Mellanox's CEO Presents at NASDAQ Annual Conference (Transcript)

Executives

Eyal Waldma - Co-Founder, President and CEO

Analysts

Mellanox Technologies, Ltd. (MLNX) NASDAQ Annual Conference Transcript December 3, 2013 8:45 AM ET

Unidentified Analyst

Good afternoon. [Stephen Brandreth] from NASDAQ. Our next presenting company is Mellanox Technologies, a leading supplier of end-to-end InfiniBand and Ethernet internet solutions and services for servers and storage. Eyal Waldma is Co-Founder, as well as President and CEO. Eyal?

Eyal Waldma

Hi. Thank you very much and welcome to the Mellanox presentation. So we have this Safe Harbor which states a lot of things. But I think the most important thing and strategic explanation of what Mellanox is this on the first file.

So what’s happening and why are we so strategic is what you see on the right, data is growing exponentially. Not only that, it's amazing, today people start to archive on flash and not on tape anymore.

And the reason is they want to keep using the data that they're using now for the next 10, 15 years to make conclusions out of that. So on the right-side, you see data growing exponentially and add on top of that that people don’t want to archive thing but want to have the data accessible all the time.

On the top left side, every day more people are writing more applications to use the data. You see students write for their phones, Android, iPhone, whatever, tablets and so on, but this is growing.

And then the number of devices that can access the data and you will soon talk also about internet of things is growing also in a very fast pace. People say that in five years more than 36 billion devices will communicate produce more data that will be stored and utilized by other applications. All this requires a very efficient mechanism to move data around. We have the largest pipes. We have the best pipes and that’s why we are so strategic to all of the ecosystem out here. We will talk about this in a few slides again.

So Mellanox, we are the interconnect company, we do both Ethernet and InfiniBand. We are the only guys that shipping end-to-end 40 gigabit per second solutions both Ethernet and InfiniBand. We also have 56 gigabit per second. Everybody else is still stuck at 10 gigabit per second except for some switches.

We are based both in Israel and in California, but we're in about 15 countries around the world. We now have R&D also here in Europe in Denmark. Last year we did $0.5 billion in revenues and this year we’ll expect to do about $390 million, $400 million.

So if you are going to the data center, it comp -- it is composed of compute, engines or platforms and storage, but both of them use CPUs, memory and I/O. CPU is growing faster and faster that's in term responsible for memory. There are multiple vendors.

The interconnect becomes the ball neck for efficiency, for utilization of the CPU performance and the memory. And then exactly what we do, we have the lowest latency in the highest bandwidth interconnect.

We already talked about data growing in a very fast manner and more applications and what you see, we are now seeing this effect of people providing cloud infrastructure, whether it's infrastructure as a service, platform as a service, software as a service, you are getting more and more platforms out there. But underneath that, someone needs to move the data.

We are seeing today everybody is using 1 gigabit, some people already move to 10 gigabit, we are now experiencing Web 2.0 and cloud guys start transitioning and using our stuff at 40 and 56.

Basically what it means is that their platforms are going to be more efficient, more productive than they are today. And some of the applications will be able to do things that they were not capable before and we will talk about this in a few slides.

But this just shows you the power of data. For example what happened in the Philippines with some of the weather phenomena’s out there, could have been simulated more accurately and could have prevented some of the effects we have seen happened there. It cost about $10 million to evacuate one-mile of coast in the U.S. for example. So if you’re more accurate, you can save money, you can save lives, you could do many more things. And you can predict the strength and other stuff.

Internet of things we talked about are close, are going to communicate with the data center, our cars. Everything we do is going to be recorded and know -- and you know in few -- not a few but a number of years, this is going to be a doctor. Actually it’s going to have another shape but it's going to tell you, you're limping because you have this. You need to take those pills, you need to do this. They will know exactly what's your situation ahead of time and we’ll try to take care of this.

Security, it's public that the NSA is using us for many things. We were part of the machine that they discovered who the guys in the Boston Marathon were. Healthcare, everything out there needs more data faster and more efficient.

So what we do is we transfer the data from the sensors that capture the things into the computer engines and there are multiple of them, storage and archive and again today archiving doesn't go anymore to tapes and saves. People always want to have the data available so they can make more conclusions.

This just shows that with today's Ethernet which maybe it's hard to see but this is the green bars, you can start seeing them here, very, very small. Ethernet by itself when other vendors does not keep up with the CPU performance and the PCI Express performance, we do with both our Ethernet and our InfiniBand interconnect.

So we really are connecting the most efficient data centers out there and what we've done, we've started our technology building supercomputers. In the past 60%, 70% of our revenues came from supercomputers.

Last year or this year, it's less than 50% and actually storage became our second largest revenue segment for Mellanox. So what we need is we started with supercomputers but now we're taking the same technology and you think about this, Web 2.0 is a supercomputer for one application. Cloud is a supper computer for multiple applications.

