Databricks’ Ali Ghodsi Never Wanted to Be CEO. Now He’s Among the Best
In a Nutshell
Ali Ghodsi became Databricks CEO despite never wanting the role, transforming a commercially failing $1.5M-revenue company into a lakehouse leader by identifying the single biggest bottleneck—commercial execution—and focusing the entire organization on it for years. He abandoned product-led growth for professional sales talent, hired Ron Gresco to build a real sales machine, and made the controversial lakehouse bet despite internal resistance and external ridicule, proving the strategy after 3-4 years. His approach centers on first-principles thinking, extreme focus on one strategic priority annually, being conflict-ready as CEO, and maintaining conviction through non-consensus bets that take multiple years to pay off.
These notes were generated by AI and may contain inaccuracies.
We're talking about humans. Why should I like you or trust you if I never get to meet you? If you're married, do you want to meet your wife and kids at all? Can't you just have a Google doc? We just write down the stuff you want to teach your kids and put it in the goddamn Google doc. Maybe you don't even need to respond to the Google Doc. Cloud cam.
Ali Ghodsi is the founder and CEO of Databricks. Ben Horowitz from A16Z named him as the best CEO without hesitation when asked. After interviewing him, the host agrees that Ghodsi is really sharp.
There were seven co-founders at Databricks. A few years in, the company wasn't performing well commercially. The board decided it wanted to make a CEO change. Ghodsi heard through the grapevine that they were interviewing other CEOs, including at least two specific people he knew about. He assumed they weren't going to pick him.
2015 was a turbulent year for Databricks. The company had great success with its open source project Apache Spark, but was not having commercial success. Gap revenue came in at $1.5 million that year. The vibe wasn't great. Ghodsi, who describes himself as a glass half empty guy, thought he was probably going to leave and that all the founders were probably going to leave.
Ghodsi actually applied for a faculty job at Berkeley and was going to take it. He faced the dilemma of whether to become CEO, which he never wanted to be, or go back to his dream of being a professor, which is what he always wanted to do.
The board did not conduct a formal CEO interview asking about strategy. There may have been some one-on-ones, but all of this was very hush-hush. Ghodsi was not allowed to know that there was a process going on.
The board's problem was that the company had very little commercial success. One would have expected them to hire someone with extensive commercial background. Ghodsi had negative amounts of that. a16z and Ben Horowitz were religious about founder-led companies because they had been through it themselves. Ben famously was asked when he was going to bring a real CEO into his company and wrote about it in his book. Horowitz didn't want to do that to the Databricks founders. He believed in picking a founder. Horowitz later said it was just a trial and they didn't even give Ghodsi a CEO salary. From Horowitz's point of view, they would try it out for six months or twelve months, and if it didn't work, they would put another guy in place.
Ghodsi thinks about this. Academia is not doing super well right now, so it would have probably been a big mistake. During that period, he introspected a lot about which path to choose. His dream was professor, and the CEO thing he wasn't even sure he wanted to do. He didn't have much respect for the business world and saw it as going to the dark side.
Ghodsi looked over his life and the decisions he had made up until that point and realized a pattern. When he was a kid growing up in Sweden, he wanted to be a game programmer. His dream was to move to LA and work for Electronic Arts. In high school, someone floated that he could come to the US and get paid $80K to be a game programmer. That was his dream, but his parents said no, he was going to university. He thought they were destroying his life, but then he went to university and learned a lot about computer science that he didn't know.
Ghodsi did an internship at a startup in the year 2000. The startup did really well and offered him to be number two guy there, drop out of school and be number two. The startup was called Room 33, and it was Europe's largest mobile startup at the time. It seemed like the opportunity of his life, but he went back to university. The IT bubble burst, and it turned out to be the best decision of his life.
The university offered him a professorship even without a PhD. He faced the choice of taking the professorship now and being a full professor, or doing an actual PhD at the top university, which was very difficult to get into. He picked the challenging option, and it opened his horizons. It was way harder and much more difficult, but he learned so much.
Ghodsi had the opportunity to come to Berkeley. Berkeley wouldn't take him as a professor, only as a postdoc. He took the chance, went to Berkeley, took the postdoc, and it again widened his horizon. Facing the CEO decision, he realized he should probably take the thing that challenges him. He knew nothing about being CEO, nothing about business. Going back to Berkeley was something he already knew how to do.
