AI transcript
0:00:03 – Welcome to the A16Z podcast.
0:00:05 I’m Doss Rush, our enterprise technology editor,
0:00:07 and in this podcast, I moderate a panel discussion
0:00:09 on some of the most heated topics in open source
0:00:12 with two of the leading founders of open source companies,
0:00:15 Armand Dodger, co-founder and CTO of HashiCorp,
0:00:18 which does open source tools for managing multi-cloud,
0:00:20 and Oli Goetze, the CEO of Databricks,
0:00:22 a SaaS offering of Apache Spark.
0:00:23 Joining them in conversation
0:00:26 is A16Z general partner, Peter Levine,
0:00:28 who’s invested and been on the board
0:00:29 of numerous open source companies,
0:00:31 such as GitHub and Netlify.
0:00:33 It’s a great discussion and it takes on everything
0:00:34 from making money on open source,
0:00:37 while managing community, to the nuance of partnering
0:00:40 and sometimes competing with big cloud vendors.
0:00:42 Just to note that this was recorded at a live event,
0:00:44 so there are some audio issues.
0:00:47 The first voice that you’re gonna hear in this is Armand’s,
0:00:49 then Oli, and a few minutes into the conversation,
0:00:50 Peter joins them.
0:00:52 Finally, please note that the content here
0:00:54 is for informational purposes only.
0:00:56 Should not be taken as legal, business tax,
0:00:58 or investment advice, or be used to evaluate
0:01:00 any investment or security.
0:01:02 And it is not directed at any investors
0:01:05 or potential investors in any A16Z fund.
0:01:09 For more details, please see a16z.com/disclosures.
0:01:12 Throughout the history of open source,
0:01:14 talking about making money on open source
0:01:16 has been a pretty controversial topic
0:01:18 with a lot of different views.
0:01:20 So, I’m curious, Oli and Armand,
0:01:23 how have you thought about commercializing open source
0:01:26 and why did you choose to turn a project into a business?
0:01:29 – For us, it didn’t start necessarily
0:01:31 as thinking about turning the open source
0:01:32 into the business.
0:01:36 It was more around recognizing that there’s a clear market gap
0:01:37 in terms of, in our case, sort of DevOps tooling,
0:01:39 how do we actually provision things
0:01:41 and sort of cloud infrastructure.
0:01:43 And then realizing it’s very hard
0:01:46 to become a large sustainable project
0:01:48 if you have negative cash flow forever.
0:01:50 And if you’re at a university and great,
0:01:51 you have grants and things like that
0:01:52 that can sort of fund it,
0:01:53 or it’s a little hobby project
0:01:56 and it’s two, three people doing it on a weekend, fine.
0:01:58 But if you’re solving a large enough problem,
0:02:00 you eventually need teams of dozens,
0:02:03 hundreds, thousands to work on that problem.
0:02:04 You kind of need a business.
0:02:07 There has to be sort of a top line connected
0:02:08 to the bottom line,
0:02:09 otherwise it doesn’t make a lot of sense.
0:02:12 And so I think for us, it started pretty pragmatically
0:02:14 in terms of, hey, we are passionate about the technology,
0:02:15 passionate about the space.
0:02:18 We want this to be viable long term.
0:02:20 Well, the only way for it to be viable long term
0:02:21 was if you make money.
0:02:24 – I’m just going to be honest.
0:02:25 We were academics.
0:02:27 We just wanted to have impact
0:02:29 and we wanted to publish papers.
0:02:32 And the software we built, people didn’t want to adopt it.
0:02:34 So we went to all the companies that were out there,
0:02:36 went to cloud error, all these guys,
0:02:37 and said, please take the software, take it with you,
0:02:39 take credit for it.
0:02:41 And they all just refused.
0:02:44 They said, this is just an academic mambo jumbo,
0:02:45 not interesting.
0:02:47 What if these PhD students just leave?
0:02:51 And then this is enterprise software we’re selling.
0:02:52 So they just rejected us.
0:02:54 So in 2013, we were kind of frustrated and we said,
0:02:56 if we want to actually succeed,
0:02:57 the only way we can get these projects off the ground
0:02:59 is if we actually start something ourselves.
0:03:00 They wasted so much of our time.
0:03:02 We put interns into these companies.
0:03:03 – What do you really think?
0:03:05 (laughing)
0:03:06 – We were sending interns into these companies,
0:03:07 hoping that they would adopt our software.
0:03:08 It never happened.
0:03:10 Since we were frustrated in 2013,
0:03:13 we just said, let’s create a company ourselves to do it.
0:03:15 I don’t think revenue was sort of top of mind for us.
0:03:16 The first two years of Databricks,
0:03:19 our only goal was how can we make spark
0:03:21 in the software that we have take over the world.
0:03:23 – Let’s talk about the flip side.
0:03:26 You choose to go out and then have a commercial venture.
0:03:28 How have you gone about managing your community
0:03:29 and communicating that with them,
0:03:32 kind of keeping their support for what you’re doing?
0:03:35 – I think for us, it goes back to Peter’s point around
0:03:38 having a clear product management framework
0:03:40 that you can articulate where your community
0:03:42 doesn’t feel like you’re just randomly picking
0:03:44 what goes one way or the other.
0:03:47 And I think for us, it was really trying to draw that line
0:03:49 and saying, okay, great, the things that are truly
0:03:52 organizational complexity problems, right?
0:03:54 You need role-based access control.
0:03:55 You want audit logging.
0:03:58 You need PCI ISO SOC compliance, things like that.
0:04:00 You’re like, okay, great, if you have those problems,
0:04:02 you’re probably a global 10,000 business.
0:04:04 You’re not a SMB hobbyist.
0:04:06 And I think we drew that line and articulated to the community
0:04:08 and said, hey, things in this bucket, great.
0:04:10 They go into enterprise and the people who are gonna pay
0:04:12 for that are the people who have that problem.
0:04:15 You as a hobbyist don’t care about, you know,
0:04:17 hardware security device integration
0:04:18 for your compliance, right?
0:04:19 Like it’s not a problem you have.
0:04:21 And so I think if you articulate it clearly that way
0:04:23 and have the discipline to stick to it,
0:04:24 then the community doesn’t feel like
0:04:26 they’re sort of randomly being jerked around.
0:04:28 And they don’t feel like they’re losing value
0:04:30 ’cause those aren’t problems they have.
0:04:31 – Yeah.
0:04:32 For us, we’re different.
0:04:33 I mean, these guys are much smarter than us.
0:04:35 We didn’t think these things through.
0:04:37 I mean, we just wanted Spark to take over the world
0:04:38 as open source projects.
0:04:40 So we only had one roadmap.
0:04:41 It was the open source roadmap.
0:04:42 We wanted to sort of, and we were frustrated
0:04:43 that no one would adopt it.
0:04:44 And there was a lot of fun in the market
0:04:48 that the technology we built won’t work for this
0:04:49 and won’t work for that.
0:04:50 It only works if it’s in memory,
0:04:52 not if it’s on this, this kind of thing.
0:04:53 So we were super frustrated.
0:04:56 So the first three years, our only goal was get adoption.
0:04:57 We didn’t care about any revenue.
0:04:58 I mean, three, four years in,
0:04:59 we only had one million revenue.
0:05:01 So we only had one roadmap.
0:05:03 We only managed the community.
0:05:05 And then in 2015, it just exploded.
0:05:06 Like it exploded.
0:05:08 Like, I mean, you saw some of the curves.
0:05:10 It’s like, you know, thousands of developers
0:05:12 started contributing code to it.
0:05:14 And at that point, that’s when things started,
0:05:15 the community was so big
0:05:17 and this thing became way bigger than we are.
0:05:19 Like, we were like, unknown.
0:05:21 And then the project was huge.
0:05:22 So at that point, we started thinking through,
0:05:24 hey, how do we actually monetize this thing?
0:05:26 – Can you talk a little bit more about,
0:05:27 you know, that moment in 2015
0:05:29 when you have this huge community
0:05:30 and now you’re starting to think
0:05:32 about the secondary roadmap
0:05:33 where you’re commercializing something.
0:05:35 How did the conversations change?
0:05:36 – Yeah, well, so what happened is,
0:05:37 remember those guys that said,
0:05:39 hey, we don’t want your stuff?
0:05:40 It’s crap.
0:05:42 In 2015, they all took it
0:05:43 and then they went and said,
0:05:45 we are the Spark company.
0:05:46 – Right.
0:05:47 – And they took credit for it.
0:05:50 So, you know, so at that point, we were like, wow, okay.
0:05:52 So everyone was adopting Spark
0:05:53 and actually these established vendors
0:05:55 were actually taking credit for the project
0:05:56 and we were unknown, small company
0:05:58 with one million revenue, right?
0:06:00 So it was at that point that we had to start figure out,
0:06:01 okay, how do we do this?
0:06:02 And that’s when we actually started
0:06:03 leveraging product management
0:06:06 and really listening to, you know,
0:06:09 what are the customers really need?
0:06:12 What’s, enterprise customers, what do they need to succeed?
0:06:13 And we realized actually
0:06:14 that the open source project itself
0:06:16 is far from what it is.
0:06:18 It just covers a small portion of it.
0:06:19 So we started building all the other things
0:06:22 that they need to have a managed solution for enterprises.
0:06:23 And then we started building that
0:06:24 and we kept it, a lot of that proprietary,
0:06:27 frankly speaking, starting in 2016.
0:06:30 We kind of like swung the pendulum the other way.
0:06:31 – You mentioned there a little bit
0:06:33 some of these other companies starting
0:06:34 and I think that’s an interesting space
0:06:37 that happens with these open source communities
0:06:38 is you’re not just the only company
0:06:40 contributing back to this code base.
0:06:42 How have you navigated some of those relationships
0:06:45 where you have competitors in there as well?
0:06:46 – You have this natural advantage
0:06:49 if you’re the sort of spiritual center of the project.
0:06:51 Yes, you could take one of our projects
0:06:53 and fork it and go contribute to yourself.
0:06:55 But in any conversation, right, they’re like,
0:06:58 why would I use your version instead of the Hashikur version?
0:06:59 They’re the author, they’re the creator,
0:07:01 they’re the one with the roadmap control.
0:07:03 So I think, you know, you see some of it
0:07:06 but it just falls by the wayside so rapidly, right?
0:07:09 Because it’s so hard for someone else to build a community
0:07:12 when there’s already sort of an orbit and universe
0:07:15 around sort of the spiritual center of the project.
0:07:16 So I think it’s just super tough.
0:07:18 You just don’t see a lot of that.
0:07:21 – When you say that spiritual center of the project,
0:07:22 like what gives you that?
0:07:24 Where does that come from?
0:07:25 – I think a lot of it comes from the credibility
0:07:27 of having the founders still there, right?
0:07:29 I think it’s super hard if you don’t have,
0:07:31 if not the founding team,
0:07:33 at least the core contributing team.
0:07:35 I think if you have that core development team,
0:07:37 the core founders still there,
0:07:40 it’s very much the spiritual center, right?
0:07:42 And it matters a lot in sales conversations, right?
0:07:43 When you can say, hey, we have the creator,
0:07:46 we have the top 20 contributors, right?
0:07:47 And I think that’s what gives it
0:07:49 that sort of spiritual center.
0:07:52 – And you mentioned that you had all these other competitors
0:07:54 starting around that same time.
0:07:56 How did that play out for you with the community,
0:07:58 kind of making people realize like you guys
0:07:59 were the ones to go to for that?
0:08:01 – Yeah, I mean, I have a controversial opinion
0:08:03 and it’s that most open source projects
0:08:05 are actually just led by one company.
0:08:08 Like there’s really one company that’s contributing to it.
0:08:09 And if you look at the,
0:08:12 in fact, most open source projects are super brittle.
0:08:13 If you actually look really closely,
0:08:15 you’ll notice it’s actually five people only.
0:08:18 It’s six guys or gals that are building the project.
0:08:19 And that’s just one company.
0:08:21 There are some counter examples.
