While many scrutinize the low gross margins of AI companies, the real threat is low switching costs. As foundation models rapidly leapfrog each other, application-layer revenue lacks the durability of traditional infrastructure software, making customer retention a primary investor concern.
Historically, software was built for predictable human workflows. Now, with AI agents executing thousands of unpredictable, low-latency queries simultaneously, product design must prioritize their needs. These agents will eventually select their own infrastructure, fundamentally changing the B2B buying process.
For AI agents to move beyond human oversight, they'll need their own identities, budgets, and authorization to consume services. This creates a new enterprise tooling category focused on agent governance, ensuring they don't "run wild" with resources or access sensitive data.
Contrary to the belief that open-weight models will dominate, large enterprises will stick with frontier providers like OpenAI. The crucial factor isn't performance but legal protections and indemnification against issues like output inference and data privacy, which open models currently lack.
ClickHouse deliberately emulated Datadog's developer-led, self-service model for its initial years. This put pressure on product and engineering to build something developers wanted. Only after establishing that strong PLG foundation did they layer on an expensive, Snowflake-style enterprise sales motion.
Two widely held beliefs are being challenged. First, even innovative digital-native companies are considering moving infrastructure back on-premise from the cloud. Second, sales cycles into traditionally slow sectors like finance are compressing as large firms adopt new technologies faster than ever before.
While the market may currently reward high revenue concentration (e.g., NVIDIA's reliance on a few customers), operators should view it as a significant risk. ClickHouse's CEO actively manages exposure, ensuring no single customer or industry exceeds 10% of total revenue to maintain durable growth.
ClickHouse's revenue ramp (0, 12, 50, 200, >500M) is faster than any database before it, yet more gradual than today's AI apps. This highlights that infrastructure adoption, while explosive, builds on durability and high switching costs, not just viral growth.
The rationale for an IPO has shifted. Companies can now offer employee liquidity through structured secondaries, a key historical benefit of going public. This allows them to avoid the downsides of being public, namely the impact of daily stock volatility on employee morale and the pressure from short sellers.