Ali Ghodsi, the co-founder and CEO of Databricks, has spent a decade building one of the most valuable private companies in enterprise software. But in a recent interview, he offered a confession that should humble every startup founder convinced their product sells itself: the shiniest technology rarely wins the deal.
Ghodsi's insight is deceptively simple. When you're a scrappy upstart trying to land Fortune 500 contracts, you don't compete head-to-head with entrenched vendors. You find a giant—a Microsoft, an Amazon, a Salesforce—and you hitch your wagon to their sales machine. Let their army of enterprise reps, who already have the relationships and the quarterly quotas, bring you into deals you'd never access alone.
The math of enterprise sales
Databricks reportedly crossed $2.4 billion in annualized revenue earlier this year, a figure that would make it one of the fastest-growing enterprise software companies in history. But Ghodsi is candid that raw product superiority explains only part of that trajectory. Enterprise buyers are risk-averse creatures. They trust vendors they already know, procurement processes they've already survived, and contracts their legal teams have already blessed.
For a startup, breaking into that ecosystem cold is prohibitively expensive. A single enterprise sales rep in the United States commands a base salary north of $150,000, plus commission structures that can double or triple that figure. Multiply that by the dozens of reps needed to cover major accounts, and you're burning through venture capital at an alarming rate—often before closing a single deal.
Ghodsi's alternative: partner aggressively with cloud hyperscalers who already employ thousands of such reps. When an AWS or Azure salesperson is trying to hit quota, they'll happily bundle a Databricks contract into a larger deal if it sweetens the pot. Databricks gets the customer; the hyperscaler gets the credit.
The dependency trap
This strategy is not without risk. Databricks has spent years navigating the delicate politics of being simultaneously partnered with and competing against the cloud giants. AWS, Microsoft, and Google all offer their own data analytics and AI services. At any moment, a hyperscaler could decide that Databricks is more competitor than complement and quietly deprioritize it in sales conversations.
Ghodsi's response has been to make Databricks indispensable enough that the hyperscalers can't afford to cut it loose. The company's lakehouse architecture—a hybrid of data lakes and data warehouses—has become a de facto standard for enterprises trying to unify their sprawling data estates. More recently, Databricks has pushed aggressively into AI model training and deployment, betting that the next wave of enterprise spending will flow to whoever controls the data infrastructure underneath large language models.
Our take
Ghodsi's candor is refreshing in an industry addicted to the myth of product-market fit as destiny. The truth is messier: enterprise software is a distribution game as much as a technology game, and sometimes the smartest move is letting someone else do the selling. Databricks' playbook won't work for every startup—you need a product that genuinely complements the giants rather than threatening them—but for those who can pull it off, it's a masterclass in leveraging asymmetric resources. The best technology doesn't always win. The best-distributed technology usually does.
