The Index
Everything on this page is a public number. What is not public is the arithmetic between them.
The cross-connections
Each of these joins two figures that are published separately and never together.
How it is calculated
66 games a day on Steam plus 400 on itch.io equals 466 a day.How it is calculated
AI launches went from 13 to 530 a month, up 517. Non-AI went from 1,030 to 1,320, up 290.How it is calculated
23,958 games times an assumed 8 hours equals 191,664 hours. At two hours a day it would be 263 years.How it is calculated
300 of 19,112 titles. The rest, 98.4 percent, stayed below.How it is calculated
1,170 games in 14 days equals 84 a day, against 66 a day on Steam.How it is calculated
170,090 games a year against 121 million new repos.How it is calculated
Assuming 15 titles played a year against 170,090 new ones.How it is calculated
Illustration only: PC market 39.9 billion USD divided by 19,112 titles. In reality over 98 percent sit below it, because the distribution is extremely skewed.The raw numbers
| Metric | Value | as of | Source |
|---|---|---|---|
| Global games market | 205.0bn USD | 2026-12-31 | Newzoo Global Games Market Report |
| Players worldwide | 3.6bn | 2025-12-31 | Newzoo |
| Steam releases per year | 19,112 | 2025-12-31 | SteamDB |
| Steam releases per day | 66 | 2026-06-30 | SteamDB, eigene Ableitung |
| Steam releases with AI disclosure | 30.8 % | 2026-07-15 | Sulka Haro / PC Gamer |
| Titles above 1m USD | 300 | 2025-12-31 | Branchenauswertungen 2025 |
| Games on itch.io | 1.0m | 2026-07-31 | itch.io |
| New itch.io games per day | 400 | 2026-02-19 | itch.io, Thank You Itch |
| New GitHub repos per year | 121.0m | 2025-12-31 | GitHub Octoverse 2025 |
| Repos importing an LLM SDK | 1.1m | 2025-08-31 | GitHub Octoverse 2025 |
| Games in the first vibe coding jam | 1,170 | 2025-04-01 | Vibe Code Game Jam 2025, Rueckblick auf vibecoding.app |
Method and sources
Every raw figure carries a source and the date the source refers to, not the date we fetched it. Where sources disagree we keep both numbers and say why they differ, instead of quietly averaging them.
Rounding happens in the display, never in the data. The exact value stays one click away under How it is calculated.
Derived figures are marked as ours. Two of them rest on an explicit assumption: the pile uses eight hours of playtime per game, and the attention share assumes fifteen games played per year. Both are stated so you can disagree with the assumption rather than with the arithmetic.
One figure is deliberately labelled a warning. Average revenue per release is a real division, but almost nobody earns it, because the distribution is extremely skewed. It is shown to make that point, not to describe a typical game.