So we are taking the same technology that got about 50% of the supercomputer market, now into Web 2.0, into Cloud, into database, into big data, into financials and so on. And now we’re seeing actually in effect of technologies going back to the supercomputers. For example, we’re starting to seeing high performance computers start to use visualization because with our I/O, with our interconnect, the virtualization offload does not have any overhead associated with it and you can get nearly bare-metal performance with that.

We did talk about the data explosions and those are the markets we’re playing in. So first, HPC is our largest market. This year was a down year for us from 2012 to 2013 in terms of HPC. We think it's going to grow into 2014. Web 2.0 grew significantly. We’re shipping hundreds of thousands of 10 gigabit per second Ethernet endpoint sneaks into multiple Web 2.0 customers’ database, Oracle, IBM DB2, SQL, and then businesses just like Teradata, a pilot data warehousing.

We believe also SAP will use us in the future and more are using our stuff. Cloud, you will see well above this, more and more people use this. And in Cloud, it’s mainly Infiniband because it’s a very competitive market, financial services for latency and storage overlay on top of everything.

So if you think what’s going on with storage, we are replacing Fiber Channel and proprietary interconnect in the back end. And now we’re also coming to the front end with 40 gigabit Ethernet and 56 gigabit Infiniband and basically every place you see -- I shouldn’t say everyplace but most if not everyplace, you see flash interconnect it’s with us. This shows with the TOP500 list, the efficiency we bring to the platforms using us. The blue dots at the top are people using us with Infiniband.

You see 80%, 90% efficiency of the systems. The green and the orange dots are people that are not using us. They are usually in the 60%, which is the average of system efficiency utilization. So basically, we give you 50% more, 50% more utilization from the infrastructure we pay a lot for.

So we are in 50% of the Petaflop machines out there. So we have about 50% of the market and it is growing and it's all over the world. It's not just in the U.S. It is in Europe, Asia, the U.S and we like that. This is for Web 2.0 -- in 2011, we started looking into Web 2.0 applications in Q1 of 2011. 2012 was a great year for us with Web 2.0. It is continued to grow and I think we will see more. But you see for example, 2X, 13X, 4X, so it’s not like we're giving you 20%. We are giving you sometimes 4X, 5X performance. So you can do much more things with us, serve many more customers without increasing your cost in a significant ways.

These are some real life examples that without us could not be. A great example is PayPal. They have 4 billion database inserts everyday. Without us, they could not detect if there are frauds or not frauds. With us, we give them enough bandwidth and latency to manipulate work, understand the data with 4 billion. Now this number of 4 billion is going to grow to 40 billion insertions everyday and they will need more capacity, more bandwidth and more machines. We are probably going to be the only guys being able to connect them.

Hy-Vee is a supermarket chain, somewhere in the Midwest I guess, and we reduced their data queries from 20 minutes to 26 seconds -- 10X, very, very significant. There are multiple examples like that. One of things we like is this guy here. There's a Web 2.0 guy that did an operation every so often. And that operation that had to do with data manipulation took seven days, seven days to do this operation. We went down to four hours. We gave them more than six days to go to the beach or do something else with the same equipment, very significant.

So we are not talking about something that's in terms of percentage but in a significant way. Those are the guys that today are using us or allow us to put their names here, are using us for their cloud infrastructure, they are more coming online. And basically to think of that, cloud is a very competitive market in terms of cost of compute and storage infrastructure because that's kind of their COGS. And if it cost less, it means they are going to make more money. So we are seeing more people use us there.

By the way here, most of people use us for Infiniband because it’s competitive. With the Web 2.0, I'd say -- well, some of them, two out of the five most users for Infiniband, but the most users for Ethernet, especially Linux today not yet the switches. Oracle has based most if not all of their products on top of Mellanox, Exadata, Exalogic, Exalytics, ZFS and they are now moving to the new Infiniband stuff. They also bought about 10% percent of the company in the market sometime ago, also participate as their secondary.

Storage is a very big thing for us. Again, I can’t say everybody but, most if not everybody is using us and we are replacing both Fibre Channel and some of the proprietary fabrics in the back end of the data center and of the data devices, data storage devices.

EMC in the Analyst Day, not this year, but last year said we have more than six platforms that were being designed into and that will go to production in the future. But you see us here in the backend it’s IBM XIV, its netapp, its EMC/Isilon today, Data Direct Networks, Xyratex, Oracle, Teradata and more.

And what’s nice, now we are starting to appear also on the front-end of those boxes, both with 40 gigabit Ethernet whether its iSCSI or sometimes also iSER using RDMA for iSCSI or with our InfiniBand we can do both.

This is the market we are playing in. We are talking about more than 14.6 million devices we can connect; 10 million servers, 3.5-storage, 0.5 million embedded. We take into account about $400 ASP per endpoint and that’s a conservative number. We believe with a 100 gigabit per second is going to be more than that. So we are talking about a $6 billion market. We did $0.5 billion last year. We can grow 10x in the future. So, there is a huge growth potential in front of us.

