Databricks sales growth tops 80%, but margin are shrinking from swarm of AI agents

Databricks said annualized revenue rose more than 80% year over year to $6.9 billion, with AI product revenue increasing to $1.7 billion from $1.4 billion in February. The company is benefiting from surging AI-agent usage and new products like Genie, Agent Bricks, and CustomerLake, but management warned that heavier model usage will الضغط margins lower as costs rise. Databricks also announced its acquisition of Panther and expansion into cybersecurity, reinforcing growth across AI and data software.

Analysis

The key second-order read is that Databricks is becoming a toll booth on AI usage, not just a software vendor. If agentic workloads keep expanding, the revenue mix should skew more toward usage intensity than seat growth, which is bullish for top-line durability but structurally pressure-testing margins because inference/model costs rise faster than software-like gross profit can absorb. That dynamic is a warning sign for every “AI platform” name that relies on external model consumption rather than owning the model stack.

For Snowflake, the competitive risk is not a direct feature-by-feature displacement; it is budget share. Databricks is broadening from analytics into workflow-specific vertical products, which can pull more data platform spend into a single vendor relationship and reduce the number of tools a customer is willing to keep funded. The bigger risk over the next 6-12 months is procurement tightening: if enterprises shift from experimentation to usage caps, the winners will be vendors that can prove marginal ROI per query, while pure volume monetization gets penalized.

The contrarian angle is that lower margins may be accepted by public investors if they are clearly tied to faster consumption growth, but in private markets it can compress valuation multiples if the market starts treating AI revenue as low-quality pass-through. That matters because listed comparables will be judged on take rate and gross margin trajectory, not just growth, and any signal that AI products are driving revenue while diluting economics can cap rerating potential. Near term, the main catalyst for SNOW is whether it can prove it captures similar agentic usage without sacrificing margin discipline; otherwise the relative multiple gap can widen further.

Key Decisions for Investors

• Short SNOW vs long a basket of data/AI infrastructure beneficiaries for a 1-3 month relative-value trade; thesis is that Databricks-style consumption growth will not translate equally across platforms, and Snowflake risks being viewed as the weaker margin-quality story if enterprise AI budgets stay constrained.
• Buy SNOW put spreads 3-6 months out on rallies into any AI platform enthusiasm; risk/reward improves if the market starts pricing in margin compression across data software names as agents increase query intensity and model costs.
• Pair trade: long security/data governance exposure vs short data platform beta over 3-9 months; if enterprises respond to runaway AI spend by enforcing controls, budget shifts should favor tooling that reduces waste rather than the platform that monetizes it.
• Avoid chasing long-duration multiple expansion in AI software until there is evidence of stable gross margin capture; use any post-summit strength to trim exposure in names where usage growth is being bought at the expense of unit economics.