0:08:23 In the case of Hadoop, there was two companies
0:08:24 and that created a huge mess.
0:08:26 So they were like fighting each other.
0:08:27 One would contribute code to it.
0:08:29 The other one would delete it.
0:08:31 And then the other one would add it back.
0:08:32 The Kubernetes maybe has two companies,
0:08:35 like maybe Google and it used to be Red Hat.
0:08:37 So it’s usually anyway,
0:08:38 just only one company or two companies.
0:08:40 So in our case, while other people started saying,
0:08:43 hey, we also offer Spark on the ground,
0:08:46 they weren’t really actually digging in.
0:08:47 They were just selling it.
0:08:48 – They were just packing it up.
0:08:48 – They were packing it up.
0:08:50 So I mean, I think in case of Hashtag Corp,
0:08:52 you guys are the only real major contributors.
0:08:54 But if you look at GitHub,
0:08:56 I’m sure you sort on contributions and commits,
0:08:58 you’ll find that the absolute majority
0:08:59 is just probably Hashtag.
0:09:00 We saw the exact same thing.
0:09:02 It’s in sort of same thing.
0:09:04 Everyone said, yeah, we’re like a lot of,
0:09:07 there’s a lot of people who look at the code
0:09:10 and maybe put a comment in or whatever,
0:09:13 but fundamentally looking at like really innovating
0:09:15 and all that really happened.
0:09:18 It’s like down to really a handful of folks who do it.
0:09:20 It’s actually very interesting.
0:09:22 Notwithstanding how large the company’s getting all that,
0:09:25 there’s always the core group that knows it.
0:09:26 Absolutely.
0:09:28 – The way we sort of articulate is almost all of our products
0:09:30 make a distinction between what we’ll call a core
0:09:32 and then sort of the extension points around the edge.
0:09:34 And if I look at the contributor graph,
0:09:35 if you look at core, it’s exactly that.
0:09:37 Yeah, it’s like five, 10, 20 people
0:09:40 working at Hashtag Corp, the 99% of the core.
0:09:42 It’s the contributions set at the edge
0:09:44 where you have these integration points where it’s,
0:09:45 you know, take a terraform, for example,
0:09:48 its integration surface is infinite, right?
0:09:50 And so at that edge is like where you go from, you know,
0:09:53 20 contributors to 2,000 contributors on the outside.
0:09:55 And I would guess for you guys is a similar thing
0:09:57 around there’s the core versus maybe some of the algorithms
0:09:59 or plugins that sit at the edge
0:10:00 where it’s easier to contribute,
0:10:02 you don’t have to be a core expert.
0:10:03 – Yeah, I agree.
0:10:05 I’m just saying in terms of like core ownership
0:10:06 of the project, it’s one company.
0:10:11 – And I would say that where open source sort of degrades
0:10:14 is the opposite of that where you have many companies
0:10:16 all arguing with each other.
0:10:18 I mean, OpenStack was a really great,
0:10:22 to me, and a great example of that where, you know,
0:10:25 like it was a jump-all on every company
0:10:27 had their own version, had their own thing,
0:10:29 and there was no consistency with it
0:10:34 because there was, in my mind, no leadership of that project.
0:10:35 – While we’re kind of on the topic
0:10:38 of participating in these communities,
0:10:40 how have you gone about managing kind of the engineering
0:10:44 function within an organization and keeping them involved
0:10:46 and how do they interact with that community?
0:10:49 – So the way I see it these days is you run a company,
0:10:50 you have an engineering department,
0:10:51 you have your product management,
0:10:52 and you’re building an awesome product
0:10:55 that’s gonna, you know, wow your customers.
0:10:56 That’s it.
0:10:59 A portion of it happens to be open source for us these days.
0:11:01 And that portion, we manage a community
0:11:03 and we give them roadmaps and, you know, do that.
0:11:08 But really, by and large, Databricks is, you know,
0:11:10 software company, I focus on building software.
0:11:13 The fact that some portions of it happen to be open source,
0:11:15 that’s just an amazing lead gen machine
0:11:18 that, you know, makes us be able to walk into accounts
0:11:20 and get ahead of the competition
0:11:21 because that’s how we know you guys.
0:11:23 You guys create its part.
0:11:25 But really, the way I look at the company is, you know,
0:11:28 build amazing software that you can monetize
0:11:29 with enterprise customers.
0:11:31 That’s the only way I look at it.
0:11:32 – Yeah, I’d guess our engineering
0:11:33 probably looks pretty similar.
0:11:36 It’s not like a open source side of engineering
0:11:37 and then an enterprise side.
0:11:39 It’s sort of one team and they just work
0:11:41 against two different roadmaps and, you know,
0:11:42 some of the features land in open source,
0:11:43 some of them land in enterprise,
0:11:45 but it’s the same engineering team,
0:11:47 same sort of product management team.
0:11:50 – Do you guys have a framework for how to think about
0:11:52 what goes in open source and not?
0:11:56 And is that consistent over time or for each release?
0:11:57 Do you debate that?
0:11:59 – Yeah, so we sort of articulated something
0:12:02 probably early 2017, which was we think about
0:12:05 sort of our split as what’s technical complexity
0:12:06 versus what’s organizational.
0:12:08 So if we’re solving something that’s fundamentally
0:12:11 caused by the organization, right?
0:12:13 You have, for example, a silo between networking
0:12:15 and security in ops, right?
0:12:16 You have a collaboration problem now.
0:12:19 Or you have a PKI team that’s distinct
0:12:20 from your security team, right?
0:12:21 – And those are close.
0:12:23 – Those are enterprise things, right?
0:12:26 Versus is it a core technical thing that we’re solving?
0:12:27 Right, like the tool fundamentally needs it?
0:12:29 That’s open source.
0:12:31 – SaaS software is very different
0:12:34 from the Red Hat support services on-prem.
0:12:37 That’s really the big difference for Databricks.
0:12:40 So kind of what we open source and what we don’t open source
0:12:42 doesn’t really matter to our customers, you know?
0:12:44 And that’s just because we’re SaaS company.
0:12:45 If we’re on-prem, it would have been a different business,
0:12:47 but because we’re in the cloud,
0:12:49 they’re renting the service from us.
0:12:51 They’re not trying to run our software.
0:12:52 They’re just renting it from us
0:12:55 and they’re paying us rent for that service, right?
0:12:56 They just want this to work in the cloud
0:12:58 and we manage it for them.
0:13:00 Where it would maybe get, iffy,
0:13:02 is if someone else decides to take all of our software
0:13:03 and offer it as well.
0:13:05 So it’s just that competitive angle.
0:13:07 Otherwise, our customers don’t care.
0:13:08 And I think the perfect example of that
0:13:10 is Amazon Web Services.
0:13:12 People use Amazon Web Services as an enterprise.
0:13:14 They have over a million customers now, right?
0:13:18 They never ask, hey, is EC2 from Amazon open source?
0:13:21 It is S3 from Amazon open source.
0:13:23 It’s Redshift from Amazon open source.
0:13:25 They’re not and no one seems to care.
0:13:28 And the truth is when you’re renting a service in the cloud,
0:13:31 you know, it’s just a different dynamic.
0:13:32 So we don’t have to worry too much about these things.
0:13:35 What becomes open source and what’s not.
0:13:36 – That’s an interesting thing because I think
0:13:39 we sit in phase one.o and you’re sort of in phase two.o,
0:13:41 which is like, we’re by and large on-premise
0:13:42 desktop software vendor.
0:13:45 – This is why we didn’t go to on-prem at Dead Rix.
0:13:48 ‘Cause we wanted to have this model.
0:13:50 We didn’t want to have to worry about this.
0:13:52 So you just rent the service from us in the cloud.
0:13:53 That’s it.
0:13:56 It’s very inspired by Amazon sort of business model,
0:13:59 rather than the red hat model, which is what exists on-prem.
0:14:00 Harder to monetize, I think.
0:14:01 – Absolutely.
0:14:02 – Well, since you’re talking about
0:14:05 kind of these cloud vendors, you mentioned AWS.
0:14:07 Peter said in this presentation, you know,
0:14:10 hey, we’ve maybe over-rotated on this threat.
0:14:12 Agree or disagree with that?
0:14:14 – I think it depends on the type of software you run.
0:14:16 And what I mean by that is, you know,
0:14:18 there’s things that are super compelling
0:14:20 for the clouds to want to run.
0:14:21 And there’s things that, you know,
0:14:22 maybe they care less so, right?
0:14:24 So I think about like HashiCorp tooling, for example.
0:14:26 You know, Terraform, for example,
0:14:28 it allows you to consume more cloud.
0:14:29 And so in that sense,
0:14:31 anything that allows you to consume more cloud,
0:14:32 wonderful, what do they care?
0:14:34 You know, they don’t care if you’re running Windows
0:14:35 or Linux or whatever you want to do.
0:14:37 Just draw more power, basically.
0:14:38 And so in that sense, it’s like,
0:14:41 is there value in them co-opting Terraform
0:14:42 or console or ball and running it themselves?
0:14:45 Like, okay, you’re gonna run two more extra nodes in Amazon.
0:14:48 Like the amount, it’s a rounding error for them.
0:14:50 So I think by the nature of being management tooling,
0:14:52 it doesn’t drive what they really monetize,
0:14:57 which is CPU hours, network I/O, disk gigs, right?
0:14:57 That’s it, right?
0:15:01 Everything else is just different packaging of that.
0:15:03 So I think, you know, it depends kind of where you sit
0:15:06 in that, you know, how aligned
0:15:08 are you to what the cloud cares about, I think.
0:15:09 – I think you guys are an interesting case,
0:15:12 usually if you sort of drive all three of those,
0:15:14 you know, probably heavy users of compute network
0:15:15 and storage.
0:15:17 So I think it’s probably a different interest.
0:15:19 – We don’t get any paid on those things.
0:15:20 We only get paid on the software.
0:15:22 We, they get a separate bill from the cloud vendors.
0:15:24 I mean, I kind of agree with Peter.
0:15:26 I see it differently from like the community
0:15:30 or the media, how they describe this problem.
0:15:32 The way I see it is basically there’s a bunch
0:15:35 of on-prem vendors, ISVs, you know, startups like us,
0:15:38 they’re running successful open source projects.
0:15:40 And then their customer base is moving to the cloud.
0:15:41 So they talk to their customers and the customer’s like,
0:15:43 hey, we’re about to move into the cloud.
0:15:45 So they say, oh, okay, we’ll also offer our service
0:15:46 in the cloud.
0:15:47 So they try to offer it in the cloud.
0:15:50 Then it turns out it’s actually extremely hard
0:15:51 to offer a cloud service.
0:15:53 And it takes you many years to get good at it.
0:15:55 In our case, we’ve only been in the cloud
0:15:56 from the very beginning.
0:15:57 We’ve never been on-prem.
0:15:58 We’re good at running cloud services.
0:16:01 There’s no problem in, you know, offering services
0:16:03 that has a lot of value to customers and to pay for it.
0:16:05 And we can run it really well in the cloud.
0:16:08 So as an example of that, we had a proprietary thing
0:16:10 called Delta, which had massive adoption
0:16:11 in the last few years.
0:16:12 It’s completely proprietary.
0:16:13 And we decided to open source it this year.
0:16:16 And we open sourced it with completely liberal open licenses
0:16:18 with no shenanigans in there.
0:16:20 You don’t need to freak out and be afraid of the cloud
0:16:22 vendors if you know how to run a cloud service.
0:16:23 But it’s hard to run a cloud service.
0:16:25 Running the SaaS model has been very hard.
0:16:26 It takes a long time to get good at it.
0:16:30 So like when we did the shift over sort of to start monetizing
0:16:33 it, you tell your engineers, hey, dude, can you, you know,
0:16:34 have this pager duty?
0:16:36 And, you know, might have to wake you up at 3 AM
0:16:37 if the thing goes down.
0:16:38 And then you’re like, what?
0:16:39 Like, I don’t want to.
0:16:40 It’s like, yeah, you’re on call.
0:16:41 This is the rotation.
