Are the LLM Wars the Database Wars?

Back in 1994, the database was the hottest thing in tech. Oracle, Sybase, Informix and IBM slugged it out on benchmark charts, drew crowds at keynotes, and made careers out of picking the right platform. The pitch was simple and thrilling: dump every record your company owns into one engine and ask it anything you want.

Thirty-two years on, the database has won so completely that it vanished. It runs every app you touched today and nobody writes think pieces about it anymore. It's a line in a config file, chosen once, forgotten, remembered only when it breaks.

But here's the part that should needle anyone watching the current AI race: none of the loudest names from 1994 actually won. Sybase got bought by SAP and went quiet. Informix got bought by IBM and became a footnote. Oracle survives mostly as a recurring invoice. The engines that actually ended up everywhere weren't in that race at all. PostgreSQL, an open project rather than a product from any of those companies, became the default. SQLite ended up inside roughly a trillion phones, browsers and apps — and nobody sat in a room and chose it. It just showed up everywhere until it was infrastructure.

So if models follow that same arc, the launch event is the wrong thing to be watching. The interesting question isn't who wins the benchmark this quarter. It's which model quietly becomes the SQLite of this era, embedded everywhere, credited nowhere.

And to be clear, this isn't offered as a prediction. Important and exciting are two different things, and most technologies eventually slide from the second column into the first — the database did, electricity did. Whether large language models are partway through that same slide, or whether the comparison just doesn't hold, nobody actually knows yet.