יום רביעי, 29 בספטמבר 2021

421 The Cost of Cloud, a Trillion Dollar Paradox with Martin Casado

[קישור לקובץ mp3]


שלום וברוכים הבאים לפודקאסט מספר 421 של רברס עם פלטפורמה, והיום הולך להיות פרק מעניין ומיוחד - באנגלית.
Money Sign Money Bag Icon Vector Clip Art - Dollar Sign Clipart Black And White@clipartmax.com


Hello and welcome to Reversim! This is episode number 421, and today we’re delighted to host Martin Casado. We also have Nati Shalom with us, and we’re going to speak with Martin about a topic we actually previously covered in our Podcast [though in Hebrew - 418 Carboretor 31 Cost of cloud paradox] - “The Cloud Paradox” or “The Trillion Dollar Paradox” article by Martin and Sarah Wang.

(Ran) So - welcome, Martin, we’re delighted to have you here. Let’s have a quick introduction and get to meet you . . .
  • (Martin) Sure - so, the quick introduction:
    • I did my PhD in Computer Science at Stanford, and I focused on Systems and Infrastructure.
    • From there, I started a company called Nicira, where we worked on building out software  networking in data centers, working with large Clouds 
      • so we were very, in the early days, close to Google, with Rackspace, had lots of conversations with Amazon . . . 
    • That [Nicira] was acquired by VMware, and I ran all the Network and Security for VMware, where I sold into the data centers.
      • I did that for about 4 years, I ran the business; When I left it was about $600M/year business
      • Right now it’s about $2B/year business . . . .
    • And for the last five years I've been at Andreessen-Horowitz, where I'm an investor and I focus on infrastructure 
      • and in particular, you know - Cloud infrastructure.

(Ran) Thank you. So, as mentioned, this is a continuation-episode to a previous episode that we recorded about your blog post, and now we'll have a chance to dive deeper.
So Nati - you want to take the lead on that?

(Nati) Yes, thanks so much again, Martin, for accepting this call. I know that you've been very busy, probably not just around this topic - and I wanted to maybe kind of start the discussion with background, like - what brought you to this topic?
I looked at your blog-post, and it looks like you've been talking about this cost-efficiency - or the lack of cost-efficiency, to be more specific - in a couple of . . . I'll say in the past.
So it looks like this is a hot topic for you and not a new one - maybe you could walk us through the background?
  • (Martin) So yeah, listen - I've been in Cloud infrastructure and data centers infrastructure for a very long time, and what's very interesting about the shift to Cloud is that you go from, you know - 
    • from the perspective of a business - you go from being a Software business, where you're shipping software, to basically being a Services business, where you're offering a service 
    • and the economics of these businesses just look different - and we're still kind of coming to grips with what that means.
  • Now listen - I'm an investor, and I sit on a lot of Boards . . . - and as investors, we have to become students of business models, I mean - that's what we do.
  • And so I sat on all these boards, and I noticed that the economics were quite different for these companies that were born in the Cloud and the folks on the Cloud, 
    • and so, you know - a lot of this was just born out of trying to understand what the economics were, or what the economics are - 
    • as opposed to traditional software companies, where you basically ship the software and it runs on somebody else’s infrastructure, so you're not kind of syncing that, 
  • and it comes from a deep background in Cloud and data center infrastructure - and being an investor sitting on the boards and then trying to understand these new businesses.