0:16:43 You know, you have to wake up at 4 AM.
0:16:45 Between 4 and 6, you’re covering if there’s
0:16:47 any outrage of any security breach and so on.
0:16:49 That’s the hard thing that you have to do,
0:16:50 which the on-prem vendors don’t have to do really
0:16:51 for their service, right?
0:16:53 Because it’s the responsibility of the IT
0:16:55 department of that private data center of your customer
0:17:00 to handle outages, security breaches, SLAs, yada, yada.
0:17:03 Whereas in our case, we had to tell our engineers, hey, sorry.
0:17:06 Like, that’s why we’re getting paid the big bucks.
0:17:07 Carry this pager.
0:17:09 It feels like there’s an interesting analogy, right?
0:17:11 Which is, you know, there was an era where, as everything
0:17:13 went from hardware to software, you
0:17:14 saw the hardware companies really struggle.
0:17:17 Because fundamentally, if your core competency is hardware,
0:17:19 it doesn’t translate super well to software.
0:17:21 And I think as you go from being a software vendor to saying,
0:17:23 hey, I want to be a cloud service provider,
0:17:26 the skill set, the core competency of writing
0:17:28 and developing software actually doesn’t translate that well
0:17:29 into being good operationally.
0:17:31 It’s a completely different skill set.
0:17:33 So I think as you go through from being from a software vendor
0:17:35 into saying, I want to be a cloud SaaS vendor,
0:17:37 you might find that actually your internal core competency
0:17:38 isn’t there.
0:17:42 My sort of opinion on this, it’s very hard to do both.
0:17:45 And we’re all tempted to go, many startups
0:17:49 are tempted to do both so that we have optionality.
0:17:53 Hey, if they, well, a customer can buy in any case, right?
0:17:56 But I think you guys have pointed out a very important part
0:17:59 here, that doing both is really, really
0:18:03 hard to do as a large company, let alone as a startup.
0:18:05 The misunderstanding is in the media,
0:18:08 they say, oh, you know, these big cloud vendors,
0:18:10 they’re just taking other people’s open source software,
0:18:13 not contributing anything back and just exploiting that.
0:18:16 What they forget to tell you is they’re really, really good
0:18:18 at running that software in the cloud.
0:18:19 And almost no one else can do it.
0:18:20 It’s really hard to do.
0:18:23 And actually, they’re getting paid the big bucks for that.
0:18:24 That’s what they don’t tell you.
0:18:28 It’s because ruin is the villain and hero story.
0:18:30 Yeah, well, I think that speaks to the fact
0:18:32 that there’s a lot more nuance to the relationship
0:18:35 between an open source company and a cloud vendor
0:18:37 than maybe what we see in the media.
0:18:40 How have you or how have you seen other open source companies
0:18:43 navigate the nuance of a cloud vendor relationship
0:18:46 or other partnerships around open source?
0:18:48 I think you can partner with them.
0:18:48 They’re good partners.
0:18:51 We have extremely close partnership with Microsoft.
0:18:53 We also had good partnership with Amazon and other cloud
0:18:54 vendors.
0:18:55 You can partner with them.
0:18:57 It’s like all these workloads on the planet
0:18:59 are moving into the cloud.
0:19:01 There’s just so much for us all to eat.
0:19:03 Figure out what the cloud vendors are good at.
0:19:05 Let them add value there.
0:19:07 Look at where they’re not adding value.
0:19:08 You can go there and focus on that.
0:19:09 And then partner with them.
0:19:11 It’s a win-win situation.
0:19:12 You can do that.
0:19:15 So I do think one has to figure out how to align with them.
0:19:17 And I think one mistake a lot of big companies are doing
0:19:21 is they don’t align their Salesforce comp models
0:19:24 to be compatible with the big company’s comp model.
0:19:26 The way it works at Databricks is our customers
0:19:30 get two bills, one bill from the cloud vendor
0:19:31 and then one bill from us.
0:19:34 You get a bill for the hardware storage, the watts
0:19:36 from the cloud vendors, and you get a bill from us
0:19:37 on the software.
0:19:38 The reason that is really important
0:19:42 is that that other bill that they get from the cloud vendors
0:19:45 is actually paying the sales compensation of the cloud
0:19:47 company salespeople.
0:19:48 Hence, they like us.
0:19:51 So they partner with us in the field in every account.
0:19:53 However, if we change the pricing model
0:19:54 so that they don’t get paid, then they
0:19:56 would hate us and block us from all the accounts.
0:19:58 So I think that’s like a minor nuance
0:20:00 that some companies haven’t figured out
0:20:03 and they end up in a really fierce competitive situation
0:20:04 with the cloud vendors.
0:20:05 We don’t have that problem.
0:20:07 The cloud vendors are very friendly to us.
0:20:08 I kind of want to zoom in a little bit and put
0:20:10 Peter in the hot seat.
0:20:12 He shared kind of that four stage funnel from developer
0:20:15 community management to product management
0:20:19 to kind of the lead jet and sales dev and then those sales.
0:20:20 I’m curious, how has that played out for you?
0:20:22 Has that held true?
0:20:25 What parts didn’t necessarily hold true?
0:20:27 I think it goes through those exact phases
0:20:29 that sort of Peter laid out.
0:20:31 I think then those phases actually map pretty well
0:20:32 into the funnel as well, which is
0:20:34 that early phase of product market fit.
0:20:37 It’s a lot about developer advocacy, building the community,
0:20:40 things like that as you go into a repeatable sales cycle.
0:20:42 Well, that only works if you have a tight fit
0:20:45 between product management and product marketing in terms
0:20:47 of, OK, great, we need the futures customers are asking for.
0:20:48 And then we should tell them about that.
0:20:51 So there has to be sort of an integration there, that phase.
0:20:53 And then I think as you start going to scale,
0:20:57 you get to those phases three and four in the funnel
0:20:58 to really be able to amplify that message
0:21:01 and bring in the cloud partners as part of your channel.
0:21:04 I think the only thing that maybe we experience slightly
0:21:06 differently actually would be on that fourth phase.
0:21:10 I think you laid out sort of start with SaaS, self-service,
0:21:12 then go departmental, then go enterprise.
0:21:14 For us, it’s almost been exactly the opposite.
0:21:16 And I think it, again, I think it’s
0:21:19 because we were sort of a phase 1.0 versus a phase 2.0.
0:21:22 I also think it depends on what product you’re actually selling.
0:21:25 You may not have– your value may not accrete
0:21:28 to an individual user, in which case,
0:21:30 I just want to make it clear, you may not have that line
0:21:33 because the product doesn’t support that line.
0:21:36 And so then you may start– many companies
0:21:38 start with field sales first.
0:21:42 So it was an example of how to layer it up as opposed to that’s
0:21:44 every company ought to be that.
0:21:46 But I think it goes to your point that you made of having
0:21:48 a framework in terms of what do you decide
0:21:49 goes open versus enterprise.
0:21:51 Because as I describe our framework,
0:21:53 our dividing line is what accretes to an individual.
0:21:54 Well, that’s open source.
0:21:57 And what accretes to an enterprise.
0:22:00 And so because of the divide we use in product management,
0:22:03 it’s very hard for us to have a sort of a self-service model.
0:22:04 There is no self.
0:22:05 That’s given away for free.
0:22:05 We’ve given away.
0:22:07 So you could look at that bottom curve
0:22:10 and say this is the open source line.
0:22:13 It’s not dollars per customer, whatever the y-axis would be.
0:22:16 And then you’d build your revenue model on top of that.
0:22:16 Exactly.
0:22:18 Exactly.
0:22:20 So I think for us, once you sort of acknowledge, hey,
0:22:22 that’s the divide, it makes sense
0:22:25 to start on the enterprise side.
0:22:29 In an article that Peter wrote in 2013, I think,
0:22:31 that was same time as we started Databricks,
0:22:34 in which he really accentuated the big difference
0:22:38 between this Red Hat model and the SAS model.
0:22:39 And it really resonated with us.
0:22:42 And we really thought, our view was really
0:22:44 like this on-prem Red Hat open source model is dead.
0:22:45 It’s bad.
0:22:46 We looked down on it.
0:22:48 We didn’t want to have anything to do with it.
0:22:52 And we really saw this SAS model as extremely powerful.
0:22:55 And it’s pretty prescient, because 2012, ’13,
0:22:59 and you know, AWS was not the huge phenomenon it is today.
0:23:01 Now it’s absolutely exploded.
0:23:03 It’s like taking over the whole planet.
0:23:05 So I agree with it, but I think the thing
0:23:08 that I would really emphasize is, for me,
0:23:11 the difference between SAS and non-SAS
0:23:12 makes a huge difference.
0:23:15 I think churn is higher if you’re not SAS.
0:23:17 I think net expansion rates are lower.
0:23:19 I think everything is worse if you’re not SAS.
0:23:21 Because what can happen– I’ll give you an example.
0:23:23 I’m not going to name the names.
0:23:25 But basically, there was an open source event.
0:23:27 This is one of those cases where the open source software
0:23:30 actually had two companies fighting over it.
0:23:32 They had contributors.
0:23:36 One would run it on-prem and give support and services.
0:23:38 And then what would happen is, after a couple of years,
0:23:41 the customer would keep the software, because it’s free,
0:23:43 but then migrate and get support from the other cheaper
0:23:46 vendor, but keep the same software.
0:23:48 That you can’t do with SAS services, right?
0:23:49 If you want to move away– if you
0:23:51 don’t want to have a contract with us,
0:23:52 we would shut down the service.
0:23:53 You can’t use it anymore.
0:23:55 And so hence– and then what they would do
0:23:56 is, then they get the cheaper vendor.
0:23:57 And then after a couple of years,
0:23:59 they would not even renew with the cheaper vendor,
0:24:00 because they would say, I actually
0:24:03 hired your support guy into my company,
0:24:05 so I don’t need to buy any support from you anymore.
0:24:07 And I’m still going to keep the software.
0:24:08 With SAS, you can’t do that.
0:24:10 So churn ends up being lower.
0:24:11 Expansion rates are higher.
0:24:12 Everything is just better.
0:24:15 It’s super interesting as I think about it.
0:24:17 Well, there’s an exact example of commodity
0:24:21 when I can go swat like all this software has been developed.
0:24:25 And yet, it’s total commodity that I can basically
0:24:28 go from one vendor to another to hiring a person
0:24:30 to actually support this thing.
0:24:32 And that’s what was happening during that era
0:24:35 when I mentioned the difference between open source
0:24:37 valuations versus proprietary.
0:24:41 This was exactly that characteristic that propagated
0:24:42 that particular dynamic.
0:24:45 So open source– the software itself
0:24:47 has zero intrinsic value.
0:24:48 Anyone can download it.
0:24:50 So if these companies were selling,
0:24:53 their value was support and services,
0:24:55 which quickly gets commoditized.
0:24:56 And it goes down to who can do that most efficiently
0:24:59 where on the planet, and they can manage that P&L.
0:25:01 And then you have a company like Red Hat, which
0:25:03 has sort of a scale advantage.
0:25:04 Exactly.
0:25:05 That’s their value at.
0:25:07 It’s a scale advantage to do exactly that.
0:25:10 So for people who aren’t familiar with that article
0:25:13 from back in 2013, could you give kind of a quick summary
0:25:13 of what your–
0:25:16 I mean, the title of this blog was
0:25:19 “Why There Will Never Be Another Red Hat,” was the title.
0:25:23 And I made the argument to Ali’s point
0:25:28 that the support model was pretty broken at the time.
0:25:31 And thinking about going to a service–
0:25:35 to a open source service model, hosted service,
0:25:40 was a way to really uncover and accentuate
0:25:43 the value of the product that you’re bringing to the market.
0:25:44 So that was–
0:25:47 I mean, there’s basically all the points that we argued here.
0:25:51 Red Hat had the scale and capacity to go and do that.
0:25:52 Don’t get me wrong.
0:25:53 Red Hat is a great company.
0:25:55 It’s just very hard for a startup
0:25:59 to go replicate what they have done because their value at
0:26:00 is the scale.
