Ali Ghodsi transformed Databricks: the coordination is the war
By Ray Dombrowski · Reporting from Youngstown ·
Ali Ghodsi transformed Databricks by embracing the difficult coordination required to ship products.
The software shop that ran like a mill until the spreadsheet cleared
On the Sequoia Capital podcast, Databricks CEO Ali Ghodsi recounted how he transformed his software firm in 2015. At the time, commercial revenue had stalled at just over a million dollars. The board hunted for his replacement while Ghodsi applied for faculty jobs at Berkeley. Ghodsi said a first-time chief executive survives by obsessing over one massive bottleneck rather than scattering attention across daily crises. He replaced his executive staff, swapped a product-led growth model for a fleet of enterprise sales veterans, and pushed the company toward a proprietary category called the "lakehouse." He did this despite internal resistance and a wave of public ridicule that painted the concept as a fantasy.
Scaling a business requires overcoming conflict aversion, Ghodsi said. Leaders must stop making excuses and face the facts of their operation. "I think I have a chip on my shoulder, you know," Ghodsi said. "Immigrant in Europe gave me a lot of chips on my shoulders and a lot of things to prove to myself and the world." He added that managing thousands of employees requires treating the organization like a managed machine—with clear inputs, predictable outputs, and rigorous maintenance—rather than a democratic club where everyone has a vote on every decision.
The fallacy of the frictionless office
Ghodsi mocked the fantasy that managers can engineer human friction out of existence with wikis and bots. When the host suggested that memos might replace meetings, Ghodsi pushed back. Trust requires a face. AI can write code in days, but shipping a product to a corporation still takes months of arguing over requirements, gathering signatures, and managing the egos of clients. The software is the easy part; the coordination is the war.
This is where the myth breaks. The Valley preaches automation, but floor managers and logistics chiefs know that people aren't code. They know that a missed shipment or a broken machine cannot be fixed with a Jira ticket. Ghodsi saw this when his engineers tried to ignore the sales process; revenue stalled until he hired veterans who knew how to navigate a corporate boardroom and speak the language of procurement. Software speeds up the work, but it cannot build the institution.
Conviction outweighs the quarterly report
Ghodsi moved from a Swedish classroom to a multi-billion-dollar empire. He survived through stubborn conviction. Executives often panic when numbers dip, killing strategic bets before the concrete has even set. Databricks endured years of ridicule for the "lakehouse" before the market shifted in their favor. It is a slow burn, similar to how rust-belt towns wait for trade policies to match the rhetoric of politicians. Both require a level of patience that the modern market rarely rewards.
A company is a series of audited bets, not a ribbon-cutting ceremony. When leadership maintains discipline over a five-year horizon, the headcount grows and the payroll stabilizes. That is the only metric that matters, whether you run a machine shop or a cloud empire. The goal is not the flash of the launch, but the steady accumulation of wins that make the business an inevitability.