(Nati) Excellent. So I think one of the things that comes out of this analysis is that there is a gap here, in terms of . . . between, you know, the trajectory to velocity versus efficiency - and that kind of led to this paradox that you were referring to.
Maybe you can walk us through this paradox - what is that specific paradox and why do you think, you know, companies that get to that at scale and not before or not . . .?
  • (Martin) So here's the paradox - the paradox is if you're small or doing a new project, it only makes sense to use the Cloud.
    • Why? because you don't have to deal with infrastructure
    • You don't have to deal with the talent of the infrastructure 
    • and so all of that you can outsource, right?
  • And so any startup that doesn't start using the Cloud, like, had better have a really good reason . . .
    • I would say the vast-vast-vast majority of any company should start just by using the Cloud.
    • and that's, you know, is just a reality.
  • Another thing about one-year-small-company is that margins don't really matter because, you know, you're really focused on growth, really focus on innovation  - and why worry about margins if you don't even have Product-Market fit, right?
    • And this is something you worry about much later.
  • On the other hand, once companies start getting larger, a number of things happen:
    • One of them is that you have more specialized talent.
    • Another thing that happens is that you have more predictable workloads - or at least you understand what your workloads look like.
  • And it turns out - we looked at, and we can go through the actual study -we looked at 50 public companies . . .
    • These are relatively new companies, you know, more than 70% of them had IPO in the last six years.
    • And across those 50 companies, if you look at their COGS - their Cost of Goods [Sold] - on average, 50% of those were the Cloud.
  • And so if you're a very large company and you care about increasing your share price, the share price of your company, what you normally do is you look at what line-items can you reduce
    • and the the most significant one is Cloud 
    • and it's actually, in this case, in these companies - it's depressive, you know . . .
    • that line item - if you could reduce it, it would increase your share price by a whole bunch.
  • So now they have this paradox - which is early on, it makes a lot of sense to use the Cloud, but in some cases later on, that's the number one impact to your share price . . .
  • And so the question is, you know - what do you do?

(Nati) In that context, again, when you gather this information and all these analysis, I saw a lot of discussion about the methodology and the data itself.
So you mentioned that you gathered information about 50 companies that are at scale and the analysis that you've done, I think, was interesting, from the fact that you . . . you know, normally when we do the comparison of the efficiency, we compare how much hardware costs outside the Cloud and how much it costs in the Cloud - and you kind of turn it to be a measure of valuation.
So it's interesting how you came about this methodology and what was the methodology . . . .
  • (Martin) This is a great question, which is like . . . normally, the way people do this analysis is well - you can buy a server for 5,000 bucks where you can rent one in the Cloud for this much money, you know -  and they compare that.
  • and that makes sense in the “old world” - so what was the “old world”?
    • I grew up in “the First Cloud Wars” - it was the Cloud versus internal IT, and internal IT is not part of COGS, right? it's a back-office function 
    • so in the old Cloud Wars - let's say between 2005 and 2015 - the entire argument was exactly that -
    •  You said you're comparing against servers and then, you know - it's pretty easy to make the argument that you go to the Cloud so you don't have to buy servers.
  • But that argument doesn't make any sense, or that cost analysis doesn't make any sense, to a SaaS company - and the reason is because the infrastructure is part of COGS, and COGS impacts share price . . .
    • and so it's not meaningful to say like “Oh, we saved 5 Bucks!”  . . . 
    • what is meaningful to say is "We can increase our share price by a billion dollars”.
  • And so I view this almost as the “Cloud Wars 2.0”, which is as software companies move to SaaS, Cloud becomes part of COGS.
    • Again, this is a very recent phenomenon and so then you ask the question: how does that impact the share price? which of course - share price is everything:
      • It's how we recruit, it's how we get debt . . . that flows over to cash.
  • and so, you know, It's just a new question and it's worth asking - and it turns out that it dramatically impacts COGS 
  • and I want to say one more thing, 'cause you set up before, which is - people have said  “well, why would you worry about COGS, when you should be worrying about growth?”
    • That's actually a meaningless statement because, if you have an extra billion dollars in share price, you can invest in a ton of growth . . .
    • so this isn't about like efficiency versus growth - this is literally about giving away margins to another company.