0:26:03 And the things that startups don’t do very well
0:26:06 is scale, because you don’t have the money to go and do that.
0:26:09 So it’s counter– it’s sort of counterproductive
0:26:10 on that dimension.
0:26:12 Compete with Procter and Gamble and Distributor.
0:26:14 Exactly, like, it doesn’t– you can’t do that.
0:26:16 You can’t do that as a startup.
0:26:18 SAS also is killing that business model even more.
0:26:19 Because the–
0:26:20 Totally.
0:26:23 The secret sauce– like, the thing that’s weird about Red Hat
0:26:25 is that of all the open source companies that exist,
0:26:27 that for some reason, that people can analyze and debate
0:26:30 forever, they ended up being a monopoly.
0:26:31 Exactly, yes.
0:26:33 Without really any fierce competition,
0:26:35 which is generally not true about the open source software.
0:26:37 You end up– because the software has zero intrinsic value,
0:26:39 you end up getting lots of competitors, which
0:26:42 commoditizes the price and brings it down.
0:26:45 So but with the cloud vendors now,
0:26:48 you’re much more unlikely to have a monopoly like they had.
0:26:50 Because if you offer just free software
0:26:52 that you’re just distributing, they can also pick it up
0:26:53 and offer it.
0:26:55 So it’s very unlikely there will ever
0:26:58 be another Red Hat because of that.
0:27:01 So what practical advice might you give?
0:27:05 Having done on-prem, I’d say skip on-prem.
0:27:09 Go straight to SAS, save yourself.
0:27:12 Yeah, I mean, I think the advantages
0:27:16 that Ali has talked about around the SAS model are very true.
0:27:19 And I think, to Peter’s point about changing competence,
0:27:22 being very hard, if you go down the road of building software
0:27:25 competence and then realize you want to switch to SAS
0:27:28 competence, very much the bucket we’re in, to be honest,
0:27:31 you realize it’s a hard shift, right?
0:27:32 It is a different skill set.
0:27:35 It’s a different set of practices.
0:27:39 And so the earlier you can do that, ideally at inception,
0:27:41 the easier your life will be, right?
0:27:43 The further you get down one road,
0:27:46 the harder and more painful that shift really is.
0:27:49 Yeah, for me, I would say SAS is obviously the one.
0:27:50 Definitely just start with the SAS.
0:27:51 By the way, Wall Street likes SAS
0:27:53 and gives you higher multiples.
0:27:54 So you get a higher valuation.
0:27:56 So there’s that as well.
0:28:01 But ignoring that aspect, I think what is it you want to do?
0:28:04 What do you want your company to do?
0:28:05 Which space do you pick?
0:28:07 And the way I think about it is you
0:28:09 have to expect these three cloud vendors, Amazon,
0:28:12 Microsoft, Google, they each have roughly $100 billion
0:28:14 of cash sitting around.
0:28:16 And they actually have a printing press
0:28:18 that’s not the cloud business, right?
0:28:20 They either have an ad business as a printing press,
0:28:23 or they have a Windows or something server business
0:28:24 that is a printing press.
0:28:27 Or they have a retail of everything on the planet
0:28:29 as their printing press.
0:28:31 So you should just assume that they’re
0:28:34 going to get really, really good at the lower levels of the stack.
0:28:36 And the lower levels of the stack,
0:28:37 there aren’t that many things.
0:28:40 There’s like machines, there is storage,
0:28:42 there’s networking, there’s some databases.
0:28:43 That’s it.
0:28:45 You move up the stack a little bit.
0:28:46 You start having much more.
0:28:49 And as you move up the stack, it gets more and more verticalized.
0:28:51 And it actually becomes a lot of different things,
0:28:53 a lot of different products.
0:28:54 The cloud vendors can’t win all of those.
0:28:57 They can dominate and crush and just completely
0:28:58 own the bottom layers.
0:29:00 The higher up you go, there’s going
0:29:01 to be a lot of vendors.
0:29:03 Otherwise, if I’m wrong about this statement,
0:29:06 there will only be three companies on the whole planet
0:29:06 in software.
0:29:08 That’s very unlikely.
0:29:10 So pick a space higher up in the stack.
0:29:11 Competition will be much less.
0:29:13 It’s going to get much more verticalized.
0:29:18 And do the sass and you’ll probably be very successful.
0:29:20 I did want to touch on briefly your backgrounds
0:29:24 and the origin of open source, as well as your own start
0:29:27 in academia.
0:29:29 Those two have been really tightly linked.
0:29:31 Now we see with commercializing open source,
0:29:33 what does that relationship look like?
0:29:36 How are academia and open source connected in your mind?
0:29:38 I mean, I think one of the interesting things
0:29:40 about academia is going back forever.
0:29:43 It’s always had this ethos of it’s free software.
0:29:45 It’s sort of publicly funded.
0:29:47 You get in government grants to pay for the stuff.
0:29:49 So you’re sort of naturally giving it out
0:29:51 or the code is there for it so other people can reproduce
0:29:53 the work and extend it and add on it.
0:29:56 So I think if you spend time in that,
0:29:57 you kind of soak in that ethos.
0:30:00 And that’s sort of the notion that the software is free
0:30:02 and other people collaborate and extend and remix.
0:30:04 That comes normal, like it doesn’t seem weird.
0:30:05 I think the other nice thing is,
0:30:07 especially infrastructure software,
0:30:09 this stuff isn’t like, hey, I’m gonna bang it out
0:30:11 over a weekend and launch my cool new app
0:30:12 and put it on Hacker News, right?
0:30:15 It’s like spend years building this stuff, right?
0:30:19 And really getting it to a point of broad usability,
0:30:21 scalability, et cetera.
0:30:23 And so where are the environments where you can spend years
0:30:26 and years working on a thing, right?
0:30:27 You kind of have that luxury in academia
0:30:29 to be able to do that.
0:30:31 And so it’s like, the first few years of development
0:30:34 take place in a university,
0:30:37 and then it becomes sort of an industrial project from there.
0:30:39 But it would be very hard to bootstrap some of the stuff
0:30:41 from zero in an industrial setting.
0:30:46 – Yeah, I actually think academia is misconfigured.
0:30:48 I mean, I’m actually a young professor,
0:30:50 so I’m kind of, I wear that hat too.
0:30:53 But I think in the systems research field,
0:30:54 it’s misconfigured.
0:30:56 I think there’s a huge opportunity for academia
0:30:59 to come in and completely disrupt the software scene.
0:31:01 But the way it’s configured right now
0:31:03 is as academics,
0:31:04 we get incentivized on publishing papers
0:31:06 in the top conferences,
0:31:09 and that’s what we focus on typically.
0:31:11 If the focus was on push the boundaries
0:31:13 of what kind of software you can build
0:31:14 and disrupt the world with it,
0:31:16 you know, all these universities,
0:31:18 with all these students that have five years
0:31:19 to sit and create the open source project,
0:31:21 could like completely disrupt
0:31:22 how software is done on the planet.
0:31:24 It’s a gigantic opportunity for any university
0:31:26 to sort of take on.
0:31:28 Berkeley actually did it, I mean, to our credit.
0:31:30 I mean, we kind of pushed forward on that
0:31:32 with some of these labs that we had,
0:31:34 like Rad Lab, Amplab, Rise Lab,
0:31:35 but it’s just scratching the surface.
0:31:38 I think there is a scenario,
0:31:41 if all the universities kind of figured this out,
0:31:41 which they haven’t,
0:31:44 that they could like completely start owning the software space.
0:31:47 It could become like the future of how software is developed.
0:31:48 It’s like basically open source projects
0:31:50 that the different universities are leading.
0:31:52 That’s not happening right now.
0:31:55 – All right, so kind of time for like a last question here.
0:31:56 What area of open source right now
0:31:58 is the most fascinating to you?
0:32:01 Where do you think the most interesting things are happening?
0:32:03 – I mean, so I’m going to kind of not answer that
0:32:08 by saying data and how, you know,
0:32:11 the value of data and the ecosystem around it,
0:32:12 how you can buy it, how you can sell it,
0:32:13 how you can leverage it,
0:32:16 and the models that interpret that data.
0:32:18 That’s, you know, we think of software,
0:32:20 we used to think of a hardware software and so on.
0:32:22 I think data is the next thing.
0:32:24 And, you know, I mean, a lot of people have said it,
0:32:25 it’s the new oil or, you know,
0:32:27 but we’re just in the beginning of that.
0:32:29 It was the early innings of how that’s,
0:32:31 there’s going to be an economy forming around data itself.
0:32:32 I mean, it’s already kind of happening.
0:32:33 So that’s, I think, fascinating.
0:32:37 So I think that’s like the next third wave
0:32:39 of interesting sort of market trend
0:32:41 that you’re going to see in the software space.
0:32:43 Like the software will kind of be free,
0:32:45 but who has the data will actually,
0:32:46 and the models around it are going to be the,
0:32:49 that’s going to be the competitive edge.
0:32:51 – I’ve always personally been a bit of a systems guy.
0:32:53 So I love following sort of like what’s happening
0:32:54 in database research land.
0:32:57 And to me, I think, you know, it’s been interesting
0:33:00 ’cause it’s like it already BMS has sort of ruled the world
0:33:01 for decades and decades.
0:33:03 And I think what’s finally happening
0:33:05 is you’re either seeing shifts because of scale, right?
0:33:06 You’re no longer saying, okay,
0:33:08 I can fit all the data on one machine.
0:33:09 I need to, you know, fundamentally go
0:33:11 to a clustered architecture.
0:33:14 Or now as we think about sort of IoT edge,
0:33:15 fog computing, whatever you want to call it,
0:33:18 there’s these notion sort of hierarchical levels of data
0:33:19 where you might have high bandwidth,
0:33:21 high throughput, you know, cloud data centers
0:33:23 and then, you know, go out to sort of an Akamai
0:33:26 or Fastly Pop and then go out to someone’s house
0:33:27 and then go out to your phone.
0:33:29 And so how do you actually design systems
0:33:32 that can reconcile and handle data sort of planetary scale,
0:33:34 much higher volume, much lower latency
0:33:39 and reconcile and do it all in a comprehensible way.
0:33:41 So I think that’s a sort of a fascinating space
0:33:44 in terms of, you know, is RDBMS finally being challenged
0:33:48 as sort of supreme when it comes to data management?
0:33:52 – I think one of the fascinating elements of open source
0:33:54 is the origination of projects now.
0:33:58 Like these stats of Google doing 2,000 projects
0:34:01 and all of that, maybe that’s part of the answer of,
0:34:05 you know, I love your academic sort of comment.
0:34:06 It may not happen there, but the fact
0:34:09 that all of these companies are really built
0:34:12 on large backbones of open source
0:34:14 and are releasing these projects into the market
0:34:17 where there’s, you know, not a lot of strategic value
0:34:19 to them, I think it’s in quotes,
0:34:23 unlocking a huge number of opportunities
0:34:27 that I think will very much dominate the landscape
0:34:30 as we, you know, sort of roll into the future.
0:34:31 That’s really interesting.
0:34:34 – I want to thank you, thank the panelists
0:34:36 and a huge thanks, Peter, to you
0:34:37 for sharing all this information.
0:34:39 (audience applauding)
0:34:41 – You just heard Olly Goetze of Databricks,
0:34:43 Armand Adger of Hashicorp
0:34:45 and A16Z general partner, Peter Levine,
0:34:48 with me, Das Rush, moderating the discussion.
0:34:49 Thanks again for listening.
0:34:52 And if you want to learn more about open source,
0:34:53 the importance of commercializing it
0:34:55 and what it takes to turn an open source project
0:34:57 into a business, you can download
0:34:59 and/or watch Peter’s presentation
0:35:04 and other open source materials at a16z.com/opensource.