Ghodsi's philosophy became to challenge himself as much as possible because you only have so many years on this planet.
About one year in, one of the one-on-ones Ghodsi had with Ben Horowitz resulted in them giving him a much bigger equity package and bumping his salary. They talked about things they strategically need to do, and it started to become clear this was not just a practice run or temporary. A couple months after that, Ghodsi invited Ben to the company All Hands, where Ben said they didn't think Ghodsi was going to do it either and gave him a temporary salary.
When Ghodsi was stuck in the CEO chair and didn't know anything about it, he had to be on an extreme learning curve. He was working 100-hour weeks. He networked with all the best people and learned from everyone he could find. He read all the books. He was applying first principles thinking to everything he was learning, making sure it actually made sense rather than just copying what others said. He was building his own CEO playbook based on all the ingredients he was learning, similar to how Bruce Lee built his own Jeet Kune Do based on different martial arts. That way it was self-consistent.
The second thing that's been with Ghodsi the whole time is that companies are typically failing because of one major problem, maybe two or three, but usually one giant bottleneck. You should identify what that giant bottleneck is and then focus the whole company and your attention almost to an extreme, orders of magnitude of attention and focus should go to this one issue. If you overdo it and go too extreme overboard on that one issue, that's great. The problem is you most likely won't be able to actually unclog or remove this bottleneck, so you should just turn all your attention to that.
At that time, the open source project Apache Spark was super successful, the tech was super awesome, but there was no commercial success. Ghodsi put all attention on figuring out how to get the commercial engine humming. That was laser attention. He got the whole company mobilized, talked about this at every All Hands, thought only about this one issue. He still operates the company that way, trying to identify the main bottleneck at the time and focusing on unclogging that. There are so many other things you have to do every day to hire people, handle drama, people quitting, board issues, legal department, customers, revenue missing targets, millions of things coming your way every day, and you don't have time to move the needle on anything, so focus on the main bottleneck.
The cycle on tackling bottlenecks is a longer cycle. If the answer was months or weeks, that's what everybody's doing every day. That's all CEOs do because disaster lands in your lap and you have to drop everything else to fix that thing. That's not strategic focus, that's tactical stuff. Ghodsi would say a year, two years, sometimes three. Creating the lakehouse category took multiple years. When Databricks announced the lakehouse category, they were ridiculed online. It was a long journey to get to the launch because there was so much resistance internally and from consultants. When they finally launched it, people were laughing at them. You have to give things a few years. You can't just give up immediately. It's really important that you identify that focus really clearly because if you pick the wrong thing and focus on it for two or three years, needless to say you're not going to go anywhere.
One of the first things Ghodsi did on that bottleneck was revamp the whole executive staff. A year to a year and a half later, the executive staff running the company was completely different from the one he inherited. That was the one thing he knew they had to change because it wasn't working. Engineering and product he handed off to a co-founder and basically ignored it for two years. He focused his attention on everything else, building up the rest of the muscle of the company.
Databricks had been doing product-led growth. They loved AWS where people swipe a credit card. They didn't want to have any sales people. That was codified into the ethos and culture of the company. They hired sales engineers who were actually the sales department. They had a couple of AEs. It became clear that PLG wasn't working. In 2015 before Ghodsi became CEO, for a little while they ran an experiment called zero touch where they shouldn't even ever talk to a customer, it should all be automated funnel because it's PLG. Revenue was going up and it just flatlined for two quarters when they started doing that. At $1.5 million revenue, if they just get one AE and that person crushes it with a quota of a million or two million, that would be the whole thing they've done for the last few years. So they should just double down on hiring sales people.
Ghodsi ran an experiment in his head: let's look at companies where this is working. Are the best salespeople technical, super technical, super smart? Do they have PhDs? The PhD technical kids are the best sales guys? No, absolutely not. So he learned two things. Probably that's not the main thing. Secondly, it probably doesn't even have any advantage because if it had an advantage, there would be some of these sellers that were great salespeople who also happen to have a PhD who would be at the top making all the money, but there were none of those guys. Ghodsi doesn't think it's only not an advantage, it's a disadvantage to have a PhD.
It became pretty clear that the way they were hiring AEs needed to change. They were testing them on how technical they were, how smart they were, but looking at the best, none of the best are technical. They had an amazing sales leader they hired, Ron Gresco, who knew what great looked like. Ghodsi worked with him, and Ron should get 99% of the credit for what happened on the sales side. It became clear they shouldn't just get only the technical people. There are other skills that sales people have that you should probably go after.