(Nati) That's a very nice argument about this measurement, and again - if I go back to the discussion around this and what I heard from . . . you know, I had a couple of discussions after that with multiple engineering people, and they said “but it's not necessarily related to Public Cloud versus on-Prem” -  and it looks like you made a lot of references to Dropbox and to Repatriation as a way of optimization, and that kind of steered the discussion, in my view, in the wrong way.
So, first of all, why did you find Dropbox so interesting and why do you think it's a reference? because it sounds very unique, in the lens . . .
  • (Martin) You know what's actually unique about Dropbox? it’s that they’re public about it . . .
    • So, you know, we weren't even going to talk about repatriation - this is the irony of this whole thing.
    • Like, the point of the piece was just to look at the impact of Cloud on COGS . . . like, when we started, it had nothing to do with repatriation.
  • But it's interesting, because we talked to all of these companies - and let's say we talked, I don't know how many . . .  let's say a dozen companies, where we actually talked to the architects
    • and almost every one of them had either were penciling out a repatriation, planning on repatriation - or had done a repatriation, right?
  • and now listen -  by repatriate, it doesn't mean they take all the workloads off the Cloud, that's silly, doesn't make sense for anybody to do that. 
    • But if you're very large, you know, it may make sense for some core workloads.
  • So it . . .  the Dropbox thing is in no way like an anomaly or unique - but what is very unique about it is that they were so public and they published this study to show the impact.
    • so we used that study as an illustration - it didn't have anything to do with the analysis, by the way, 
    • and so in all of the discussion around the . . . I think it's actually worth saying: in all of the discussion around the piece which there's been so much
      • I mean, there's been dozens of follow-on pieces . . .
    • In all of the analysis, nobody has disagreed with the numbers - and instead, you know, people say “well, you know, like, oh, you should have used Dropbox as an example . . .”
      • which, again, had nothing to do with the analysis 
    • or they will go to these “Cloud 1.0” arguments - the old arguments, which have also nothing to do with the piece . . . 
      • The pieces is about SaaS companies . . . Impact margin and share price and so.
  • I actually think this is a new result, and I think it's a new discussion point, and I think it's a very interesting one, and I'm hoping that the discourse will evolve, 
    • so we're actually talking about the problem at hand and not kind of dredging up, you know, these kind of stale arguments that don't really apply.

(Nati) Right . . . so Repatriation, if I summarize it - Repatriation is (1) not that rare , it was a good example to latch on, based on your analysis.
And (2) - would you agree that when we talk about this efficiency, it doesn't have to be Repatriation, it could be anything . . . even with optimization within the Cloud - it’s just that you have to be more aware and put the right resources into that cross-park . . . 
  • (Martin) Well, let's just go through what the results actually said . . . let me just state very clearly what the results - what the analysis was and what were the results. 
  • So the analysis was first - if you look at modern SaaS companies, what percentage of their COGS is Cloud? We did that by looking through S-1’s
    • The result was that, on average, 50% of their COGS was Cloud - that's a HUGE number, OK? 
    • so that was the 1st result - 50% of COGS is Cloud for these top-companies.
  • Then the 2nd result is - we asked the question: if you reduce that by a factor of two, what is the impact to share price?
    • And of the 50 companies we analyzed, the impact of share price was, you know, over $100 Billion . . .  
    • meaning, if those 50 companies reduced their Cloud costs by a factor of 2, their share price would increase by $100 Billion - and the number is actually probably more like 200, we were very conservative . . .
  • So that's the results of the paper, that's it - Beginning and End.
    • And those numbers . . . You know, I very much stand by, I mean - like make the data public, like whatever.
  • So then the question is - what are the implications?
    • Well, a very clear implication is that if you're one of these companies, you should find a way to reduce your Cloud costs, right? . . . . it's a very clear implication . . . 
    • and a way to reduce your Cloud costs, as [there are] many of them, to your point, right?
      • So you could just focus on Software optimization
      • You could buy a 3rd-party tool which helps you with Software Optimization,
      • or - you could consider Repatriating . . .
    • And when we spoke to practitioners - and remember, you know, I used to run a very large data center business, so this is not like, you know, like some business-guy just like looking at this - I mean, I used to run this business.
      • They talked to practitioners - and they employ all three of them . . .
    • Some will . . . you know, some ignore margins, some focus on just optimizing their code, some work with 3rd-party systems and some have actually pulled workloads off the Cloud.
    • Not all of them - normally just some subset of workloads, increased costs.
    • And that's really, to me, the kind of, you know, the “So what?” or the “What can you do about the results?”
      • Is that you can consider one of these three things.