0:00:05 I’m Doss Rush, our enterprise technology editor,
0:00:07 and in this podcast, I moderate a panel discussion
0:00:09 on some of the most heated topics in open source
0:00:12 with two of the leading founders of open source companies,
0:00:15 Armand Dodger, co-founder and CTO of HashiCorp,
0:00:18 which does open source tools for managing multi-cloud,
0:00:20 and Oli Goetze, the CEO of Databricks,
0:00:22 a SaaS offering of Apache Spark.
0:00:23 Joining them in conversation
0:00:26 is A16Z general partner, Peter Levine,
0:00:28 who’s invested and been on the board
0:00:29 of numerous open source companies,
0:00:31 such as GitHub and Netlify.
0:00:33 It’s a great discussion and it takes on everything
0:00:34 from making money on open source,
0:00:37 while managing community, to the nuance of partnering
0:00:40 and sometimes competing with big cloud vendors.
0:00:42 Just to note that this was recorded at a live event,
0:00:44 so there are some audio issues.
0:00:47 The first voice that you’re gonna hear in this is Armand’s,
0:00:49 then Oli, and a few minutes into the conversation,
0:00:50 Peter joins them.
0:00:52 Finally, please note that the content here
0:00:54 is for informational purposes only.
0:00:56 Should not be taken as legal, business tax,
0:00:58 or investment advice, or be used to evaluate
0:01:00 any investment or security.
0:01:02 And it is not directed at any investors
0:01:05 or potential investors in any A16Z fund.
0:01:09 For more details, please see a16z.com/disclosures.
0:01:12 Throughout the history of open source,
0:01:14 talking about making money on open source
0:01:16 has been a pretty controversial topic
0:01:18 with a lot of different views.
0:01:20 So, I’m curious, Oli and Armand,
0:01:23 how have you thought about commercializing open source
0:01:26 and why did you choose to turn a project into a business?
0:01:29 – For us, it didn’t start necessarily
0:01:31 as thinking about turning the open source
0:01:32 into the business.
0:01:36 It was more around recognizing that there’s a clear market gap
0:01:37 in terms of, in our case, sort of DevOps tooling,
0:01:39 how do we actually provision things
0:01:41 and sort of cloud infrastructure.
0:01:43 And then realizing it’s very hard
0:01:46 to become a large sustainable project
0:01:48 if you have negative cash flow forever.
0:01:50 And if you’re at a university and great,
0:01:51 you have grants and things like that
0:01:52 that can sort of fund it,
0:01:53 or it’s a little hobby project
0:01:56 and it’s two, three people doing it on a weekend, fine.
0:01:58 But if you’re solving a large enough problem,
0:02:00 you eventually need teams of dozens,
0:02:03 hundreds, thousands to work on that problem.
0:02:04 You kind of need a business.
0:02:07 There has to be sort of a top line connected
0:02:08 to the bottom line,
0:02:09 otherwise it doesn’t make a lot of sense.
0:02:12 And so I think for us, it started pretty pragmatically
0:02:14 in terms of, hey, we are passionate about the technology,
0:02:15 passionate about the space.
0:02:18 We want this to be viable long term.
0:02:20 Well, the only way for it to be viable long term
0:02:21 was if you make money.
0:02:24 – I’m just going to be honest.
0:02:25 We were academics.
0:02:27 We just wanted to have impact
0:02:29 and we wanted to publish papers.
0:02:32 And the software we built, people didn’t want to adopt it.
0:02:34 So we went to all the companies that were out there,
0:02:36 went to cloud error, all these guys,
0:02:37 and said, please take the software, take it with you,
0:02:39 take credit for it.
0:02:41 And they all just refused.
0:02:44 They said, this is just an academic mambo jumbo,
0:02:45 not interesting.
0:02:47 What if these PhD students just leave?
0:02:51 And then this is enterprise software we’re selling.
0:02:52 So they just rejected us.
0:02:54 So in 2013, we were kind of frustrated and we said,
0:02:56 if we want to actually succeed,
0:02:57 the only way we can get these projects off the ground
0:02:59 is if we actually start something ourselves.
0:03:00 They wasted so much of our time.
0:03:02 We put interns into these companies.
0:03:03 – What do you really think?
0:03:05 (laughing)
0:03:06 – We were sending interns into these companies,
0:03:07 hoping that they would adopt our software.
0:03:08 It never happened.
0:03:10 Since we were frustrated in 2013,
0:03:13 we just said, let’s create a company ourselves to do it.
0:03:15 I don’t think revenue was sort of top of mind for us.
0:03:16 The first two years of Databricks,
0:03:19 our only goal was how can we make spark
0:03:21 in the software that we have take over the world.
0:03:23 – Let’s talk about the flip side.
0:03:26 You choose to go out and then have a commercial venture.
0:03:28 How have you gone about managing your community
0:03:29 and communicating that with them,
0:03:32 kind of keeping their support for what you’re doing?
0:03:35 – I think for us, it goes back to Peter’s point around
0:03:38 having a clear product management framework
0:03:40 that you can articulate where your community
0:03:42 doesn’t feel like you’re just randomly picking
0:03:44 what goes one way or the other.
0:03:47 And I think for us, it was really trying to draw that line
0:03:49 and saying, okay, great, the things that are truly
0:03:52 organizational complexity problems, right?
0:03:54 You need role-based access control.
0:03:55 You want audit logging.
0:03:58 You need PCI ISO SOC compliance, things like that.
0:04:00 You’re like, okay, great, if you have those problems,
0:04:02 you’re probably a global 10,000 business.
0:04:04 You’re not a SMB hobbyist.
0:04:06 And I think we drew that line and articulated to the community
0:04:08 and said, hey, things in this bucket, great.
0:04:10 They go into enterprise and the people who are gonna pay
0:04:12 for that are the people who have that problem.
0:04:15 You as a hobbyist don’t care about, you know,
0:04:17 hardware security device integration
0:04:18 for your compliance, right?
0:04:19 Like it’s not a problem you have.
0:04:21 And so I think if you articulate it clearly that way
0:04:23 and have the discipline to stick to it,
0:04:24 then the community doesn’t feel like
0:04:26 they’re sort of randomly being jerked around.
0:04:28 And they don’t feel like they’re losing value
0:04:30 ’cause those aren’t problems they have.
0:04:31 – Yeah.
0:04:32 For us, we’re different.
0:04:33 I mean, these guys are much smarter than us.
0:04:35 We didn’t think these things through.
0:04:37 I mean, we just wanted Spark to take over the world
0:04:38 as open source projects.
0:04:40 So we only had one roadmap.
0:04:41 It was the open source roadmap.
0:04:42 We wanted to sort of, and we were frustrated
0:04:43 that no one would adopt it.
0:04:44 And there was a lot of fun in the market
0:04:48 that the technology we built won’t work for this
0:04:49 and won’t work for that.
0:04:50 It only works if it’s in memory,
0:04:52 not if it’s on this, this kind of thing.
0:04:53 So we were super frustrated.
0:04:56 So the first three years, our only goal was get adoption.
0:04:57 We didn’t care about any revenue.
0:04:58 I mean, three, four years in,
0:04:59 we only had one million revenue.
0:05:01 So we only had one roadmap.
0:05:03 We only managed the community.
0:05:05 And then in 2015, it just exploded.
0:05:06 Like it exploded.
0:05:08 Like, I mean, you saw some of the curves.
0:05:10 It’s like, you know, thousands of developers
0:05:12 started contributing code to it.
0:05:14 And at that point, that’s when things started,
0:05:15 the community was so big
0:05:17 and this thing became way bigger than we are.
0:05:19 Like, we were like, unknown.
0:05:21 And then the project was huge.
0:05:22 So at that point, we started thinking through,
0:05:24 hey, how do we actually monetize this thing?
0:05:26 – Can you talk a little bit more about,
0:05:27 you know, that moment in 2015
0:05:29 when you have this huge community
0:05:30 and now you’re starting to think
0:05:32 about the secondary roadmap
0:05:33 where you’re commercializing something.
0:05:35 How did the conversations change?
0:05:36 – Yeah, well, so what happened is,
0:05:37 remember those guys that said,
0:05:39 hey, we don’t want your stuff?
0:05:40 It’s crap.
0:05:42 In 2015, they all took it
0:05:43 and then they went and said,
0:05:45 we are the Spark company.
0:05:46 – Right.
0:05:47 – And they took credit for it.
0:05:50 So, you know, so at that point, we were like, wow, okay.
0:05:52 So everyone was adopting Spark
0:05:53 and actually these established vendors
0:05:55 were actually taking credit for the project
0:05:56 and we were unknown, small company
0:05:58 with one million revenue, right?
0:06:00 So it was at that point that we had to start figure out,
0:06:01 okay, how do we do this?
0:06:02 And that’s when we actually started
0:06:03 leveraging product management
0:06:06 and really listening to, you know,
0:06:09 what are the customers really need?
0:06:12 What’s, enterprise customers, what do they need to succeed?
0:06:13 And we realized actually
0:06:14 that the open source project itself
0:06:16 is far from what it is.
0:06:18 It just covers a small portion of it.
0:06:19 So we started building all the other things
0:06:22 that they need to have a managed solution for enterprises.
0:06:23 And then we started building that
0:06:24 and we kept it, a lot of that proprietary,
0:06:27 frankly speaking, starting in 2016.
0:06:30 We kind of like swung the pendulum the other way.
0:06:31 – You mentioned there a little bit
0:06:33 some of these other companies starting
0:06:34 and I think that’s an interesting space
0:06:37 that happens with these open source communities
0:06:38 is you’re not just the only company
0:06:40 contributing back to this code base.
0:06:42 How have you navigated some of those relationships
0:06:45 where you have competitors in there as well?
0:06:46 – You have this natural advantage
0:06:49 if you’re the sort of spiritual center of the project.
0:06:51 Yes, you could take one of our projects
0:06:53 and fork it and go contribute to yourself.
0:06:55 But in any conversation, right, they’re like,
0:06:58 why would I use your version instead of the Hashikur version?
0:06:59 They’re the author, they’re the creator,
0:07:01 they’re the one with the roadmap control.
0:07:03 So I think, you know, you see some of it
0:07:06 but it just falls by the wayside so rapidly, right?
0:07:09 Because it’s so hard for someone else to build a community
0:07:12 when there’s already sort of an orbit and universe
0:07:15 around sort of the spiritual center of the project.
0:07:16 So I think it’s just super tough.
0:07:18 You just don’t see a lot of that.
0:07:21 – When you say that spiritual center of the project,
0:07:22 like what gives you that?
0:07:24 Where does that come from?
0:07:25 – I think a lot of it comes from the credibility
0:07:27 of having the founders still there, right?
0:07:29 I think it’s super hard if you don’t have,
0:07:31 if not the founding team,
0:07:33 at least the core contributing team.
0:07:35 I think if you have that core development team,
0:07:37 the core founders still there,
0:07:40 it’s very much the spiritual center, right?
0:07:42 And it matters a lot in sales conversations, right?
0:07:43 When you can say, hey, we have the creator,
0:07:46 we have the top 20 contributors, right?
0:07:47 And I think that’s what gives it
0:07:49 that sort of spiritual center.
0:07:52 – And you mentioned that you had all these other competitors
0:07:54 starting around that same time.
0:07:56 How did that play out for you with the community,
0:07:58 kind of making people realize like you guys
0:07:59 were the ones to go to for that?
0:08:01 – Yeah, I mean, I have a controversial opinion
0:08:03 and it’s that most open source projects
0:08:05 are actually just led by one company.
0:08:08 Like there’s really one company that’s contributing to it.
0:08:09 And if you look at the,
0:08:12 in fact, most open source projects are super brittle.
0:08:13 If you actually look really closely,
0:08:15 you’ll notice it’s actually five people only.
0:08:18 It’s six guys or gals that are building the project.
0:08:19 And that’s just one company.
0:08:21 There are some counter examples.
0:08:23 In the case of Hadoop, there was two companies
0:08:24 and that created a huge mess.
0:08:26 So they were like fighting each other.
0:08:27 One would contribute code to it.
0:08:29 The other one would delete it.
0:08:31 And then the other one would add it back.
0:08:32 The Kubernetes maybe has two companies,
0:08:35 like maybe Google and it used to be Red Hat.
