The discipline behind turbopuffer's $100M ARR and 50-person team

The standard assumption in infrastructure software is that scale requires capital, and capital requires a growth story that satisfies investors. turbopuffer is a clean counter-example, and the story of how they built it is worth reading carefully.

The company powers search infrastructure for many AI companies, has over $100 million in annualized run rate, and employs about 50 people. They raised $700,000 initially — at a conservative $14 million valuation — and didn't raise again for two and a half years, and then only to fund a tender offer for employees. Founders Simon Hørup Eskildsen and Justine Li came from Shopify, started in Justine's living room in 2023, got Cursor as their first customer by promising a 95% cost reduction, and built from there.

That's the setup. Here's what's actually worth learning from it.

Hiring as a discipline, not a strategy

turbopuffer waits until the pain is undeniable before hiring. No pre-staffing for anticipated growth. Justine Li puts it simply: "We are okay with pain, but we are not okay with lowering the bar."

That's a real trade-off. Features ship later. Some customers wait longer than they'd like. But the upside is a team where everyone has genuine leverage and where the people you hire are solving real problems rather than ones you invented to keep them busy. The instinct when things are going well is to hire ahead of need. turbopuffer's experience suggests that instinct is worth resisting. The short-term discomfort of staying lean buys you something that's very hard to recover once lost: a high bar and a coherent culture.

Customers as the real accountability

turbopuffer's customer orientation looks almost like a quirk on the surface — Simon canceling an afternoon of meetings to help a customer with a Postgres incident, Justine tracking down a private fiber-optic connection to shave seven milliseconds of latency off a customer's requests. But there's a structural explanation. When customers pay your salaries instead of venture capital, you can't afford to treat them as vanity metrics.

This changes how you build. You're not optimizing for demo-ability or a growth chart that impresses a board. You're optimizing for whether the thing works for the person paying you. It's a simpler and more honest feedback loop, and it produces a different kind of product instinct across the whole team.

Honest ambition

Simon coined the phrase "ambition bullying" to describe investors pressing him to claim unicorn ambitions before he had evidence for them. He told his eventual early investor, Lachy Groom, that if turbopuffer hadn't found product-market fit after a year, they'd wind the company down. Groom said fine.

That honesty is rarer than it should be. Most fundraising conversations reward confident projections over careful thinking. The consequence is that founders internalize goals that don't belong to them and then build organizations shaped by those externally imposed goals rather than by what they actually know about the market. turbopuffer's alternative — letting ambition grow alongside contact with the market — isn't timidity. It's accuracy.

None of this means the model transfers everywhere without friction. Infrastructure search is a real and defensible niche, and Eskildsen and Li had the technical depth and professional network from Shopify to bootstrap it credibly. They also funded initial salaries from their personal savings, which is its own kind of constraint not everyone can absorb.

But the core idea holds: staying lean means staying uncomfortable, and staying uncomfortable keeps you honest about what the product is, who it serves, and what the team can actually do. That discipline — not the whimsy, not the knit pufferfish — is what's worth copying.

Sources

• A love letter to a new kind of company (original)
• Found via the TLDR newsletter