(Nati) So in your view, why are we in that stage? I mean, I'm assuming that those companies are not stupid, they look at the numbers and they see the numbers that you're seeing . . .
Why haven't then got to the same conclusion before? why haven’t they got to the point where 50% of their cost is Cloud-cost?
  • (Martin) Yeah, so this is a great question . . . so I've got 2 answers - Well, I've got 3 answers for that -
    • The first one [is that] this is actually a relatively new phenomenon - 
      • It's just that we don't have a lot of maturity with SaaS business models . . .  like, we've been talking about SaaS forever, but there hasn't been a lot of SaaS infrastructure, there hasn't been a lot of Cloud companies that have been fully on the Cloud that are public - this is a new thing.
      • So the first one is just [that] I don't think people have been looking at it.
    • The second one is - companies, early on, get private funding from people like me -  and we just don't care about margins as much, because we're just interested in seeing if, like, the product will work, or if people will buy it, right?
      • And so, you're not investing in efficiency, and you're not investing in COGS - you're just investing in software growth.
      • So the second one is that the private markets don't look at it
    • And then the third one is - even the public markets these days, because cash has been so cheap - like, debt is basically free - they're just investing in growth rather than efficiency, because that's what the public markets say . . .
  • Now, the question we should all ask ourselves is “listen - we looked at 50 companies, 70% of them had IPO in the last six years, they're all growing 'cause that's relatively new - what happens when (A) they slow-down growing or (B) the market starts to value margins - What happens then?”
    • You know, and the reality is . . . the impact that the Cloud has on the share prices can be even more dramatic, right? Instead of $100 billion or $200 billion, it could be, you know, $400 billion . . .
    • And so - if you extrapolate our results to the industry, you could estimate 500-billion-to-a-trillion dollars of share prices being impacted. 
      • You could estimate that, you know, I mean just to . . . just look at this, you know, the total number of similar companies,
      • that's a lot of money that's being impacted, and so I suspect we're going to see a pretty dramatic shift to address this.
      • I don't know what it's going to look like, but I suspect something will happen.

(Nati) So if I summarize what you just said, which is again very interesting - you basically point out, rightfully so in my view, to the maturity of the market, to the maturity of investors that are, you know, running those new companies, their new growth companies - the investors obviously measure things that are not necessarily related to efficiency, but related to growth, and that's why companies are very much optimized towards that.
And similarly, if we look at the companies themselves - they’re also being measured at growth, so that's how they optimize themselves.
There's also the movement between what you said - the “Cloud Wars 1.02” to “Cloud Wars 2.0”, which is part of that maturity in which we kind of think that moving to Cloud and Cloud native automatically give us that efficiency - and we don't necessarily look into what is efficiency within the Cloud . . .
So we think that people using the Cloud in itself - in it by itself - it gives us that efficiency, and I think clearly, from your study, there are today many options in the Cloud that wouldn't necessarily fit into that definition. 
So is that a good summary?
  • (Martin) Yeah, I think that's a great summary
  • And again - I just think it's very important to whoever is listening . . .  I think a lot of feedback on the the post presents a false dichotomy between efficiency and growth.
    • They're saying like “why do you care about efficiency,  when you should be focused on growth?” - but it's a false dichotomy.
    • The reality is - if you have more money, you can grow faster
    • So, like - if your share price is being knocked down by a billion dollars, you can't invest in growth as much, and so you know . . .
    • Again, I think that people don't understand that we're in a quite a new world with companies in SaaS, and we need to all sit back and reevaluate it, and think of what it means.
  • And listen - I'm an investor now, right? And so, like - I don't care what the answer is . . . I don't care if it's Cloud, I don't care but . . . I don't care.
    • I just want to know what the economics are, so I can invest my money, right?
    • I mean it's, you know, like - I'm totally neutral in all this.