0:08:37 So it’s usually anyway,
0:08:38 just only one company or two companies.
0:08:40 So in our case, while other people started saying,
0:08:43 hey, we also offer Spark on the ground,
0:08:46 they weren’t really actually digging in.
0:08:47 They were just selling it.
0:08:48 – They were just packing it up.
0:08:48 – They were packing it up.
0:08:50 So I mean, I think in case of Hashtag Corp,
0:08:52 you guys are the only real major contributors.
0:08:54 But if you look at GitHub,
0:08:56 I’m sure you sort on contributions and commits,
0:08:58 you’ll find that the absolute majority
0:08:59 is just probably Hashtag.
0:09:00 We saw the exact same thing.
0:09:02 It’s in sort of same thing.
0:09:04 Everyone said, yeah, we’re like a lot of,
0:09:07 there’s a lot of people who look at the code
0:09:10 and maybe put a comment in or whatever,
0:09:13 but fundamentally looking at like really innovating
0:09:15 and all that really happened.
0:09:18 It’s like down to really a handful of folks who do it.
0:09:20 It’s actually very interesting.
0:09:22 Notwithstanding how large the company’s getting all that,
0:09:25 there’s always the core group that knows it.
0:09:26 Absolutely.
0:09:28 – The way we sort of articulate is almost all of our products
0:09:30 make a distinction between what we’ll call a core
0:09:32 and then sort of the extension points around the edge.
0:09:34 And if I look at the contributor graph,
0:09:35 if you look at core, it’s exactly that.
0:09:37 Yeah, it’s like five, 10, 20 people
0:09:40 working at Hashtag Corp, the 99% of the core.
0:09:42 It’s the contributions set at the edge
0:09:44 where you have these integration points where it’s,
0:09:45 you know, take a terraform, for example,
0:09:48 its integration surface is infinite, right?
0:09:50 And so at that edge is like where you go from, you know,
0:09:53 20 contributors to 2,000 contributors on the outside.
0:09:55 And I would guess for you guys is a similar thing
0:09:57 around there’s the core versus maybe some of the algorithms
0:09:59 or plugins that sit at the edge
0:10:00 where it’s easier to contribute,
0:10:02 you don’t have to be a core expert.
0:10:03 – Yeah, I agree.
0:10:05 I’m just saying in terms of like core ownership
0:10:06 of the project, it’s one company.
0:10:11 – And I would say that where open source sort of degrades
0:10:14 is the opposite of that where you have many companies
0:10:16 all arguing with each other.
0:10:18 I mean, OpenStack was a really great,
0:10:22 to me, and a great example of that where, you know,
0:10:25 like it was a jump-all on every company
0:10:27 had their own version, had their own thing,
0:10:29 and there was no consistency with it
0:10:34 because there was, in my mind, no leadership of that project.
0:10:35 – While we’re kind of on the topic
0:10:38 of participating in these communities,
0:10:40 how have you gone about managing kind of the engineering
0:10:44 function within an organization and keeping them involved
0:10:46 and how do they interact with that community?
0:10:49 – So the way I see it these days is you run a company,
0:10:50 you have an engineering department,
0:10:51 you have your product management,
0:10:52 and you’re building an awesome product
0:10:55 that’s gonna, you know, wow your customers.
0:10:56 That’s it.
0:10:59 A portion of it happens to be open source for us these days.
0:11:01 And that portion, we manage a community
0:11:03 and we give them roadmaps and, you know, do that.
0:11:08 But really, by and large, Databricks is, you know,
0:11:10 software company, I focus on building software.
0:11:13 The fact that some portions of it happen to be open source,
0:11:15 that’s just an amazing lead gen machine
0:11:18 that, you know, makes us be able to walk into accounts
0:11:20 and get ahead of the competition
0:11:21 because that’s how we know you guys.
0:11:23 You guys create its part.
0:11:25 But really, the way I look at the company is, you know,
0:11:28 build amazing software that you can monetize
0:11:29 with enterprise customers.
0:11:31 That’s the only way I look at it.
0:11:32 – Yeah, I’d guess our engineering
0:11:33 probably looks pretty similar.
0:11:36 It’s not like a open source side of engineering
0:11:37 and then an enterprise side.
0:11:39 It’s sort of one team and they just work
0:11:41 against two different roadmaps and, you know,
0:11:42 some of the features land in open source,
0:11:43 some of them land in enterprise,
0:11:45 but it’s the same engineering team,
0:11:47 same sort of product management team.
0:11:50 – Do you guys have a framework for how to think about
0:11:52 what goes in open source and not?
0:11:56 And is that consistent over time or for each release?
0:11:57 Do you debate that?
0:11:59 – Yeah, so we sort of articulated something
0:12:02 probably early 2017, which was we think about
0:12:05 sort of our split as what’s technical complexity
0:12:06 versus what’s organizational.
0:12:08 So if we’re solving something that’s fundamentally
0:12:11 caused by the organization, right?
0:12:13 You have, for example, a silo between networking
0:12:15 and security in ops, right?
0:12:16 You have a collaboration problem now.
0:12:19 Or you have a PKI team that’s distinct
0:12:20 from your security team, right?
0:12:21 – And those are close.
0:12:23 – Those are enterprise things, right?
0:12:26 Versus is it a core technical thing that we’re solving?
0:12:27 Right, like the tool fundamentally needs it?
0:12:29 That’s open source.
0:12:31 – SaaS software is very different
0:12:34 from the Red Hat support services on-prem.
0:12:37 That’s really the big difference for Databricks.
0:12:40 So kind of what we open source and what we don’t open source
0:12:42 doesn’t really matter to our customers, you know?
0:12:44 And that’s just because we’re SaaS company.
0:12:45 If we’re on-prem, it would have been a different business,
0:12:47 but because we’re in the cloud,
0:12:49 they’re renting the service from us.
0:12:51 They’re not trying to run our software.
0:12:52 They’re just renting it from us
0:12:55 and they’re paying us rent for that service, right?
0:12:56 They just want this to work in the cloud
0:12:58 and we manage it for them.
0:13:00 Where it would maybe get, iffy,
0:13:02 is if someone else decides to take all of our software
0:13:03 and offer it as well.
0:13:05 So it’s just that competitive angle.
0:13:07 Otherwise, our customers don’t care.
0:13:08 And I think the perfect example of that
0:13:10 is Amazon Web Services.
0:13:12 People use Amazon Web Services as an enterprise.
0:13:14 They have over a million customers now, right?
0:13:18 They never ask, hey, is EC2 from Amazon open source?
0:13:21 It is S3 from Amazon open source.
0:13:23 It’s Redshift from Amazon open source.
0:13:25 They’re not and no one seems to care.
0:13:28 And the truth is when you’re renting a service in the cloud,
0:13:31 you know, it’s just a different dynamic.
0:13:32 So we don’t have to worry too much about these things.
0:13:35 What becomes open source and what’s not.
0:13:36 – That’s an interesting thing because I think
0:13:39 we sit in phase one.o and you’re sort of in phase two.o,
0:13:41 which is like, we’re by and large on-premise
0:13:42 desktop software vendor.
0:13:45 – This is why we didn’t go to on-prem at Dead Rix.
0:13:48 ‘Cause we wanted to have this model.
0:13:50 We didn’t want to have to worry about this.
0:13:52 So you just rent the service from us in the cloud.
0:13:53 That’s it.
0:13:56 It’s very inspired by Amazon sort of business model,
0:13:59 rather than the red hat model, which is what exists on-prem.
0:14:00 Harder to monetize, I think.
0:14:01 – Absolutely.
0:14:02 – Well, since you’re talking about
0:14:05 kind of these cloud vendors, you mentioned AWS.
0:14:07 Peter said in this presentation, you know,
0:14:10 hey, we’ve maybe over-rotated on this threat.
0:14:12 Agree or disagree with that?
0:14:14 – I think it depends on the type of software you run.
0:14:16 And what I mean by that is, you know,
0:14:18 there’s things that are super compelling
0:14:20 for the clouds to want to run.
0:14:21 And there’s things that, you know,
0:14:22 maybe they care less so, right?
0:14:24 So I think about like HashiCorp tooling, for example.
0:14:26 You know, Terraform, for example,
0:14:28 it allows you to consume more cloud.
0:14:29 And so in that sense,
0:14:31 anything that allows you to consume more cloud,
0:14:32 wonderful, what do they care?
0:14:34 You know, they don’t care if you’re running Windows
0:14:35 or Linux or whatever you want to do.
0:14:37 Just draw more power, basically.
0:14:38 And so in that sense, it’s like,
0:14:41 is there value in them co-opting Terraform
0:14:42 or console or ball and running it themselves?
0:14:45 Like, okay, you’re gonna run two more extra nodes in Amazon.
0:14:48 Like the amount, it’s a rounding error for them.
0:14:50 So I think by the nature of being management tooling,
0:14:52 it doesn’t drive what they really monetize,
0:14:57 which is CPU hours, network I/O, disk gigs, right?
0:14:57 That’s it, right?
0:15:01 Everything else is just different packaging of that.
0:15:03 So I think, you know, it depends kind of where you sit
0:15:06 in that, you know, how aligned
0:15:08 are you to what the cloud cares about, I think.
0:15:09 – I think you guys are an interesting case,
0:15:12 usually if you sort of drive all three of those,
0:15:14 you know, probably heavy users of compute network
0:15:15 and storage.
0:15:17 So I think it’s probably a different interest.
0:15:19 – We don’t get any paid on those things.
0:15:20 We only get paid on the software.
0:15:22 We, they get a separate bill from the cloud vendors.
0:15:24 I mean, I kind of agree with Peter.
0:15:26 I see it differently from like the community
0:15:30 or the media, how they describe this problem.
0:15:32 The way I see it is basically there’s a bunch
0:15:35 of on-prem vendors, ISVs, you know, startups like us,
0:15:38 they’re running successful open source projects.
0:15:40 And then their customer base is moving to the cloud.
0:15:41 So they talk to their customers and the customer’s like,
0:15:43 hey, we’re about to move into the cloud.
0:15:45 So they say, oh, okay, we’ll also offer our service
0:15:46 in the cloud.
0:15:47 So they try to offer it in the cloud.
0:15:50 Then it turns out it’s actually extremely hard
0:15:51 to offer a cloud service.
0:15:53 And it takes you many years to get good at it.
0:15:55 In our case, we’ve only been in the cloud
0:15:56 from the very beginning.
0:15:57 We’ve never been on-prem.
0:15:58 We’re good at running cloud services.
0:16:01 There’s no problem in, you know, offering services
0:16:03 that has a lot of value to customers and to pay for it.
0:16:05 And we can run it really well in the cloud.
0:16:08 So as an example of that, we had a proprietary thing
0:16:10 called Delta, which had massive adoption
0:16:11 in the last few years.
0:16:12 It’s completely proprietary.
0:16:13 And we decided to open source it this year.
0:16:16 And we open sourced it with completely liberal open licenses
0:16:18 with no shenanigans in there.
0:16:20 You don’t need to freak out and be afraid of the cloud
0:16:22 vendors if you know how to run a cloud service.
0:16:23 But it’s hard to run a cloud service.
0:16:25 Running the SaaS model has been very hard.
0:16:26 It takes a long time to get good at it.
0:16:30 So like when we did the shift over sort of to start monetizing
0:16:33 it, you tell your engineers, hey, dude, can you, you know,
0:16:34 have this pager duty?
0:16:36 And, you know, might have to wake you up at 3 AM
0:16:37 if the thing goes down.
0:16:38 And then you’re like, what?
0:16:39 Like, I don’t want to.
0:16:40 It’s like, yeah, you’re on call.
0:16:41 This is the rotation.
0:16:43 You know, you have to wake up at 4 AM.
0:16:45 Between 4 and 6, you’re covering if there’s
0:16:47 any outrage of any security breach and so on.
0:16:49 That’s the hard thing that you have to do,
0:16:50 which the on-prem vendors don’t have to do really
0:16:51 for their service, right?
