up to 75 roles and Netflix closed two studios, one six weeks after shipping a game it had praised.
Amazon handed two big multiplayer games back to their makers and Epic sold ArtStation and Sketchfab to focus on Fortnite and its engine.
Trustar is reported close to a $1.5bn deal for Alibaba's games arm, and Tencent's revenue rose 11% to about $30bn.
Union members planted a flag for each laid-off Xbox developer outside Bethesda's Maryland office, with more rallies set for 18 August.
A shooter climbed Steam's charts after its makers listed ten reasons not to buy it, and a free eight-year solo project launched to real players.
How it unfolded
- Mon Sega reports 17.3% higher sales; Unity's quarter lands
- Tue Supermassive opens talks to cut up to 75 jobs
- Wed Union flags outside Bethesda; Amazon drops two multiplayer games; Epic sells ArtStation
- Thu Netflix closes Night School and Moonloot; Xsolla ships player-retention tools
- Fri Trustar reported near a $1.5bn deal for Alibaba's games arm
Where this points
Watch whether the next round of studio earnings repeats Unity's split - selling to existing players growing fast while the tools for making new games barely move.
Full briefing
The two numbers that explain the week
Unity had what its chief executive Matthew Bromberg called arguably its best quarter ever
Look at where the growth came from and the week snaps into focus. Unity has two halves. “Create” is the game engine, the software studios use to build things. “Grow” is the advertising network that places ads inside mobile games. Create grew 2%. Grow grew 69%, driven by developer spending on Vector AI, Unity’s ad-placement system
Bromberg described the business as a flywheel
That is the shape of the whole week. The money is moving out of making games people finish and into running systems people stay inside.
Everyone is building for what happens after the sale
Xsolla, which handles payments and storefronts for game studios, released a set of tools this week aimed squarely at that gap
Sega, reporting for the three months to the end of June, saw net sales rise 17.3% to 95 billion yen, about $600 million
Mattel formed Mattel Game Studios, nearly 300 staff across Los Angeles and Hangzhou
Who is paying for the shift
The same week, the people who make finite games kept losing their jobs.
Supermassive Games, the British studio behind Until Dawn and The Dark Pictures Anthology, opened a redundancy consultation that could cut up to 75 roles
Netflix shut Night School Studio in Los Angeles, the maker of Oxenfree, and Moonloot in Helsinki, which had not released a game
Amazon is handing back the two big always-online games it published. Throne and Liberty moves to NC by the end of the fourth quarter of 2026; Lost Ark goes to Smilegate in early 2027
Epic sold the artist portfolio site ArtStation and the 3D model marketplace Sketchfab to KitBash, five years after buying both
Then there was Bethesda’s office in Rockville, Maryland, on the day Xbox chief executive Asha Sharma visited. Union members from the Communications Workers of America put up a giant inflatable rat outside
Money is still going into new teams, but note what kind. 2K opened Small Axe Studios in Vancouver under Aaron McHardy, who spent nearly 18 years at EA as executive producer on Madden and College Football
The buyers have not left
Private equity firm Trustar is close to buying Alibaba’s video game business, Bloomberg reported, citing people with knowledge of the talks
Tencent reported second-quarter revenue of 204.8 billion yuan, about $30.36 billion
A new fighting game where the content is other people
Arc System Works and Sony released Marvel Tokon: Fighting Souls on PlayStation 5 and PC. It is a four-against-four tag fighter with a cast of 20, one shared health bar per team, and teammates you unlock during the match by filling a gauge
What is striking is the design decision underneath it. Simplified inputs let a beginner throw a finisher with one button
That is the design end of the same trade. A game whose content is other players never runs out of content, and never needs a sequel to keep earning.
Two smaller stories cut the other way. Ahead of its September early-access release, Bulkhead published a 15-minute video called “10 Reasons NOT to Buy” its $39.99 shooter Wardogs
Also this week
Twitch told users their streams, clips and chat will be used to train Amazon’s AI models unless they opt out
Apple proposed the fees it wants on purchases made through links out of the App Store
Saber Interactive will add an AI disclosure to the Steam page for Rideshare “Stimulator”
The far end of engineered play
Finish in Pennsylvania, where for 12 years machines called “skill games” have spread through gas stations and laundromats
The New York Times sat down at one in Allentown and played the memory game through. It took 9 minutes and 50 seconds and earned 3 cents
02 · Lesson · why it matters
Why getting better will not win you more games
A ranked ladder measures you, and it learns most from matches it cannot call - so it keeps finding opponents you can barely beat.