One quality is that great technical people tend to be more introverted, generalizing here, and they're maybe not the most aggressive, opinionated people that are toe-stepping. But the best salespeople are actually aggressive, very aggressive. The first one of the first AEs that Ron hired, Dave, got them a meeting with a customer that they couldn't get a meeting with. In the meeting, the exec that had gotten them the meeting said get that AE out of here, I don't want to see that guy ever again. He's emailed my boss, everybody else, this is unacceptable. After the meeting, Ghodsi asked Dave what happened. Dave said Ron told him to get into the building, figure out some way of getting in. Nobody was responding to email, so he went through every door he could, had to break glass, got them the meeting. What are you complaining about? It became clear they need to be professionally aggressive. It's the squeaky wheel gets the grease. That's counterintuitive for engineers from professor land.
There was another meeting with one of the early AEs, Keith. They were in the room and the customer somehow got pissed at this guy and told him who the hell are you, what are you even telling me. Ghodsi thought that was pretty bad. Then Keith suddenly just started asking her these weird questions like what is your team working on. I thought we were supposed to do sales. At the end of the meeting, she was hugging him saying I'm so sorry for saying that. Ghodsi asked what he did in there. Keith said he noticed she was emotional, so he wanted to bring the subject to something she's comfortable with. Everybody's comfortable talking about what their team members are working on. He moved the subject to an area she was comfortable with. Ghodsi learned these guys have amazing EQ for handling really difficult conversations that get really uncomfortable. They're the kind of people that it's hard to ignore them. If they email you, you have to respond back to them.
It started becoming clear there's a rubric. They need to be a little bit professionally aggressive. They need to have amazing EQ. They need to have command of the room. Eventually Ghodsi started learning that the really smart ones actually figure out what's called the power base of the company, how decisions get made high up in an organization. It's actually complicated. Not a single person will say I'm going to buy this for $100 million. It's a complicated web of influencers and people inside the customer that make this decision. So they have to work that power base. This is an art and it's a very different art from what Ghodsi grew up doing. They should just hire the pros. Since it's so humanly intensive, sales people are dealing with other humans, a lot of the things learned about how to deal with computers and math and logic and physics doesn't apply over here. Humans haven't changed, so probably these techniques that they've learned over the last 30 years should be trusted.
Ghodsi revamped their sales team that way, and a lot of this he learned from Ron. He pushed back and they had all these arguments, but eventually Ron turned out to be right and they landed this model. Ron was unique because he's a classic salesperson. They call them meat eating, doing push-ups. He's the classic PTC, BMC style school seller. Ghodsi grew up at PTC. Ron had been selling FTP, which is free, and making a lot of money on selling free FTP. It's hard to sell free stuff. He was good at this, a phenomenal sales guy. But Ron was special because he also happened to have an engineering degree. He had an undergrad engineering degree, and then when he did his GSB degree at Stanford, he also got an engineering masters just for fun because he had cycles to burn. He could push back on them and talk in their language, but he was really from that breed but could speak their language and they could talk to each other. It wasn't like oil and water.
Ron was smart, but also the other types of smart that maybe an engineer won't have, but he had both types of smart. Those people are very rare that they not only have an engineering undergrad but they have masters in engineering with Stanford GSB. The second thing Ghodsi looked for is can they build the machine or can they just run somebody else's machine. The joke is do you have a driver's license, are you a good driver, okay, can you build the car for me? Very different. Ron had been in a startup called Cyclone and they went from zero to $50 million ARR. The company eventually sold to Axway and there he had seen the growth up until hundreds of millions. He had really seen zero to 50, 50 to 100, understood that scale in a very detailed way. He was very cerebral. You could argue with him like why don't we put the engine on the back or why don't we change the design this way, and he could explain why that would or wouldn't make sense. He had been in these two companies, basically not changed company for 11 years. That meant he had longevity. A lot of these folks will just bomb out. If you're doing something strategic like figuring out go-to-market motion, it's going to take you two, three, four years. If the guy or gal is bombing out after two years, you've got to reset and start over. He'd been there a long time, so Ghodsi knew he was going to be loyal. If the going gets rough, he's not going to quit in the first instance, which also meant that Ghodsi could probably be
Ron is still running sales as CRO, which is very rare in the industry. VPs of sales typically churn through companies every two to three years because companies are growing faster than they used to, and these executives are said to not be able to scale.