0:16:53 Because it’s the responsibility of the IT
0:16:55 department of that private data center of your customer
0:17:00 to handle outages, security breaches, SLAs, yada, yada.
0:17:03 Whereas in our case, we had to tell our engineers, hey, sorry.
0:17:06 Like, that’s why we’re getting paid the big bucks.
0:17:07 Carry this pager.
0:17:09 It feels like there’s an interesting analogy, right?
0:17:11 Which is, you know, there was an era where, as everything
0:17:13 went from hardware to software, you
0:17:14 saw the hardware companies really struggle.
0:17:17 Because fundamentally, if your core competency is hardware,
0:17:19 it doesn’t translate super well to software.
0:17:21 And I think as you go from being a software vendor to saying,
0:17:23 hey, I want to be a cloud service provider,
0:17:26 the skill set, the core competency of writing
0:17:28 and developing software actually doesn’t translate that well
0:17:29 into being good operationally.
0:17:31 It’s a completely different skill set.
0:17:33 So I think as you go through from being from a software vendor
0:17:35 into saying, I want to be a cloud SaaS vendor,
0:17:37 you might find that actually your internal core competency
0:17:38 isn’t there.
0:17:42 My sort of opinion on this, it’s very hard to do both.
0:17:45 And we’re all tempted to go, many startups
0:17:49 are tempted to do both so that we have optionality.
0:17:53 Hey, if they, well, a customer can buy in any case, right?
0:17:56 But I think you guys have pointed out a very important part
0:17:59 here, that doing both is really, really
0:18:03 hard to do as a large company, let alone as a startup.
0:18:05 The misunderstanding is in the media,
0:18:08 they say, oh, you know, these big cloud vendors,
0:18:10 they’re just taking other people’s open source software,
0:18:13 not contributing anything back and just exploiting that.
0:18:16 What they forget to tell you is they’re really, really good
0:18:18 at running that software in the cloud.
0:18:19 And almost no one else can do it.
0:18:20 It’s really hard to do.
0:18:23 And actually, they’re getting paid the big bucks for that.
0:18:24 That’s what they don’t tell you.
0:18:28 It’s because ruin is the villain and hero story.
0:18:30 Yeah, well, I think that speaks to the fact
0:18:32 that there’s a lot more nuance to the relationship
0:18:35 between an open source company and a cloud vendor
0:18:37 than maybe what we see in the media.
0:18:40 How have you or how have you seen other open source companies
0:18:43 navigate the nuance of a cloud vendor relationship
0:18:46 or other partnerships around open source?
0:18:48 I think you can partner with them.
0:18:48 They’re good partners.
0:18:51 We have extremely close partnership with Microsoft.
0:18:53 We also had good partnership with Amazon and other cloud
0:18:54 vendors.
0:18:55 You can partner with them.
0:18:57 It’s like all these workloads on the planet
0:18:59 are moving into the cloud.
0:19:01 There’s just so much for us all to eat.
0:19:03 Figure out what the cloud vendors are good at.
0:19:05 Let them add value there.
0:19:07 Look at where they’re not adding value.
0:19:08 You can go there and focus on that.
0:19:09 And then partner with them.
0:19:11 It’s a win-win situation.
0:19:12 You can do that.
0:19:15 So I do think one has to figure out how to align with them.
0:19:17 And I think one mistake a lot of big companies are doing
0:19:21 is they don’t align their Salesforce comp models
0:19:24 to be compatible with the big company’s comp model.
0:19:26 The way it works at Databricks is our customers
0:19:30 get two bills, one bill from the cloud vendor
0:19:31 and then one bill from us.
0:19:34 You get a bill for the hardware storage, the watts
0:19:36 from the cloud vendors, and you get a bill from us
0:19:37 on the software.
0:19:38 The reason that is really important
0:19:42 is that that other bill that they get from the cloud vendors
0:19:45 is actually paying the sales compensation of the cloud
0:19:47 company salespeople.
0:19:48 Hence, they like us.
0:19:51 So they partner with us in the field in every account.
0:19:53 However, if we change the pricing model
0:19:54 so that they don’t get paid, then they
0:19:56 would hate us and block us from all the accounts.
0:19:58 So I think that’s like a minor nuance
0:20:00 that some companies haven’t figured out
0:20:03 and they end up in a really fierce competitive situation
0:20:04 with the cloud vendors.
0:20:05 We don’t have that problem.
0:20:07 The cloud vendors are very friendly to us.
0:20:08 I kind of want to zoom in a little bit and put
0:20:10 Peter in the hot seat.
0:20:12 He shared kind of that four stage funnel from developer
0:20:15 community management to product management
0:20:19 to kind of the lead jet and sales dev and then those sales.
0:20:20 I’m curious, how has that played out for you?
0:20:22 Has that held true?
0:20:25 What parts didn’t necessarily hold true?
0:20:27 I think it goes through those exact phases
0:20:29 that sort of Peter laid out.
0:20:31 I think then those phases actually map pretty well
0:20:32 into the funnel as well, which is
0:20:34 that early phase of product market fit.
0:20:37 It’s a lot about developer advocacy, building the community,
0:20:40 things like that as you go into a repeatable sales cycle.
0:20:42 Well, that only works if you have a tight fit
0:20:45 between product management and product marketing in terms
0:20:47 of, OK, great, we need the futures customers are asking for.
0:20:48 And then we should tell them about that.
0:20:51 So there has to be sort of an integration there, that phase.
0:20:53 And then I think as you start going to scale,
0:20:57 you get to those phases three and four in the funnel
0:20:58 to really be able to amplify that message
0:21:01 and bring in the cloud partners as part of your channel.
0:21:04 I think the only thing that maybe we experience slightly
0:21:06 differently actually would be on that fourth phase.
0:21:10 I think you laid out sort of start with SaaS, self-service,
0:21:12 then go departmental, then go enterprise.
0:21:14 For us, it’s almost been exactly the opposite.
0:21:16 And I think it, again, I think it’s
0:21:19 because we were sort of a phase 1.0 versus a phase 2.0.
0:21:22 I also think it depends on what product you’re actually selling.
0:21:25 You may not have– your value may not accrete
0:21:28 to an individual user, in which case,
0:21:30 I just want to make it clear, you may not have that line
0:21:33 because the product doesn’t support that line.
0:21:36 And so then you may start– many companies
0:21:38 start with field sales first.
0:21:42 So it was an example of how to layer it up as opposed to that’s
0:21:44 every company ought to be that.
0:21:46 But I think it goes to your point that you made of having
0:21:48 a framework in terms of what do you decide
0:21:49 goes open versus enterprise.
0:21:51 Because as I describe our framework,
0:21:53 our dividing line is what accretes to an individual.
0:21:54 Well, that’s open source.
0:21:57 And what accretes to an enterprise.
0:22:00 And so because of the divide we use in product management,
0:22:03 it’s very hard for us to have a sort of a self-service model.
0:22:04 There is no self.
0:22:05 That’s given away for free.
0:22:05 We’ve given away.
0:22:07 So you could look at that bottom curve
0:22:10 and say this is the open source line.
0:22:13 It’s not dollars per customer, whatever the y-axis would be.
0:22:16 And then you’d build your revenue model on top of that.
0:22:16 Exactly.
0:22:18 Exactly.
0:22:20 So I think for us, once you sort of acknowledge, hey,
0:22:22 that’s the divide, it makes sense
0:22:25 to start on the enterprise side.
0:22:29 In an article that Peter wrote in 2013, I think,
0:22:31 that was same time as we started Databricks,
0:22:34 in which he really accentuated the big difference
0:22:38 between this Red Hat model and the SAS model.
0:22:39 And it really resonated with us.
0:22:42 And we really thought, our view was really
0:22:44 like this on-prem Red Hat open source model is dead.
0:22:45 It’s bad.
0:22:46 We looked down on it.
0:22:48 We didn’t want to have anything to do with it.
0:22:52 And we really saw this SAS model as extremely powerful.
0:22:55 And it’s pretty prescient, because 2012, ’13,
0:22:59 and you know, AWS was not the huge phenomenon it is today.
0:23:01 Now it’s absolutely exploded.
0:23:03 It’s like taking over the whole planet.
0:23:05 So I agree with it, but I think the thing
0:23:08 that I would really emphasize is, for me,
0:23:11 the difference between SAS and non-SAS
0:23:12 makes a huge difference.
0:23:15 I think churn is higher if you’re not SAS.
0:23:17 I think net expansion rates are lower.
0:23:19 I think everything is worse if you’re not SAS.
0:23:21 Because what can happen– I’ll give you an example.
0:23:23 I’m not going to name the names.
0:23:25 But basically, there was an open source event.
0:23:27 This is one of those cases where the open source software
0:23:30 actually had two companies fighting over it.
0:23:32 They had contributors.
0:23:36 One would run it on-prem and give support and services.
0:23:38 And then what would happen is, after a couple of years,
0:23:41 the customer would keep the software, because it’s free,
0:23:43 but then migrate and get support from the other cheaper
0:23:46 vendor, but keep the same software.
0:23:48 That you can’t do with SAS services, right?
0:23:49 If you want to move away– if you
0:23:51 don’t want to have a contract with us,
0:23:52 we would shut down the service.
0:23:53 You can’t use it anymore.
0:23:55 And so hence– and then what they would do
0:23:56 is, then they get the cheaper vendor.
0:23:57 And then after a couple of years,
0:23:59 they would not even renew with the cheaper vendor,
0:24:00 because they would say, I actually
0:24:03 hired your support guy into my company,
0:24:05 so I don’t need to buy any support from you anymore.
0:24:07 And I’m still going to keep the software.
0:24:08 With SAS, you can’t do that.
0:24:10 So churn ends up being lower.
0:24:11 Expansion rates are higher.
0:24:12 Everything is just better.
0:24:15 It’s super interesting as I think about it.
0:24:17 Well, there’s an exact example of commodity
0:24:21 when I can go swat like all this software has been developed.
0:24:25 And yet, it’s total commodity that I can basically
0:24:28 go from one vendor to another to hiring a person
0:24:30 to actually support this thing.
0:24:32 And that’s what was happening during that era
0:24:35 when I mentioned the difference between open source
0:24:37 valuations versus proprietary.
0:24:41 This was exactly that characteristic that propagated
0:24:42 that particular dynamic.
0:24:45 So open source– the software itself
0:24:47 has zero intrinsic value.
0:24:48 Anyone can download it.
0:24:50 So if these companies were selling,
0:24:53 their value was support and services,
0:24:55 which quickly gets commoditized.
0:24:56 And it goes down to who can do that most efficiently
0:24:59 where on the planet, and they can manage that P&L.
0:25:01 And then you have a company like Red Hat, which
0:25:03 has sort of a scale advantage.
0:25:04 Exactly.
0:25:05 That’s their value at.
0:25:07 It’s a scale advantage to do exactly that.
0:25:10 So for people who aren’t familiar with that article
0:25:13 from back in 2013, could you give kind of a quick summary
0:25:13 of what your–
0:25:16 I mean, the title of this blog was
0:25:19 “Why There Will Never Be Another Red Hat,” was the title.
0:25:23 And I made the argument to Ali’s point
0:25:28 that the support model was pretty broken at the time.
0:25:31 And thinking about going to a service–
0:25:35 to a open source service model, hosted service,
0:25:40 was a way to really uncover and accentuate
0:25:43 the value of the product that you’re bringing to the market.
0:25:44 So that was–
0:25:47 I mean, there’s basically all the points that we argued here.
0:25:51 Red Hat had the scale and capacity to go and do that.
0:25:52 Don’t get me wrong.
0:25:53 Red Hat is a great company.
0:25:55 It’s just very hard for a startup
0:25:59 to go replicate what they have done because their value at
0:26:00 is the scale.
0:26:03 And the things that startups don’t do very well
0:26:06 is scale, because you don’t have the money to go and do that.
0:26:09 So it’s counter– it’s sort of counterproductive
0:26:10 on that dimension.
0:26:12 Compete with Procter and Gamble and Distributor.
