My first Condensing the Cloud post centered on some simple — but important — advice for founders: “To Build Value in Tech, Build Different.” In that post, I used Databricks* as an example of building different, as the company continues to develop new technology to meet the needs of emerging data science and AI workloads.
To that end, I invited Databricks CEO Ali Ghodsi onto the Condensing the Cloud podcast — for our first-ever episode — to discuss the Databricks journey, in the hopes that up-and-coming entrepreneurs could learn from his experience.
Watch the full episode here:
Highlighted Excerpts
Lightly edited for ease of reading.
Max: Ali, great to have you here. Maybe I should start by asking: Is there something in particular that you now think, "Oh, if only I had known this earlier, I would've done so much better, gotten here so much faster?"
Ali: Yeah, I mean, I think there's so many things, so it's hard to pick, but I would say in the early days, I didn't have any experience outside of just the core things that we were innovating on. We were building Apache Spark at the time, which was a data-processing engine, but how do you do marketing? How do you do sales? How do you do finance? And so on. And I think if I had gotten one piece of advice, it would've been to be more trusting of the people that are doing those things in those fields.
And how do you figure out how to find the best in those fields to come work for you? It's just much faster if you can find the right people and leverage them rather than having to reinvent the wheel from scratch every time. So I think that was the biggest challenge. We knew how to do engineering and how to build our products, but that whole other thing that's needed for building a B2B company… it just took a very long time to figure out the steps.
Max: Well, engineering is logical, it's deductive. The people who do engineering are smart in a similar way to how you're probably smart, but then sales, sometimes there's a lot more intuition and interpersonal stuff that comes into it. And so sometimes the great salespeople are great because of a totally different set of skills than what you're used to. So I imagine it was some adaptation to learn to value and appreciate those skills and understand the mix of them that it takes to be successful.
Ali: Yeah, I mean, the kind of salespeople you want to hire as an engineer are probably not the best salespeople. They're probably more like engineers than salespeople, and they're not necessarily the most effective. It's, "How do you find the people that really excel at their job in these other fields?" They're not going to approach it the way you approach it or think about it the way you think about it. So you have to resist this urge of just kind of finding people that are, again, engineers just like you, but dressed up as marketing or sales or something else.
Max: So what was your learning curve to figure that out? What approach did you take? You knew you had this problem that was totally different from the sorts of problems you had grown up solving. It wasn't an algorithm problem. It was like, "How do I find people who are good at sales and find people who are good at marketing and find people that are good at finance?" Did you decide to learn some of those fields yourself or did you rely on your network? How did you even approach that problem?
Ali: People that have the kind of background I have, they'll read the books, they will learn all the stuff about that field, kind of go bottoms up and build up that kind of... "How hard is this? I'm going to learn everything about it." That's good. But I think another thing that was kind of useful that maybe I didn't get advice to do is just spend a lot of the time in those fields and find the best in the market.
Don't just go read the books and become book smart in sales. Think about how to figure out who is the best in that field and learn from them. Repeat the same thing in marketing, repeat the same thing in finance, repeat the same thing in customer success. Frankly, also repeat the same thing in engineering because while we knew how to engineer systems with a few researchers in a research lab at UC Berkeley, building an engineering factory with product management, with design, with technical program management and all those kinds of things, that itself is art that we didn't know. So we had to repeat it and do it again in all of them.
Max: In your journey, a lot of the hypergrowth came at a time where capital was relatively available. But now that capital is a little tighter, companies are needing to sometimes blend in more of the efficiency much earlier in the cycle. If you look back, do you think it would've helped you maybe or hurt you a lot if you had to blend in efficiency earlier? You could almost see it either way. What do you think?
Ali: I really firmly think that for Databricks' case, for us, it would've hurt us if we grew up in this era. I think a lot of it depends on the company and the CEO. I was paranoid from day one, so I was never like, “Oh, let's just burn all this money. Who cares? This is going to be trillions of dollars, and none of this matters.” I was really keeping tabs on how much time, how much runway do we have? How far do we need to get? How big are these shoes that we're buying? How fast are we going to grow into these shoes? I was doing those kinds of calculations all the time, every year. And so we were paranoid, we were careful. But having said that, it gave us the ability to grow extremely fast.
