The company Databricks has announced a new funding round that values it at 188 billion dollars (about 172 billion euros) - a figure larger than the annual gross product of several Balkan states combined. The exact sum wasn't disclosed, the money isn't in the account yet, and the round only closes somewhere at the end of summer. In other words: the valuation was announced before the money arrives. That's how big the trust - or the pressure - is on the artificial intelligence market.

This isn't a random jump. Over the past eighteen months Databricks has gone from round to round at a rare pace. In February it raised 5 billion dollars at a valuation of 134 billion. In September 2025 - a billion at 100 billion. And in December 2024 - a record 10 billion at 62 billion. So many funding rounds that people online are already joking about when „Series AA” will arrive, because the alphabet is running out.

The trick that opens the safes is as old as the AI fever itself: repackaging. Databricks was founded in 2013 as a company for storing and rapidly analysing huge volumes of data in the cloud - a classic enterprise software firm from the time before ChatGPT. Nothing about that is AI by origin. But access to the data of big companies put it in an ideal position to serve precisely those now chasing artificial intelligence with corporate-grade security standards.

And that's the point worth holding on to. Databricks isn't an AI lab - it has rebranded itself into an AI company. It rolled out a string of products with genuinely AI names - a database for AI agents, an AI portal, a „meta-platform” that manages multiple agents - and suddenly the legend changed. Yesterday's SaaS firm now collects billions as if it were part of a new era. The effect is so powerful that even the sandwich chain Jersey Mike's mentioned „artificial intelligence” a full 22 times in its stock-market filings.

There's also one detail that says more than the valuation itself. CEO Ali Ghodsi recently published an internal measurement built to control AI costs across his 3,000 engineers. The conclusion: open models, especially the Chinese GLM 5.2, are already handling even the hardest coding tasks - and doing it cheaper than the closed models from Anthropic and OpenAI. The company raising 172 billion euros on the AI wave is at the same time quietly admitting that cheap open alternatives are catching up to the expensive infrastructure. Who ends up paying that bill - the investor or the user?