(Nati) So one comment that I heard about this argument from another large public company, who has the same problem - they said that it's not necessarily related to the efficiency, the way that you kind of articulated it . . . it basically a reference to the fact that in SaaS business model, you're almost forced to deliver more value - and when you’re asked to deliver more value, for the same price, your margins are going to be marginalized and the cost of infrastructure, which by definition will grow . . .
So is it the problem with the SaaS business model, where people are, you know, kind of  . . .  when I'm saying “people” it's mostly customers - are expecting to see more and more value for the same cost, and therefore there is some anomaly here? you’re not necessarily going to be able to match the cost of the value that you are continuously generating . . . 
  • (Martin) Yeah, I just . . . I don't understand the argument, like - I mean, you've got great SaaS companies like Datadog, with amazing margins,  and then they innovate and . . . 
    • I don't understand how SaaS has anything to do with your inability to add value.

(Nati) OK, so I think that's a short answer to a long question, which is always great . . 
I think you're right in the sense that when we look only at SaaS - clearly SaaS companies are able to generate more value - and let's talk about those public companies.
I'm curious what was their reaction? - you mentioned, you know, their names, you mentioned the names publicly - you talked about Datadog, you talked about Dropbox, Asana . . . what was their reaction to all that?
  • (Martin) Yeah, I mean . . .  honestly, I haven't heard anything from Datadog or Asana, so . . . 
  • The reason that we chose Datadog, by the way, in the post, the reason that we did is . . . it's actually like the most conservative case, because like, you know, very very few companies grow at the pace that they grow - at the size that they are.
  •  In the history of infrastructure, very few companies have been over a billion dollars and are still growing in . . . yeah, I think they're 60% or whatever
    • and so - to make the argument for Datadog is like choosing kind of the worst example, which is exactly what we wanted to demonstrate, right?
    • we didn't want to cherry-pick a great example of a company that's already slowing down
    • we're like - this is a very fast growing company that's at scale, and even then - there's billions of dollars they can save, right? 
  • Now, does it matter to them? You know, that's probably 7% of their share price - maybe or maybe not - but like it's still a billion dollars, right?
    • and there's a lot you can do with that and so - that's the specific reason we chose Datadog - it’s just 'cause it's, like - it's the worst case of our argument, and still I think it's compelling.

(Nati) And so - in your view, let's take again, let’s pick on Datadog for a second - if you were in Datadog and you were reading your article again, what would you do differently, today?
  • (Martin) I actually think these big companies are largely already doing it . . . I actually think the best example is Spotify - Spotify has been tremendously prescient on this, so Spotify . . .
  • I you know, it's . . . I'm gonna answer you by talking about what Spotify is doing, and if our data dug-up . . . you do with Spotify, and I presume that they are.
  • So Spotify has actually made Cloud costs a first - like a first-class primitive that is visible all the way down to the developers, 
    • and so they've got, you know, an internal project that's a dashboard - that shows for anything that you build, what the actual cost is.
    • And that level of visibility is going to produce, you know, much better developer behavior, that's going to consider cost.
  • And in the results - they have a great blog post about that, where they actually talk about that and why they did it is for this reason
    • And I compare that to not . . . this is another public company I spoke to, I'm not gonna mention the name, but it's kind of a large, you know, infrastructure company where they have multiple product lines - that have zero-percent margin, because that they use the Cloud, right?
    • This clearly, like - if you don't control this problem, you can end up basically just reselling the Cloud . . .
  • And so you have to do something like what Spotify is doing - and you probably have to do it early in order to rate-in this cost.
  • I mean, they're . . .  there are honest to goodness companies today, that their new products have no margins because they haven't addressed this problem early enough.

