# Supabase is worth $10.5B because agents need boring databases Supabase just raised $500M at a $10.5 billion valuation — doubled in eight months — and the reason is almost funny: over 60% of the new databases on its platform are now created by an AI tool, not a human. The flashy part of the AI boom is the agents writing the code. The part that's quietly minting money is the boring, reliable place that code has to put its data. That's not a coincidence — it's the whole lesson about where durable value lives. - **Author:** Fedor Chishchin - **Published:** 2026-06-06 - **Tags:** business, architecture, ai-native - **Source:** https://fedorthinks.com/blog/agents-need-boring-databases --- Here's a number that should make you stop scrolling. Supabase — a company that sells Postgres databases, possibly the least glamorous product in tech — just [raised $500 million at a $10.5 billion valuation](https://www.cnbc.com/2026/06/04/database-startup-supabase-raises-500-million-10point5-billion-valuation.html). That's roughly double what it was worth eight months ago. The round was led by GIC, with Stripe and Salesforce Ventures piling in. Why is a database company suddenly worth as much as a mid-size bank? The [reason](https://techcrunch.com/2026/06/05/supabase-doubles-valuation-to-10b-in-8-months/) is the genuinely wild part: new database launches on the platform are up 600% in a year, and **over 60% of them are now created by an AI tool, not a person.** CEO Paul Copplestone credits Claude Code and Codex directly, because those tools ["expand the number of people who can build."](https://www.cnbc.com/2026/06/04/database-startup-supabase-raises-500-million-10point5-billion-valuation.html) Agents are out there spinning up real databases by the millions. Sit with what that means. The headline story of this year is AI writing software. The business getting rich off it is the one selling the boring thing that software needs underneath. That's not an accident. It's a pattern as old as commerce, and it tells you exactly where the durable money is. ## In a gold rush, sell shovels The cliché exists because it's true: the people who reliably got rich in a gold rush weren't the miners — it was the ones selling shovels, jeans, and supplies to everyone digging. Most miners went home broke. The shop on the corner sold to all of them, winners and losers alike. The vibe-coding boom is a gold rush, and Supabase is selling shovels. Most of the apps being spun up by AI tools right now will go nowhere — abandoned prototypes, weekend toys, half-finished SaaS. Doesn't matter. **Every single one of them needs a database the moment it wants to remember anything.** Supabase gets paid whether the app becomes a business or dies in a week. It's positioned underneath the activity, not betting on any one outcome of it. That's the cleanest possible position in a frenzy: you don't have to pick the winner, because you supply all the contestants. ## Agents make the boring layer _more_ important, not less There's a deeper reason this is happening, and it's the part builders should internalize. An AI agent that writes your app is impressive, but it is, by itself, forgetful. The model doesn't hold your data. It plans, it writes, it acts — and then the actual state, the source of truth, has to live _somewhere reliable_ outside the model. Strip away the magic and an agent is mostly a very fast thing that reads from and writes to a database. I've argued before that [stateless agents die](/blog/stateless-agents-die) and that everything good rests on [a deterministic source of truth the model sits on top of](/blog/grounding-is-a-constraint). Supabase's valuation is that argument with a dollar sign attached. The smarter and more autonomous agents get, the _more_ they lean on a dependable place to keep state — not less. The intelligence layer is exploding and commoditizing at the same time; the storage-and-truth layer underneath just keeps getting more load-bearing. Industry write-ups put it plainly: agentic frameworks are [only as powerful as the data infrastructure under them](https://www.instaclustr.com/education/agentic-ai/agentic-ai-frameworks-top-10-options-in-2026/). ## Why "boring" is the feature, not the bug Notice what Supabase did _not_ win on. It didn't ship a clever model. It didn't invent a new paradigm. It offered plain, well-run Postgres — a database that has been quietly trusted for thirty years — and made it easy to point an agent at. The boringness is the entire value. When the layer above you is changing every month, the thing people want underneath is something that does not surprise them. This is exactly why I keep [boring tech where it can be boring](/blog/the-model-was-never-the-moat): the novelty belongs in the AI layer, and the substrate should be predictable. The market just paid ten billion dollars to confirm it. Predictable infrastructure isn't the unsexy afterthought to the AI story — it's the part that captures the value while the exciting layer races to zero. ## The takeaway for the rest of us You're probably not about to start a database company. But the lesson scales all the way down to a single product. When everyone around you is chasing the glamorous, fast-moving layer — the newest model, the slickest agent — ask a quieter question: what does all this activity _depend on_ that doesn't change, and do I own a piece of it? The agents are the gold rush. The thing they all have to write their data into is the shovel. Supabase just got a ten-billion-dollar reminder that, in a frenzy, the boring layer everyone needs is worth more than any single exciting thing built on top of it. Build accordingly — own a shovel, not a lottery ticket.