The classic PTC sellers are hard to manage. They're aggressive and a handful to manage. Ali felt that because Ron had survived the difficult early days, he wasn't going to leave from subsequent challenges, which allowed for a more aggressive management style and the ability to have huge fights without losing him.
By the time you know that you need to hire an executive, it's too late. There's a fundamental trade-off: with infinite time, you can hire the best person, but with a very short amount of time, you can hire someone quickly but the quality won't be great.
Ali tried to hire ahead of the curve, before the need arose, which gave more time to run the search for a longer period. This approach allowed for increasing false negatives and decreasing false positives.
False positives mean hiring someone thought to be great who turns out not to be great. It typically takes a year to figure out they're not working out. Even after six months, you might know, but you won't fire an executive immediately. After another six months, if you have the courage to address it, they might leave, which could take another six months. Then you have to restart the search, now gun-shy from the previous failure, potentially settling for a compromised candidate with no strengths but also no weaknesses. This process can lose two to two-and-a-half years.
The solution is to be really picky and skip on a lot of candidates who could be great. At Databricks, Ali skipped phenomenal people who went on to do great things and now have a chip on their shoulder about not being hired. The playbook was to err on the side of being extremely picky, take a decent amount of time, and start early.
Most of the executives who have been with Ali have been there for about a decade. There have been two cases where mistakes were made, with maybe more, but two really stand out.
A good search takes at least six months, maybe longer, and could take a year. Searches at Databricks have taken a year. People settle when they say they've looked at everybody and this is the best they can get, or that the role has been empty too long and they need someone in the seat now.
When Ali looked back at the two hires that didn't work out, all the signals were there even in the interview process. They told him things that were outrageous, and he knew they were outrageous, but he dismissed them. In one case, he rushed through the hire because it looked good initially.
Back doors are phenomenal. Talking to all of a person's past managers gives an insane amount of 360-degree view. About 10-20% of front-door references end up being truthful, while 80% are not.
Databricks had an interesting rivalry with Snowflake. At one point, Databricks was pretty far behind from a revenue standpoint but has since passed them. When HubSpot was evaluating both companies, they chose Snowflake.
The consensus at the time was that Snowflake was a sales-driven company with a sales guy CEO, while Databricks was product-driven. Ali felt this was unfair because they had Ron and an amazing team, and Andy built up an amazing machine like an army. However, it was true that Databricks was half the revenue and growing slower.
To overtake someone with double the revenue requires several years. Good strategies take multiple years and can't be done overnight.
The approach was to study the enemy carefully, understand their weaknesses, and apply strengths to those weaknesses. Snowflake built an amazing, game-changing product that disrupted all the data warehouses of the hyperscalers.
Snowflake had three key weaknesses:
- It was proprietary, with all data stored in their proprietary format. People didn't want to get locked in.
- No real support for AI, which was pretty weak despite their claims.
- It was pretty expensive, though this was probably because it was a great product and they could extract margin.
Databricks went after these three weaknesses directly. The open lakehouse approach meant customers own their own data in a completely open format without lock-in. AI capabilities were pushed hard because Databricks' roots were in AI and machine learning since 2009.
Cost was addressed by ensuring TCO was typically a third of Snowflake's, and this still wins on TCO today. The strategy was hammered extremely aggressively account after account with a very careful playbook.
The playbook wasn't about saying Snowflake sucks. It was a coexist strategy: identify the Achilles heel and press on those. Go into accounts, look at exact workloads amenable to machine learning, pull those out, and move the format to open source.
The lakehouse category exemplified this playbook. It was a three-to-four-year strategy to create this category, which was controversial at the time. People didn't believe in it, but eventually the category took off. Now everybody claims to have a lakehouse and open formats.
A professor once said to watch the competition but never follow the competition. The second part is really important. Some companies are not customer-obsessed, and some are not competition-obsessed, which is a mistake.
From the Art of War, all CEOs should study their enemy, their position, whether high or low, how many troops they have. But some companies become obsessed with competition and just copy them, which is the most common failure mode.
Product managers tend to just build the same features as the competition. Either you don't obsess over the competition, or you obsess and just copy them. Databricks didn't just build a data warehouse and say they had a better one than Snowflake, because no one would have believed them since Snowflake created that category.
