Databricks once charged customers about $1,000 for every user on top of what they consumed, and the fee was slowing the usage it is paid on, CRO Ron Gabrisko said on SaaStr AI’s podcast.
When Databricks made users free, “consumption went out the roof”, he said. Gokul Rajaram, writing on X, calls it a case of “Dashboard vs Pipes” pricing.
Business Pill · Take Rate
An evergreen Business Engineer explainer on take rates. It teaches the general idea only and does not discuss Databricks.
The key insight: Databricks’ CRO says a per-user fee of about $1,000 slowed the consumption Databricks is paid on, and making users free sent consumption up; Rajaram calls it dashboards versus pipes.
What Gabrisko Said
Early on, Databricks “put some user-based pricing on there” and a platform fee, Gabrisko said. He spoke with Monaco co-founder Sam Blond on SaaStr AI’s CRO Confidential podcast, posted on 28 October.
He asked customers, “Why aren’t you ramping faster?” Their answer: “For every user I have, it costs me a thousand bucks.”
So he asked: “Okay, if I just do all the users for free, you’ll just let everybody use it?” They said “Sure.” Consumption then rose sharply, he said.


What Was Posted on X
Gokul Rajaram wrote on X on 18 November that Databricks’ per-user fee was a gate in front of the compute consumption that is its primary revenue, and that removing it unlocked the flywheel. He splits products into dashboards, where the value is logging in, and pipes, where the value is output produced. The dashboards-versus-pipes quotes in this piece come from this post.
Dashboards vs Pipes
Rajaram’s frame: in “dashboard products” the value is logging in. In “pipes products” the value is output produced.
“Seat pricing works for dashboard products because predictability matters and users are the constraint,” he wrote. “It breaks for pipes products because the constraint becomes tasks completed, outcomes delivered, or compute consumed.”
Databricks is a pipe in that frame. “As data grows, Databricks grows. As queries grow, Databricks grows,” Gabrisko said. It bills in DBUs, its consumption unit.
What the Numbers Show
Databricks said in September 2025 it had crossed a $4 billion revenue run-rate. In August 2026 it said it had crossed $7 billion, growing more than 80% year over year.
That is 1.75 times between the two announcements, about 11 months apart, by our arithmetic. The August round valued Databricks at $190 billion, about 27 times its stated run-rate, also our arithmetic.
The company says more than 1,000 customers consume at over $1 million revenue run-rate, and more than 100 at over $10 million.
How AI Vendors Price Today
Anthropic’s price list sells both. Claude’s Team plan is $20 per standard seat a month, billed annually, and $100 for a premium seat. Its API is per token: Opus 5.5 lists at $4 per million input tokens and $20 per million output.
Salesforce’s Agentforce page offers per-user licensing, $2 per conversation, or Flex Credits at $500 per 100,000 credits, which it calls pay per action.
Gabrisko said he once told AI coding companies on per-user pricing they would never be huge without usage pricing. “Everything went to usage-based pricing now on tokens,” he said.
At Opus 5.5’s list price, the $1,000 a Databricks customer described per user would buy 50 million output tokens, by our arithmetic.
Usage pricing ties revenue to activity. This week Ramp’s CEO wrote that business spend on AI tokens went sideways after August, and Ramp’s SWE-Bench v2 measured coding agents’ cost per task. Databricks’ own data showed median AI user spend rising since July.
The Structural Read
As we read Gabrisko’s account, the fee did not fail because customers objected to price; it failed because it taxed the activity that produces Databricks’ main revenue.
As we read the vendors’ own price lists, AI pricing today mixes both models: seats for access, tokens, conversations or credits for output.
Ron Gabrisko, Databricks CRO, on SaaStr AI’s podcast
“Okay, if I just do all the users for free, you’ll just let everybody use it?”
Three Implications
FOR FOUNDERS Gabrisko says the base unit should scale over time and match the value the customer gets; Databricks uses DBUs.
FOR AI VENDORS Anthropic and Salesforce both list seat prices and usage prices on their own pricing pages.
FOR BUYERS Rajaram says seat pricing breaks where the constraint is tasks completed, outcomes delivered or compute consumed.
The Business Engineer Lens
This story maps onto the Business Engineer framework From SaaS to AGaaS.
The analysis puts it this way: “SaaS is charged for access to tools. AGaaS charges for the execution of outcomes.”
As we read Gabrisko’s account, Databricks’ fix was to stop charging for access and keep charging for consumption. The analysis predates his interview and does not discuss Databricks.
What Is Not Established
Gabrisko did not say when Databricks dropped the user fee, whether the $1,000 was monthly or yearly, or by how much consumption rose. Databricks’ figures are company-stated run-rates, not audited results.
The dashboards-versus-pipes split is Rajaram’s framework, not Databricks’. The quotes come from the podcast’s automatic captions.
The Bottom Line
Databricks removed a per-user fee of about $1,000 that customers said slowed their ramp, and consumption rose sharply, per its CRO. Rajaram’s test: seats where users are the constraint, usage where output is.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Gabrisko’s quotes are from the automatic captions of SaaStr AI’s YouTube episode (32:20-32:46 and 28:54-29:53); the SaaStr blog’s summary matches. The dashboards-versus-pipes frame and its quotes are from Rajaram’s post on X. Databricks’ run-rates and valuation are from its press releases; vendor prices are from the vendors’ own pricing pages, read on 19 November. The 1.75 times, 27 times and 50 million token figures are our arithmetic. Not investment advice; for information only.
Sources: SaaStr AI, CRO Confidential podcast with Databricks CRO Ron Gabrisko (YouTube), 28 Oct 2026 · SaaStr blog (Jason Lemkin), Databricks CRO Ron Gabrisko: 5 Interesting Learnings, 28 Oct 2026 · Gokul Rajaram on X, 18 Nov 2026 · Databricks press release, 13 Aug 2026 ($7B run-rate, $190B valuation) · Databricks press release, 8 Sep 2025 ($4B run-rate) · Anthropic, claude.com/pricing, read 19 Nov 2026 · Salesforce, Agentforce pricing page, read 19 Nov 2026









