Long Strange Trip

Ali Ghodsi on running a company by its one giant bottleneck

Ali Ghodsi· Co-founder and CEO of Databricks at Databricks
·~75 min·English·Sequoia
Business StrategyAI CompanyOpen Source
TL;DR

A CEO playbook from Databricks' Ali Ghodsi: run the company by finding its one giant bottleneck and pouring everything into it, hire for the skills you lack, attack rivals only where they are weak, and treat conflict as a discipline you learn.

01Core Operating Model

One Giant Bottleneck

Ghodsi runs the whole company by finding the single biggest thing blocking it and pouring almost all of his attention on that one clog, for as long as it takes.

usually it's like one giant bottleneck. And I think you should at that point identify what that giant bottleneck is and then focus the whole company

— Ali Ghodsi, Long Strange Trip
Key Insight
The method is really a bet against multitasking. A CEO's day is flooded with a million small fires, so Ghodsi compresses all of that and gives one clog orders-of-magnitude more attention than anything else. The danger is picking the wrong clog, because two or three years spent unblocking the wrong thing goes nowhere.
02Hiring

The Best Salespeople Are Not Technical

A company full of engineers kept scoring sales candidates on how technical they were, until Ghodsi looked at the actual top performers and found none of them were.

They need to be a little bit professionally aggressive. They need to have amazing EQ. They need to be the kind of, you know, command of the room.

— Ali Ghodsi, Long Strange Trip
Key Insight
The real lesson runs past sales: Ghodsi learned to hire for the skill he lacks, not the one he has. His engineers kept measuring candidates on their own axis and missed that selling to humans is a separate craft, refined over thirty years, that does not reward a PhD.
03Hiring

Hire Before the Seat Is Empty

By the time a role obviously needs filling it is already too late, so Ghodsi starts the search early and would rather turn down a strong candidate than rush into a hire that fails.

by the time you know that you need to hire some an exec, it's too late.

— Ali Ghodsi, Long Strange Trip
Key Insight
The cost is asymmetric, and that is the whole argument. A bad hire you keep too long burns two years or more; a good candidate you pass on costs only the candidate. So he tilts hard toward rejecting people and starts each search before the seat opens, so that time is working for him instead of against him.
04Competition

Hit Only Where the Rival Is Weak

Databricks was behind Snowflake on revenue and growth, so Ghodsi refused to copy its data warehouse and attacked only the three places it was structurally weak.

study your enemy carefully, understand their weaknesses, and apply your strength to their weaknesses.

— Ali Ghodsi, Long Strange Trip
Key Insight
He describes two opposite failure modes, and both lose. Ignore the competition and you get blindsided; obsess over it and you just copy its features. His third path was to name a new category, the lakehouse, and fight only where the incumbent could not easily follow, on openness, on AI, and on price.
05Execution

Put It in the Comp Plan

When Ghodsi is convinced a bet is right and the field will not prioritize it, he stops arguing and writes it into the sales compensation plan, which ends the debate.

don't argue with sales that's dumb. Put it in their comp plan. You know if sales doesn't want to do something then put it in the comp plan to do that thing

— Ali Ghodsi, Long Strange Trip
Key Insight
The comp plan is how he converts conviction into behavior without a fight. It also exposes his theory of scale: twelve thousand people left alone cancel each other out like atoms bouncing, so he over-communicates one black-and-white priority, even overriding ads that convert better, until it is impossible to misread.
06Second Acts

A Step Back to Go Ten Forward

Ghodsi wanted a diversified product portfolio rather than a one-product company, so he persisted with non-consensus bets for years before their payoff became visible.

you do have to have stamina. You have to do the thing you believe in.

— Ali Ghodsi, Long Strange Trip
Key Insight
What stands out is that Ghodsi took the bet from a position of success, not failure. At hundreds of millions in revenue he was, by his own account, disappointed and greedy for something much bigger, and that hunger is what let him keep pouring years into diversification while the market still saw only Spark. The scarce ingredient is not the idea, it is the stamina to hold a non-consensus bet until it is finally proven right.
07AI in Practice

The Bottleneck Was the Process, Not the Code

Ghodsi's two-day connector was only a prototype, so he dug into why production versions took his teams three quarters, and found the slow part was the requirements, integrations, and testing, not the code.

it's going to take humanity a decade at least to absorb and diffuse AI

— Ali Ghodsi, Long Strange Trip
Key Insight
The prototype separated two kinds of work. Writing the code was fast; the production version also took a quarter of requirements, an integration setup nobody enjoyed, and a quarter of testing at the end. That gap is why Ghodsi thinks AI takes a decade to diffuse: the models are already good enough, and the slow part is re-engineering the human process around them, one organization at a time.
08The Human Element

Conflict Aversion Is the Worst Trait

Ghodsi says the single most damaging thing a CEO can be is conflict-averse, because avoiding hard conversations lets a company drift instead of aligning.

the number one trait that is terrible for a CEO is if they're conflict averse.

— Ali Ghodsi, Long Strange Trip
Key Insight
He reframes a personality trait as a learnable discipline. Conflict aversion, in his telling, is no more fixed than a preference for skipping the gym; it is the default everyone has to push against. The tell he trusts is self-honesty, because the leaders who avoid conflict usually begin by lying to themselves about what is actually broken.