The idea for lakehouse was controversial internally. Seasoned folks said don't do it because it's a funny word that's technical. Strategy firms did surveys and also said it doesn't work, proposing bland names like "unified data science platform." Sales folks internally said it doesn't make sense and is too technical.
They had to march ahead with conviction despite the risk of jeopardizing a successful company on this weird thing. Initially, the people who said don't do it were vindicated because people ridiculed it, making jokes about data rapids going through the data river.
The strategy was to mobilize the whole company behind it, not create a separate team. It became completely maniacal and religious across the board. If an article said great things but didn't mention lakehouse, it was considered horrible. Customer wins were dismissed if the customer didn't mention lakehouse.
Even great Facebook ads were rejected if they didn't mention lakehouse. The head of marketing showed conversion rates proving Spark sells better than lakehouse, but Ali insisted on using the lakehouse messaging despite worse performance.
Sales reps are motivated by their comp plans, and it was easier to sell the core product. Ali was personally obsessed with comp plans, always including multipliers, extra spiffs, and gravy on top. Every year, the comp plan was adjusted so that if sales doesn't want to do something, put it in their comp plan to do that thing. If convicted about the right approach, put it in the comp plan and stop the debate.
There have been a couple of points in Databricks' history where they decided to take a step back in hopes of taking ten steps forward. This includes moving from PLG to enterprise, really doing AI, and the lakehouse initiative. Many CEOs are nervous to take that step back, worried about what investors will say if revenue slows down in the short term.
On the private side, there's more latitude. When lakehouse was initiated in 2018-2019, Databricks had multiple hundreds of millions of dollars of ARR and was successful by some metrics, but Ali was unhappy and wanted the company to be much bigger and more impactful.
When talking to public company investors, Ali would paint a big vision of what he wanted Databricks to do, but afterward would truth-seek and realize they weren't actually doing that. They were just selling Spark. It came from a place of not being content with a single-product company and wanting multiple diversified successful products.
Ali studied companies in Silicon Valley and saw that one-trick ponies like Splunk have amazing innovation but are defined by one thing, while Amazon, Google, and Microsoft have many products and can't be defined simplistically. The goal was to become one of those companies, requiring big bets that might not succeed.
Ali was stubborn and kept going with these bets for a few more years until eventually seeing success. It came from not being content and not wanting a single-product company. The same transition happened at HubSpot, going from a marketing app company to a CRM platform, taking a massive step back to take ten steps forward.
Public markets hear so much BS on earnings calls that they discount it and just want to see the money. It takes multiple years until people start paying attention to the thing that's going to pay off.
Databricks was cloud-only AI and no one cared because on-prem is where all the business was. It took five years until it took off, and then suddenly people asked how it happened so fast. The reaction is always that people can't believe how quickly all these products were built, not realizing it took five years of work they didn't pay attention to.
Stamina is required to do the thing believed in. It has to be non-consensus but also right, and pursued for four or five years before people start looking at it.
People who have worked with or for Ali describe him as a killer who plays more aggressive than seems possible. Growing up in Sweden as an immigrant, switching schools and environments, created a chip on the shoulder and the aggression needed to succeed. There's probably a lot of tough anger that piled up over the years. The United States has softened him because it's a happy place with harmony.
As a CEO, the number one terrible trait is being conflict-averse. Conflict-averse CEOs are the worst possible thing for the role. The CEO job requires being crystal clear because everybody is pushing constantly. If conflict-averse, you'll say yes to everything or lie or avoid subjects.
Being crystal clear means saying no definitively. If someone catches you in the hallway or elevator and tries to jam in an approval, that's sneaky and they need to go through proper processes. Without this clarity, the company starts going in all kinds of directions instead of laser-focusing energy on the bottleneck to solve.
Being aggressive helps with consistency and clarity so everybody gets the message. People don't understand nuance. The approach is to make it black and white with no doubt, giving everyone the cheat sheet. With 12,000 employees, big companies start canceling each other out with atoms bouncing off each other and strategies that don't make sense.
Ali Ghodsi noted that Databricks reached 12,000 employees from a much smaller base, and acknowledged the challenges companies face when scaling past 150 people. He agreed that slowdown is somewhat necessary during this transition. Ghodsi identified the CEO transition period as when this shift occurred, around the time the company moved from everyone knowing each other to requiring more formal structure.
When a company is small, the CEO can run it like a star where one or two people know everything happening across the organization. As the company grows, CEOs must transition to putting structure in place so they don't need to know everything directly. This means trusting that different parts of go-to-market and operations will function without constant CEO oversight.
