World Cup 2026Planet in Play
What you're seeing
Every hour, 90+ million rows are written to Postgres.
Millions of data points a second. Bets, touches, passes, saves, goals, live traffic. Every one gets saved into a single database at the stadium, then copied out to fans on every continent. The globe shows that traffic as a live heatmap: host countries light up the moment a play is recorded, and the rest of the planet glows as billions follow along.
One match, in real time, is a wall of demand. Scroll to watch a single game spike ↓
Round of 16 · July 4–6, 2026
This weekend's knockout games — every goal at its real minute
Under the hood · one Postgres database
Every action on those pitches is a row written to Postgres.
A pass, a shot, a save, a goal, a bet. Each one becomes a single write, saved once at the stadium's Postgres database and copied out to servers around the world in milliseconds. When every game reacts to the same goal at once, Neon adds compute to handle the surge, then drops back to zero when the action stops so it never pays for capacity it isn't using. Five games, one database.
THE STORY BEHIND THE DATABASE
130 years of database history · one knockout draw · tap a year
Flat files and the filing cabinet era
A man in Kansas City kicks a ball into a net. A number goes from 0 to 1. Two billion people on six continents have to agree on that number inside the same second - and none of them will ever think about the machine that made them agree. That's the whole job. It took 130 years. Nobody was supposed to notice the database, and if we did ours right, you won't.
The World Cup's oldest problem is also computing's: count a lot of things, fast, before a deadline.
In 1890 the US Census did it with Herman Hollerith's punch-card tabulator - eight years of manual counting compressed into three months. One card per person, facts punched into fixed columns. A squad sheet is the same object with cleats.
The 1950s inherited the model and the trap. Records sat in flat files, one after another, in a layout only the program that wrote them understood. Want shirt number 10? Read one through nine first, every time. Want to start tracking goals? Every program that touched the file hard-coded the exact byte each field started at - so adding a column didn't extend the file, it detonated it.
Data and code were poured together like concrete.
That's the original sin. Every page after this is somebody trying to pour them separately - which, 130 years later, is the entire pitch of a serverless database.
↗ Hollerith tabulator - US Census Bureau (1890)