Interview
ClickHouse CEO: AI Margins Need to Improve | Revenue Concentration Should be a Concern
- The company targets $1 billion in Annual Recurring Revenue (ARR) by December 2027, with Aaron Katz personally favoring the under on this figure, while also projecting ARR achievement within the next two years or sooner.
- Organizational scaling includes hiring nearly 1,000 employees by the end of the current year, with a strategic shift to in-person work options and the establishment of 16 to 20 global offices across 36 regions, specifically adding locations in Singapore, Sydney, and Tokyo within 12 to 18 months.
- The company plans to ship features and enter new product categories two years ahead of previous schedules, prioritizing long-term sustainability and customer acquisition over short-term valuation steps, with the public listing contingent on leadership discretion and potentially delayed until markets behave rationally.
- Financial performance expectations include gross retention exceeding 99% and net dollar retention surpassing 200%, while maintaining revenue diversity with no single category exceeding 10% of the total.
- Industry predictions forecast that within three to five years, AI agents will possess identity, budgets, and authorization for autonomous consumption, with 50% of tokens processed through open models and every company utilizing specialized models trained on proprietary data.
- The company anticipates a migration of stacks from hyperscalers back to on-premises infrastructure within the next few years, alongside a belief that open weight model security concerns will be resolved within one to two years.
- Strategic advantages will be maintained through proprietary features and cloud offerings that prevent erosion by hyperscalers or open-source redistribution, while the company expects to remain the default database for next-generation agent-built applications.
- Market dynamics suggest that the "AI finance function" dedicated to consumption management will become common within five years before becoming obsolete as agents self-govern, and that disruption will stem from undefined future technologies rather than existing competitors or "rear view mirror" tech.
- Operational constraints include reliance on mid-to-late 30s experienced talent paired with juniors, a rejection of centralized management structures, and an inability to secure star AI talent from Frontier Labs despite aggressive offers.
- Future scenarios rule out the irrelevance of doctors (with robots handling procedures), the necessity of daily office attendance, the concentration of revenue from a single category, or the need to raise $15 billion to $25 billion over the next decade, while emphasizing that human oversight will persist in architectural decision loops for the next ten years.
- Risks and limitations involve the difficulty of predicting AI sector winners and losers, the irrationality of public markets impacting morale, potential legal protection gaps for open weight models in sensitive data usage, and the preference for organic open-source development rather than government-funded model creation.