The critical change occurs when managers of managers become meaningful rather than nominal. Previously, these intermediate layers existed but didn't matter because the CEO could reach down and understand everything. Once the organization becomes too large for the CEO to reach down directly, the intermediate layer must actually function as the conduit for information and decision-making.
Ghodsi emphasized putting processes in place where everything reviews and KPIs are measured so leadership can ensure the ship is heading in the right direction without tapping people's shoulders constantly. This becomes necessary because the CEO no longer knows everyone's name or can directly monitor all activities.
The discussion turned to Jack Dorsey and Brian Armstrong's approaches to flattening organizational structures. Dorsey describes the company as an AI with employees training it and providing context, advocating for no more than two layers between the CEO and frontline employees. Armstrong similarly focuses on collapsing organizational levels rather than just shrinking headcount.
Ghodsi sees directional correctness in these approaches but argues for separating two distinct concepts: the official org chart versus how information disseminates, aggregates, and decisions get made. Twenty years ago, these were the same thing because information could only flow through the tree structure of org charts. Today, information can be broadcast top-down or bottom-up.
Ghodsi shared that when he asks audiences whether they believe AI is smarter than most people around them most of the time, approximately 90% raise their hands. He defines this as AGI - artificial general intelligence that matches most human capabilities, even if not superintelligence.
Despite this recognition, enterprises are using AI primarily as chatbots for asking questions and getting answers, and for coding assistance. Ghodsi noted that people complain about AI hallucinating, but humans make mistakes constantly as well.
The critical gap is enterprise context. Ghodsi asks audiences whether they have hundreds of agentic co-workers collaborating, providing proposals, and keeping them updated, and no one raises their hand. What's lacking is capturing the enterprise context from people's heads, daily decisions, recorded meetings, and emails.
Databricks built a product called Genie that captures this ontology and feeds it to AI systems. At Databricks, this creates what Ghodsi describes as magical capabilities where Genie can answer any question in any meeting and automate tasks. The company uses the Genie app on phones to continuously feed context and automate processes.
While agreeing philosophically with Dorsey's intelligence argument, Ghodsi questions whether this applies to human org charts. He expressed skepticism about companies moving to 25-person manager spans with player-coach models where managers are expected to code 80% of the time and manage 25 people 20% of the time.
Ghodsi argued that management involves significant human interaction around career development, unclear goals, and personal issues that can't be handled effectively with 25 direct reports. He believes this approach will break down, leading to unhappy employees who don't understand what's happening because their managers are too busy coding to meet with them.
Ghodsi confirmed he still codes and never stopped, having coded since third grade and run engineering and product functions. In November of the previous year, he began committing code to production to understand the full development lifecycle.
He started by building a connector in two days that teams were taking three quarters to build. When told his version wasn't production-ready, secure, or thoroughly tested, he investigated the gap between proof-of-concept and production deployment.
The team initially claimed they could compress the timeline from three quarters to seven and a half months using AI, but Ghodsi pushed for the real explanation. The actual bottlenecks weren't coding speed but customer feedback collection for PRDs, which took a full quarter, and setting up connections to systems like Salesforce, Workday, and NetSuite, which teams found difficult and time-consuming.
The solution involved hiring consultants for system setup work, reducing PRD development from a quarter to one week by accepting that initial versions could be refined later, and moving testing earlier in the process using AI for end-to-end validation. They also shifted from one person per connector to seven people working on seven connectors simultaneously to handle absences and maintain momentum.
Ghodsi believes it will take humanity at least a decade to absorb and diffuse AI across organizations, based on his experience watching even a tech company with super-smart, AI-willing employees struggle with process re-engineering. This observation has informed his view since the beginning of the year.
Ghodsi described his approach as focusing on three areas: identifying the main bottleneck that could 10x the company and carving out time to address it since others are busy running existing operations; handling the daily CEO responsibilities that pack calendars; and spending remaining time on product details given the 3,500-engineer portfolio.
He tries to compress calendar commitments to create free blocks for thinking about major initiatives rather than working 8 AM to 5-6 PM back-to-back meetings, which he considers being a slave to the calendar and working on everyone else's priorities.
Databricks operates with significant trust built over 10+ years with many employees. Ghodsi knows individual strengths and weaknesses, understanding how people are likely to succeed or struggle in specific areas.