How it works
- A ranked ladder holds a number that guesses how good you are
- Every match is a test of that guess
- A blowout tells it almost nothing; a close match tells it most
- So it hunts for opponents you have about an even chance against
- You improve, your number rises, the opponents get harder
- Your win rate lands back near half
The twist
Getting better does not make you win more often - it makes you fight harder people, because the ladder is measuring you, not rewarding you.
Where you've seen this
School exams
graded on a curve, so the whole class improving moves nobody up
Job levels
the bar for the next band rises as more people clear the last one
Dating apps
who you are shown is chosen by an estimate of you that updates as you swipe
Credit scores
one number stands in for you and quietly decides what you get offered
The catch
A perfect measuring instrument is a permanent coin flip, which is exhausting to play - so studios deliberately hand out easier matches to keep people around, and buy retention with a worse measurement.
Full lesson
The queue is the game
Marvel Tokon launched this week with almost nothing to do alone. Five short story episodes, a party mode, a handful of arcade ladders. Everything else is other players: casual matches, ranked matches, open lobbies.
Arc System Works went out of its way to make that queue easy to enter. One control scheme for everybody, so a beginner and a veteran stand in the same line. No separate pool for people using simple inputs.
Then the ladder takes over. And the ladder is not there to give you a fair fight for its own sake. It is running a measurement.
A match is a question
A rating system holds a number for you. It is a guess about how strong you are, and it starts out wrong.
Every match is a test of that guess. The system predicts the result, sees what actually happened, and moves your number by the size of its mistake. Win when it expected you to lose, and it moves a lot. Win when it expected you to win, and it barely moves at all.
So the value of a match depends entirely on how uncertain it was. A game you were always going to win teaches the system nothing it did not already believe. A coin flip teaches it the most, because either outcome is genuinely news.
That is why the matchmaker hunts for opponents it thinks you have about an even chance against. Not to be kind. Because that is where the information is.
The ladder moves under your feet
Here is the part that surprises people. Improve, and your win rate does not go up.
You get better. You win a few more than expected. Your number rises. The matchmaker reads the higher number and starts finding you harder opponents. Within a few dozen matches you are back to winning about half.
The gains are real. They just show up in a different column. Your rank moved; your record did not. That is why “I have got much better and I still win half my games” is not a contradiction. It is also why the number beside your name tells you more about your progress than your win-loss record ever will.
The instrument and the product pull apart
A perfect measuring instrument would be a permanent coin flip. Every match a knife edge, forever. That is exhausting to play.
So studios trade. New players get softer matches for a while. Some ratings are hidden. Some queues are quietly loosened when the population thins out. Each of those buys people staying, and each of them costs measurement: a lopsided match tells the system less, so the number takes longer to find where you actually are.
Both dials are set by the people who built the game, and neither is announced. What arrives on your screen as “your skill level” is partly a setting somebody chose. The pressures behind that choice are ones you were not in the room for. How many people are online tonight. How long a queue can run before players quit. Whether this month is about keeping newcomers or serving the top. It is a real measurement of something. It is also a business decision.
The same machine, outside the game
Once you have seen it, the shape turns up everywhere a system sorts people by an estimate and then changes what it puts in front of them.
A class graded on a curve: everyone studies harder, the grades stay where they were. A promotion band at work: the bar for the next level rises as more people clear the last one. A dating app deciding who you get shown, from a number it keeps about you and updates as you swipe. A credit score standing in for a person and quietly deciding what they are offered.
In each one, the number is not a description of you sitting still. It is a running argument between what the system believed and what you keep doing.
What the number can and cannot hold
The estimate is honest about one thing and silent about the rest. It knows who beat whom. It does not know that you played the last hour tired, or on a bad connection, or against a team that had rehearsed together for a year.
And you are inside it, not above it. From your seat you see your own matches, which is the smallest possible sample of the thing being measured. The system sees millions and still calls itself an estimate.
Pennsylvania’s gambling machines are the same claim stretched until it snaps. They are called skill games because there is, technically, a skill in them. That skill takes ten minutes to earn three cents, so nobody ever uses it. A number can be a real measurement, a rough guess, and somebody’s convenient arrangement, all at once. Usually it is all three.
03 · Lab · your turn
Two Ladders, One Player
Train, play, and watch a ranked ladder keep your win rate near half while your rating climbs.
04 · Hope · carry this
A studio published ten reasons not to buy its own game, and players bought it anyway. Plain honesty is still the cheapest thing to offer, and people can still tell the difference.
More from Gaming