The whole strategy was: let's be very calculated and take calculated risks, but let's fly close to the sun and let's fly fast. And if the strategy works, we'll be big enough that, at that point, it's hard to kill us. There were just so many small startups that were all doing the kind of stuff we did. We all sounded the same. It was all data, blah, blah, blah, analytics, something easier, better, faster. So, how do you really separate yourself from the masses? So the strategy was to raise a lot of money, execute fast, but thoughtfully and make sure that the company doesn't go under while you were doing that. So I think we would not have been the size we are today, for sure. I actually don't think even would've been as successful, but I certainly don't think it would've been at the size that we are today if funding would've been very limited and it would've been this push for efficiency from day one.
So that's my view firmly about Databricks. I'm not saying this applies to every company. I think it could be very different for different markets, different CEOs. And certainly, I see a lot of companies these days, especially in the AI field and the way they're making investments, where I'm looking at it, and I'm saying, that doesn't make any sense to me. It doesn't make sense to me if you don't have any income to go buy a gazillion GPUs, 10,000 GPUs, 20,000 GPUs, 8,000 GPUs, and then build something and say later we'll figure out how that's going to pan out. I would not do that. So that's way above my risk tolerance level. I would never do that, and I'm skeptical that it actually can work.
Max: Are there two or three things that you just wish you could buy a piece of software that would do? How come this thing doesn't exist and the software for this is so crappy, outside your core product, things that you would want as a user?
Ali: So things that people should go build right now?
Max: Yep.
Ali: Well, first of all, I think large companies are all doing this optimization game. They have big ops teams. These are people that pull things out in Excel and they do analytics, each department has lots of these people. They're called analysts, but it could be marketing analysts or finance analysts and so on. They pull the data out of these systems, they do some Excel math on it, then they produce some dashboards. Then, those dashboards go to meetings for execs and then they make decisions about how to do investments in their companies. This, I think, certainly has to get automated. This manual work to pull data out of all the systems of record, put them in Excel, do some manual basic math on it, then dashboard it and then produce PowerPoints and then back again.
It's like the backbone of how all large corporations are run today. You have to be able to automate that. And I think there's also a huge lack of AI in the analytics that those people then do. They pull it out, they put it in Excel. But largely, they're just looking at histograms and trend lines, and well, it looks like it's going that way, but you gotta be able to leverage AI for this stuff. So I think there's a huge opportunity for that to happen. I think all these systems of record software companies that are out there, not singling out any particular ones, but we have... Workday does your HR, Salesforce does your CRM, NetSuite does your finance, all of them. I'm not singling out any particular one. All of these, once AI is really infused into them... and that's not what we do, by the way.
We're not in the AI, HR field. But once AI is infused to those, you should be able to do a lot of these things automatically, who should get promoted, what should be the increase in salary. In sales, what are the accounts that are going to grow? How much should the quota be? What are the territories for them? How much should people get paid? Which accounts should you go after? Who's going to churn? All that kind of stuff. In finance, what is your revenue going to be? How are you going to forecast that? What are the confidence intervals on it? All these kinds of things are actually today done manually in large organizations with Excel. So I think there's a huge opportunity here. Basically, take any system of record and AI infused version of it in the future, I think it's a great opportunity.
Same thing for SIMS, they take massive data and they process it and they alert you on security, AI is going to be infused into those. It's not quite happened yet. Customer 360, CDP, every organization wants to understand their customers. I think there's room to innovate on all of these, pretty much the whole SaaS stack. And I haven't even gone to apps, like lawyers and doctors and teachers. I'm not even looking at those. I'm just saying all the systems of record, software, SaaS companies would pretty much have to be reinvented in this new AI era. So I think there's going to be this great opportunity.
I think the problem is it didn't quite happen in 2023. So now I'm worried that I think, in 2024, there's going to be a trough of disillusionment, or maybe even 25, of like, none of this worked, none of that was true. So this is overhyped. But then surely enough, all of these things that I said are going to happen. They just take a little bit longer than a year.
*Denotes a Battery portfolio company. For a full list of all Battery investments, please click here.
The information contained herein is based solely on the opinion of Max Schireson and Ali Ghodsi and nothing should be construed as investment advice. This material is provided for informational purposes, and it is not, and may not be relied on in any manner as, legal, tax or investment advice or as an offer to sell or a solicitation of an offer to buy an interest in any fund or investment vehicle managed by Battery Ventures or any other Battery entity. The views expressed here are solely those of the author.
The information above may contain projections or other forward-looking statements regarding future events or expectations. Predictions, opinions and other information discussed in this publication are subject to change continually and without notice of any kind and may no longer be true after the date indicated. Battery Ventures assumes no duty to and does not undertake to update forward-looking statements.
Ali Ghodsi on Databricks' Growth Journey, Lessons in Leadership and What Founders Should Build Today