(Nati) So let's talk about Spotify again, as a reference - I think you mentioned it and I think it's a good reference - so you mentioned that they made the efficiency and the cost-efficiency, First-class-citizen and . . . meaning that, at least in my words, they look at that as just another feature, and they incentivize their development team to actually optimize,  by making the cost visible to the developers, if I got your point correctly - can you elaborate what that actually means?
  • (Martin) Yeah, well no no . . . I mean, it's actually very interesting, right? I mean [that] the macro trend . . . the macro drop - backdrop - to all of this is the move from software to SaaS, right?
    • it used to be the case “I would write software, it's running on somebody else’s’ machine so all I cared about was performance” - I didn't care about cost, right?
    • And so, like, that's just not part of the developer’ culture . . .
  • Nowadays, if you're writing software and it’s sub-optimize, it actually impacts the margins to the company that you work for, and you know it turns out - to a large degree.
  • So what Spotify has done is they say “OK, if you're developing a microservice, like if you're developing a service, what we're going to do is when we show you that service ,we're going to show you the “Health Score” 
    • like - what's the availability? what's the up-time? what’s the latency?
    • but we're also going to show the cost.
  •  And by doing that, at least you know if that's going to dramatically increase or not, 
  • And because Cloud cost is such a significant part of COGS, controlling that, you know, as we've been mentioning, is a priority for the business, 
  • So I would recommend anybody listening to this - that is, doing a Cloud service - to think, you know, as your developers are building, you hold them up to a set of standards, you know - 
    • Number of bugs, whatever it is . . . Performance, Documentation  . . . 
    • Cost absolutely should be one of those - there are things you should be tracking.
  • Now, it's up to you whether you care, right? You may decide that you're in a growth phase and you don't care.
  • I mean, that's basically the core of the argument here - you need to be in a position that once you care, you can do something about it.

(Nati) So that makes me 2 follow-up questions on that regard: one of them is related to the development team itself and how they structure it, because a lot of the listeners for Reversim - the Podcast that Ran is running on and we're working sometimes together on that - is that Startups that are looking for, you know, those tools and usually what we find is that a lot of those tools (1) have, you know . . . need a lot of rights, and they're basically exposing a lot of information, and it's very hard to know, as a developer, what that actually means.
So as part of the shift-left, we're starting to see a trend towards FinOps, like a recognition that we need specialized tools that would do cost analysis - not for IT but for developers, and that needs to be part of the DevOps cycle and part of the development cycle . . .   
So is there . . . are there any tools already that is, you know, providing the cost-analysis for developers, that is different than, I would say, cost-analysis for “general IT”?
  • (Martin) Yeah, I mean - there's actually, a bunch of them . . . 
  • They developed . . . They developed their own, but it's open source and it's available and they've got cost insights, which is a plug in.
  • So if you go to like, you know, Backstage.io - there's a blog post called New Cost Insights plugin: The engineer’s solution to taming cloud costs, which does exactly that.
  • So there's a lot of companies and a lot of products out there which help you understand the Cloud costs.
    • Track them over time and then, you know - provide some guidance on how you want to reduce them.
  • And again - I do think, from the listener standpoint, the point is not that you should drain your Cloud cost.
    • That is a business decision - and I have no opinion on whether it makes sense for you
    • but I do know that it's quite likely, over time, that it will matter - and when that happens, hopefully you’re going to be in a position to do something about it.
    • And as part of that, you know, making, you know . . .  pulling these tools into your normal developer workflow - I think it’s very important.

(Nati) And in that context, do you think that we're expecting to see these FinOps now growing, very similar to the way the DevSecOps, you know, kind of started to merge?
  • (Martin) I mean, yeah . . .  so I, you know, I don't know the answer to that.
  •  I do know that like basically API catalogs, microService catalogs, you know, dashboards that show kind of performance and costs, I mean - they're just going to continue to increase in popularity.
  • And the thing for me with FinOps is like . . . I just feel like the definition has become so muddy that I don't know, 
    • but I think we can all say tooling that allows for the visibility and display of efficiencies and costs to all parts of the business, from the CFO down to the engineer, are going to continue to grow in popularity - and they should, that I'm very confidence.