0:26:14 Exactly, like, it doesn’t– you can’t do that.
0:26:16 You can’t do that as a startup.
0:26:18 SAS also is killing that business model even more.
0:26:19 Because the–
0:26:20 Totally.
0:26:23 The secret sauce– like, the thing that’s weird about Red Hat
0:26:25 is that of all the open source companies that exist,
0:26:27 that for some reason, that people can analyze and debate
0:26:30 forever, they ended up being a monopoly.
0:26:31 Exactly, yes.
0:26:33 Without really any fierce competition,
0:26:35 which is generally not true about the open source software.
0:26:37 You end up– because the software has zero intrinsic value,
0:26:39 you end up getting lots of competitors, which
0:26:42 commoditizes the price and brings it down.
0:26:45 So but with the cloud vendors now,
0:26:48 you’re much more unlikely to have a monopoly like they had.
0:26:50 Because if you offer just free software
0:26:52 that you’re just distributing, they can also pick it up
0:26:53 and offer it.
0:26:55 So it’s very unlikely there will ever
0:26:58 be another Red Hat because of that.
0:27:01 So what practical advice might you give?
0:27:05 Having done on-prem, I’d say skip on-prem.
0:27:09 Go straight to SAS, save yourself.
0:27:12 Yeah, I mean, I think the advantages
0:27:16 that Ali has talked about around the SAS model are very true.
0:27:19 And I think, to Peter’s point about changing competence,
0:27:22 being very hard, if you go down the road of building software
0:27:25 competence and then realize you want to switch to SAS
0:27:28 competence, very much the bucket we’re in, to be honest,
0:27:31 you realize it’s a hard shift, right?
0:27:32 It is a different skill set.
0:27:35 It’s a different set of practices.
0:27:39 And so the earlier you can do that, ideally at inception,
0:27:41 the easier your life will be, right?
0:27:43 The further you get down one road,
0:27:46 the harder and more painful that shift really is.
0:27:49 Yeah, for me, I would say SAS is obviously the one.
0:27:50 Definitely just start with the SAS.
0:27:51 By the way, Wall Street likes SAS
0:27:53 and gives you higher multiples.
0:27:54 So you get a higher valuation.
0:27:56 So there’s that as well.
0:28:01 But ignoring that aspect, I think what is it you want to do?
0:28:04 What do you want your company to do?
0:28:05 Which space do you pick?
0:28:07 And the way I think about it is you
0:28:09 have to expect these three cloud vendors, Amazon,
0:28:12 Microsoft, Google, they each have roughly $100 billion
0:28:14 of cash sitting around.
0:28:16 And they actually have a printing press
0:28:18 that’s not the cloud business, right?
0:28:20 They either have an ad business as a printing press,
0:28:23 or they have a Windows or something server business
0:28:24 that is a printing press.
0:28:27 Or they have a retail of everything on the planet
0:28:29 as their printing press.
0:28:31 So you should just assume that they’re
0:28:34 going to get really, really good at the lower levels of the stack.
0:28:36 And the lower levels of the stack,
0:28:37 there aren’t that many things.
0:28:40 There’s like machines, there is storage,
0:28:42 there’s networking, there’s some databases.
0:28:43 That’s it.
0:28:45 You move up the stack a little bit.
0:28:46 You start having much more.
0:28:49 And as you move up the stack, it gets more and more verticalized.
0:28:51 And it actually becomes a lot of different things,
0:28:53 a lot of different products.
0:28:54 The cloud vendors can’t win all of those.
0:28:57 They can dominate and crush and just completely
0:28:58 own the bottom layers.
0:29:00 The higher up you go, there’s going
0:29:01 to be a lot of vendors.
0:29:03 Otherwise, if I’m wrong about this statement,
0:29:06 there will only be three companies on the whole planet
0:29:06 in software.
0:29:08 That’s very unlikely.
0:29:10 So pick a space higher up in the stack.
0:29:11 Competition will be much less.
0:29:13 It’s going to get much more verticalized.
0:29:18 And do the sass and you’ll probably be very successful.
0:29:20 I did want to touch on briefly your backgrounds
0:29:24 and the origin of open source, as well as your own start
0:29:27 in academia.
0:29:29 Those two have been really tightly linked.
0:29:31 Now we see with commercializing open source,
0:29:33 what does that relationship look like?
0:29:36 How are academia and open source connected in your mind?
0:29:38 I mean, I think one of the interesting things
0:29:40 about academia is going back forever.
0:29:43 It’s always had this ethos of it’s free software.
0:29:45 It’s sort of publicly funded.
0:29:47 You get in government grants to pay for the stuff.
0:29:49 So you’re sort of naturally giving it out
0:29:51 or the code is there for it so other people can reproduce
0:29:53 the work and extend it and add on it.
0:29:56 So I think if you spend time in that,
0:29:57 you kind of soak in that ethos.
0:30:00 And that’s sort of the notion that the software is free
0:30:02 and other people collaborate and extend and remix.
0:30:04 That comes normal, like it doesn’t seem weird.
0:30:05 I think the other nice thing is,
0:30:07 especially infrastructure software,
0:30:09 this stuff isn’t like, hey, I’m gonna bang it out
0:30:11 over a weekend and launch my cool new app
0:30:12 and put it on Hacker News, right?
0:30:15 It’s like spend years building this stuff, right?
0:30:19 And really getting it to a point of broad usability,
0:30:21 scalability, et cetera.
0:30:23 And so where are the environments where you can spend years
0:30:26 and years working on a thing, right?
0:30:27 You kind of have that luxury in academia
0:30:29 to be able to do that.
0:30:31 And so it’s like, the first few years of development
0:30:34 take place in a university,
0:30:37 and then it becomes sort of an industrial project from there.
0:30:39 But it would be very hard to bootstrap some of the stuff
0:30:41 from zero in an industrial setting.
0:30:46 – Yeah, I actually think academia is misconfigured.
0:30:48 I mean, I’m actually a young professor,
0:30:50 so I’m kind of, I wear that hat too.
0:30:53 But I think in the systems research field,
0:30:54 it’s misconfigured.
0:30:56 I think there’s a huge opportunity for academia
0:30:59 to come in and completely disrupt the software scene.
0:31:01 But the way it’s configured right now
0:31:03 is as academics,
0:31:04 we get incentivized on publishing papers
0:31:06 in the top conferences,
0:31:09 and that’s what we focus on typically.
0:31:11 If the focus was on push the boundaries
0:31:13 of what kind of software you can build
0:31:14 and disrupt the world with it,
0:31:16 you know, all these universities,
0:31:18 with all these students that have five years
0:31:19 to sit and create the open source project,
0:31:21 could like completely disrupt
0:31:22 how software is done on the planet.
0:31:24 It’s a gigantic opportunity for any university
0:31:26 to sort of take on.
0:31:28 Berkeley actually did it, I mean, to our credit.
0:31:30 I mean, we kind of pushed forward on that
0:31:32 with some of these labs that we had,
0:31:34 like Rad Lab, Amplab, Rise Lab,
0:31:35 but it’s just scratching the surface.
0:31:38 I think there is a scenario,
0:31:41 if all the universities kind of figured this out,
0:31:41 which they haven’t,
0:31:44 that they could like completely start owning the software space.
0:31:47 It could become like the future of how software is developed.
0:31:48 It’s like basically open source projects
0:31:50 that the different universities are leading.
0:31:52 That’s not happening right now.
0:31:55 – All right, so kind of time for like a last question here.
0:31:56 What area of open source right now
0:31:58 is the most fascinating to you?
0:32:01 Where do you think the most interesting things are happening?
0:32:03 – I mean, so I’m going to kind of not answer that
0:32:08 by saying data and how, you know,
0:32:11 the value of data and the ecosystem around it,
0:32:12 how you can buy it, how you can sell it,
0:32:13 how you can leverage it,
0:32:16 and the models that interpret that data.
0:32:18 That’s, you know, we think of software,
0:32:20 we used to think of a hardware software and so on.
0:32:22 I think data is the next thing.
0:32:24 And, you know, I mean, a lot of people have said it,
0:32:25 it’s the new oil or, you know,
0:32:27 but we’re just in the beginning of that.
0:32:29 It was the early innings of how that’s,
0:32:31 there’s going to be an economy forming around data itself.
0:32:32 I mean, it’s already kind of happening.
0:32:33 So that’s, I think, fascinating.
0:32:37 So I think that’s like the next third wave
0:32:39 of interesting sort of market trend
0:32:41 that you’re going to see in the software space.
0:32:43 Like the software will kind of be free,
0:32:45 but who has the data will actually,
0:32:46 and the models around it are going to be the,
0:32:49 that’s going to be the competitive edge.
0:32:51 – I’ve always personally been a bit of a systems guy.
0:32:53 So I love following sort of like what’s happening
0:32:54 in database research land.
0:32:57 And to me, I think, you know, it’s been interesting
0:33:00 ’cause it’s like it already BMS has sort of ruled the world
0:33:01 for decades and decades.
0:33:03 And I think what’s finally happening
0:33:05 is you’re either seeing shifts because of scale, right?
0:33:06 You’re no longer saying, okay,
0:33:08 I can fit all the data on one machine.
0:33:09 I need to, you know, fundamentally go
0:33:11 to a clustered architecture.
0:33:14 Or now as we think about sort of IoT edge,
0:33:15 fog computing, whatever you want to call it,
0:33:18 there’s these notion sort of hierarchical levels of data
0:33:19 where you might have high bandwidth,
0:33:21 high throughput, you know, cloud data centers
0:33:23 and then, you know, go out to sort of an Akamai
0:33:26 or Fastly Pop and then go out to someone’s house
0:33:27 and then go out to your phone.
0:33:29 And so how do you actually design systems
0:33:32 that can reconcile and handle data sort of planetary scale,
0:33:34 much higher volume, much lower latency
0:33:39 and reconcile and do it all in a comprehensible way.
0:33:41 So I think that’s a sort of a fascinating space
0:33:44 in terms of, you know, is RDBMS finally being challenged
0:33:48 as sort of supreme when it comes to data management?
0:33:52 – I think one of the fascinating elements of open source
0:33:54 is the origination of projects now.
0:33:58 Like these stats of Google doing 2,000 projects
0:34:01 and all of that, maybe that’s part of the answer of,
0:34:05 you know, I love your academic sort of comment.
0:34:06 It may not happen there, but the fact
0:34:09 that all of these companies are really built
0:34:12 on large backbones of open source
0:34:14 and are releasing these projects into the market
0:34:17 where there’s, you know, not a lot of strategic value
0:34:19 to them, I think it’s in quotes,
0:34:23 unlocking a huge number of opportunities
0:34:27 that I think will very much dominate the landscape
0:34:30 as we, you know, sort of roll into the future.
0:34:31 That’s really interesting.
0:34:34 – I want to thank you, thank the panelists
0:34:36 and a huge thanks, Peter, to you
0:34:37 for sharing all this information.
0:34:39 (audience applauding)
0:34:41 – You just heard Olly Goetze of Databricks,
0:34:43 Armand Adger of Hashicorp
0:34:45 and A16Z general partner, Peter Levine,
0:34:48 with me, Das Rush, moderating the discussion.
0:34:49 Thanks again for listening.
0:34:52 And if you want to learn more about open source,
0:34:53 the importance of commercializing it
0:34:55 and what it takes to turn an open source project
0:34:57 into a business, you can download
0:34:59 and/or watch Peter’s presentation
0:35:04 and other open source materials at a16z.com/opensource.
Today, despite the critical importance of open source to software, it’s still seen by some as blasphemous to make money as an open source business. In this podcast, Armon Dadgar, Cofounder and CTO of HashiCorp; Ali Ghodsi, CEO of Databricks; and a16z General Partner Peter Levine explain why it’s necessary to turn some open source projects into businesses.
They also cover the most important questions for open source leaders to answer: How do you keep community engaged while building a business? What new opportunities does SaaS (software-as-a-service) present? And if you are a SaaS business, how should you approach cloud service companies, like Amazon Web Services (AWS)?