The organization holds staff meetings three times weekly: Mondays for 60-90 minutes focusing on top priorities and KPIs tied to one or two annual goals, and 30-minute Wednesday and Friday morning meetings to address immediate bottlenecks and maintain team cohesion.
Ghodsi argued that in-person interaction is essential for building trust and preventing departmental silos, comparing it to asking why families need to meet when they could just share Google Docs. The goal is ensuring executive team members prioritize company-wide objectives over their functional teams.
The company conducts multiple types of QBRs quarterly covering product, go-to-market, and other areas to detect issues across regions, departments, and products. These reviews generate action plans for addressing problems like underperforming products or regions.
Offsites occur at least quarterly, focusing on top bottlenecks rather than abstract strategic discussions.
When asked about going public, Ghodsi stated that Databricks would likely be worth more as a public company currently, though this varies over time. The company has facilitated employee liquidity through private tenders serving approximately 15,000 people including ex-employees across nearly 100 countries, managing complex tax implications.
He believes public markets make sense for liquidity facilitation eventually, but prefers remaining private during the current major transition period, noting that companies undergoing significant changes benefit from private market flexibility.
Companies going through big transitions like moving from permanent licenses to SaaS licenses often go private, execute the transition, then go public again. During the current AI transformation, staying private makes more sense than going public. Companies like Anthropic and OpenAI have massive capital needs, while Databricks is free cash flow break-even and doesn't require additional capital. The public markets are experiencing extreme volatility with contradictory narratives about AI companies, from fears of a "SaaS apocalypse" to sudden enthusiasm for semis as a bottleneck play. This back-and-forth makes public markets unsuitable for handling revolutionary transitions.
Public markets struggle with transformation periods and perform better during normal market conditions. They initially viewed software as having zero value, then reversed course when revenue started accelerating. This inconsistency makes it preferable to wait until there's more predictability before considering going public.
When becoming a CEO at scale, developing a "BS detector" and becoming comfortable with conflict requires several key practices. The most important principle is to stop lying to yourself about what's wrong with the company, individuals, or strategy. Self-defense mechanisms often create excuses that prevent seeing the truth. Truth-seeking requires being willing to face discrepancies head-on rather than avoiding them. Even if someone is naturally conflict-averse, this can be overcome through deliberate practice.
Conflict aversion should be treated like other difficult but necessary habits, similar to going to the gym or eating vegetables. Just as working out is unpleasant for everyone regardless of natural motivation levels, facing conflict is inherently difficult. The key is pushing through this discomfort rather than using personality traits as an excuse. If someone wants to be a CEO, they cannot claim to be "the kind of person who avoids conflict" - this is incompatible with the role requirements.
Being a CEO requires getting crystal clear with people and tackling conflicts directly. While nobody enjoys conflict as it involves pushback, criticism, and unpleasantness, this is simply part of the job. The distinction between people who avoid conflict versus those who handle it better is a matter of degree, but complete avoidance is not an option for effective leadership.
Two prominent CEOs, Ali Ghodsi and Nikesh Arora from Palo Alto, share similar characteristics: strong BS detectors, comfort with conflict, and resistance to being liked as a primary goal. Over-indexing on being liked, as measured by awards like Glassdoor recognition, may not correlate with long-term success. Steve Jobs exemplified this approach of not prioritizing likability.
Successful leadership requires both the courage of convictions and staying power to maintain them. Conviction without persistence resembles a New Year's resolution that fails quickly. The most valuable form of conviction involves being right about something that everyone else believes is wrong. This combination of correct contrarian belief, strong conviction, and patience produces the largest outcomes.
While conviction around founding ideas receives significant attention, conviction around second acts proves even more critical for scale-up CEOs. The Lakehouse product launch exemplified this, representing a major competitive offering that was unique in the market. Despite skepticism from most employees and external observers, sustained focus over multiple years led to success. This demonstrates that second acts often require more conviction than initial products.
When making major strategic transitions at scale, such as moving from applications to platforms, leaders must swing the pendulum fully toward the new direction rather than taking a middle approach. Partial commitment allows organizations to default to existing first acts. Focus should be limited to one or two key areas per year, with complete organizational attention directed toward the bottleneck issue. The Lakehouse product received this level of singular focus for two to three years as the annual planning priority.
Keep Sequoia Capital in your library
Save the videos and channels worth coming back to, and find them again in one place.