(Nati) My next question would be on kind of connecting . . . Yeah, I'm saying that the other question that I have is kind of taking on the fact that you're covering both Andreessen-Horowitz as an investor and you've been an entrepreneur before - so you kind of see both sides . . .
And the question was what should change in the investor side, in your view?
  • (Martin) Well, that's what I'm trying to find out . . . 
    • No, I mean it's, you know . . . I mean you cannot . . . it's what I'm trying to find out.
  • You could argue that this actually . . . this has nothing to do with me, because this is a public-company-problem and I'm a private investor
    • and by time the the company's public, you know, I don't have any exposure to it 
  • and so, you know, I definitely think it means that there's a lot of opportunity to invest in Cloud tooling.
  • I do think, you know, if we want to hold long positions for companies at scale, like as a board member, I do think that it provides some guidance on how you instruct the company
    • but for me, the big thing is just trying to find big shifts in the industry, 
  • And you know what's interesting about this? this is my favorite thing about this entire study, and it's the following - 
    • So I was involved in the time in history when the Cloud service providers had this exact same dilemma - and then I saw what happened, so let me just walk through that, and this is the irony of this whole thing . . .
    • The irony of the whole thing is, like, this exact same thing happened to the Cloud service providers back in, you know, the late 2000s . . .  so, you know, let's say Google and Amazon and Microsoft, you know - they were building their own Cloud services, right?
    • I even have the slide, it’s kind of funny . . . if you look at the slides from, like, 2010 - they're saying “OK, well, we've got AWS, and then they did a cost analysis of their COGS . . .
      • Because now the infrastructure is COGS and they found, I guess . . . guess how much of their COGS was servers?
      • Turns out it was 50% . . . just like the Cloud is, right?
      • So they looked at their COGS and like “Oh, goodness!”, right? . . . “they're like 50% of our COGS as servers!”
      • So what did they do? they totally disrupted the server supply-chain, right? . . .
      • And at the time, you're like “these guys - they're crazy!”, because these are servers, this is a software company - how are you going to go against HP and Dell?
      • And like, you know, - there's no possible way . . . what do you know about, like, you know  . . .
      • And yet, the economic argument was so compelling that they did it, right?
    • So we've got this historical-analog, you know, that I think is coming up again. 
    • So as soon as you look at your COGS and you're like 50% as a single line-item - you disrupt that line-item . . . 
  • And so what I'm very interested in is - OK, so you've got  all these SaaS companies, they have a single line-item - what are they going to do?
    • and if they do something dramatic, the industry changes . . .
  • and I think it's very interesting to think through what that industry change might be, right?
    • It could be that they prop-up like a “Generic Cloud”, right?
    • That's what happened in Pharma - if you look at Big Pharma, like, what happened is the buyers propped up Dr. Reddy’s, which was a generic drug producer in India.
      • Maybe someone will prop-up a generic Cloud?
      • Maybe they'll repatriate some workloads, you know?
      • Maybe you'll see special-purposes Clouds?
      • Maybe they'll find a way to get the Cloud providers to erode margins . . . 
    • I don't know what's gonna happen, but I know there's a trillion dollars in the balance - and so there's gonna be some big shift . . . 
      • I just don't know what it is yet - which is why I'm glad we're having this discussion 'cause it helps me think it through.

(Nati) True, yeah . . . so basically, what you're saying is that we need to apply supply-chain methodology into how we run SaaS business, rather than, if you'd like, just think about it as someone else’ problem . . .
And that means that the whole idea of outsourcing your entire infrastructure should be questioned, again . . . because you - by definition - going to limit how much you're going to optimize that supply chain.
Again, if you . . .  and the example that you gave about the Cloud providers and the 50% and the “Aha! moment” for them, I think, should happen to a lot of SaaS companies.
And I think that, you know, makes repatriation an interesting option, because it's, in a way, the same kind of move that the Cloud providers have done when they disrupted the server market, in some sense at least.
But it makes it at least more logical than, I think, some people who are kind of still in the transition to Cloud, and all of a sudden someone tells them “repatriate!” . . . we kind of look at that as a crazy idea . . . 
But I think when you think about it from a supply chain perspective - and in supply chains you basically look at the chunks of . . .  you know, supply and the cost per each item - and if there's a, you know, an item that is a big chunk of your supply, you automatically look at how you optimize it, and that's a different kind of way to look into this.
So I think that's a very good analogy . . . kind of looking into the transformation that happened in Cloud to get them to that point, and I think that's maybe one lesson . . .

(Ran) So, with that, we'll wrap up.
I thank you again very much, Martin, and thank you Nati - it's been a pleasure talking to you. 
So thank you for your time - and hope to talk to you again.
(Martin) Wonderful - thank you so much, it was a real pleasure.

Thank you, cheers.
 

 האזנה נעימה ותודה רבה לעופר פורר על התמלול!