Commentary

Databricks vs. Snowflake trade public shots on X — Ali Ghodsi pitches his own competitor as a customer

Oct 29, 2026

Key Points

  • Databricks CEO Ali Ghodsi publicly pitched Snowflake on outsourcing AI inference entirely through Databricks' Unity Gateway, citing 30% token cost savings for recent customers.
  • The exchange began when Snowflake CEO Sridhar Ramaswamy questioned Databricks' 80% growth rate, asking whether it holds up after removing pass-through inference revenue and one-time license conversions.
  • Ghodsi's pitch exploits Databricks' private status, which allows selective growth disclosure without quarterly obligations, positioning the company as having cost advantages Snowflake cannot replicate internally.

Summary

Databricks Pitches Itself as Snowflake's AI Token Supplier

Ali Ghodsi, Databricks CEO, went after Snowflake on X with an unusual pitch: use Databricks to buy AI inference tokens instead of building that capability in-house.

The exchange started when Snowflake's Sridhar Ramaswamy questioned Databricks' disclosed growth rate. Ramaswamy asked whether the 80% year-over-year growth Databricks has cited holds up after removing pass-through inference revenue, one-time software license conversions to services, and native AI customer contributions—all metrics that would flatten the picture if stripped out.

Ghodsi's response was pointed: "I don't think we should fight. The two companies are complementary since you're not doing this AI stuff." He then offered Snowflake a direct commercial proposal. Databricks can route Snowflake's token requests through its Unity Gateway to access both open-source and frontier models, with smart budget allocation and routing. Ghodsi cited customers using Omnigen (compatible with Snowflake's harness) to cut token spend by 30% in recent months while token volumes rose.

The pitch works because it sidesteps a larger tension. Snowflake is a $115 billion public company near all-time highs. Databricks remains private and can selectively release growth numbers without quarterly disclosure obligations—giving it optionality that public markets don't allow. By suggesting Snowflake outsource AI inference entirely rather than build it, Ghodsi implicitly argues that Databricks' infrastructure and negotiating power with model providers creates a cost advantage Snowflake can't match internally.

Ramaswamy's follow-up was measured: "Generous. I was asking about the revenue, not about contributing to it." The tone stayed functional rather than hostile, though the subtext is sharp—both companies are competing for enterprise AI workloads, and Ghodsi just demonstrated that even a private competitor can publicly position itself as a better option for Snowflake's customers on the specific dimension where Snowflake is weakest.

Every deal, every interview. 5 minutes.

TBPN Digest delivers summaries of the latest fundraises, interviews and tech news from TBPN, every weekday.