It’s a very healthy funnel we are working with. We serve all of the Tier 1 server companies, now we can say all of the Tier 1 storage companies, some embedded. We work with the operating system companies, virtualization companies, applications and then people using us. It’s a very healthy funnel.

We are very proud to have the full end-to-end solution. So we do not depend on anyone else to drive 56 gigabit per second in the market. It’s not easy to do 56. No one else is doing 40 gigabit end-to-end. We are already doing 56 since 2011. And today we are the only guys that have 40 gigabits Ethernet NICs in production, no one else has that.

So what we have is we have an endpoints, we can sale either silicon that people put on the LAN on motherboards. We have boards plug-in as PCI express form factor or Mezzanine Cards. We have switches both silicon and platforms. We have software. We have the MetroX which expands between data centers up to 10's of kilometers and then we have the cables both copper and fiber optics. So we can really take responsibility from PCI Express to PCI Express.

So our ASP per endpoint is actually growing when we add more of the products connecting from PCI Express to PCI Express, obviously with 100-gig that can have more value, higher ASP, higher gross margins and so on. And again, we’ll start to see more and more people follow us doing the same thing.

When we came out in 2007, before 2007 we were only an InfiniBand company, we just had InfiniBand. But then we decided, some people will not touch InfiniBand, so let's go offensive and also defensive.

So offensive says InfiniBand has great value, let's put all of the InfiniBand offload engines on top of Ethernet layer 2, so we put RDMA over Converged Ethernet, we put visualization offloads, storage offloads on top of layer 2 transport of Ethernet.

At the same time, our defensive move says, guess what, you're buying Mellanox InfiniBand, you log -- you are not log down to InfiniBand. You take the cable out from an InfiniBand switch, plug it to Cisco or anyone else Ethernet, it works as Ethernet, it autosenses, auto-negotiates and get to the right Ethernet rate that you plugged us in.

So offensive means we bringing InfiniBand into the Ethernet market and we are selling hundreds of thousands of 10-gigabit Ethernet because we have the best 10 gigabit Ethernet mixed with RDMA, with offloads into Web 2.0, cloud and other markets.

On the defensive side, we use InfiniBand, you don’t need to be worried about future because if for the future, you want to take the same cable and plug it to Ethernet we support that, and a lot of sense go to Ethernet. So it’s both on the NIC/HCA and the switch, and by the way here we can actually bridge between Ethernet and InfiniBand in the same platforms.

So obviously we have the best InfiniBand out there and the only competitor now is Intel, they are way behind. They don’t have 56. They are actually so much behind, they think that the next version they will support something else.

We have the 40 gigabit -- we actually can run Ethernet in 56. We have like a good Web 2.0 customer in Europe that will start running 56 gigabit Ethernet, not even just 40 -- give them 40% more bandwidth. We are extending beyond the datacenters and now we have MetroX and we can run RDMA between buildings, 80 kilometers apart. It’s a big thing.

We have a large Web 2.0 guy that has multiple datacenters in a campus and they wanted to do breakup since snapshots between buildings were being able to provide them the most efficient team to connect there.

We bought three companies this year for Interconnect, two are public. We bought a small one as well, but basically now we can provide and work even on the 100 gigabit per second, cables both copper, passive and active. We can provide fiber optics VCSEL based with IPtronics driving the VCSELs and we can provide silicon photonics with IPtronics and Kotura, those are the two acquisitions we have done, one in Denmark and one in LA.

So we have the full span of interconnect required for the future for the 100 gigabit per second. And we believe also will have for the 200 gigabit per second because the Kotura silicon photonics can drive 70 gigabits per second x 4, it’s 280. So it is more than 200. So we are kind of have the technology for 200 gigabit per second.

Again the full end to end including the Metro, this is our roadmap. Today we are at 56 gigabit per second. We are going to have a 100 gigabit per second at 2014-15 timeframe. We think we are at least a generation ahead of competition. So we are think Intel is going to be behind us, but not only that even when they come out with a product now, Oracle, EMC, Teradata, the Web 2.0 guys.

Microsoft will need to port -- their software is going to take more than two years to do that. If at all they want to do it because they don’t want to give Intel more real estate on their platforms. And then we will think we are going to get to 200 gigabit per second in the 2016-17 timeframe.

So this year revenues, that been in the past, I guess five years, or six and I want to just say that we went into 2012, our numbers were about $360 million, we did more than $0.5 billion. But really how it should look is this way, and what happened here is the following. In 2011, Romley was delayed by more than nine months by Intel. So about $50 million, that were supposed to be sold and recognized in 2011 moved to 2012 and that’s why here we did only $260 million and here at $50 million more.

Another thing that happened in 2012, two very large entities bought $50 million that is now going to deploy in 2013, so we had $50 million from ‘11, $50 million from ‘13 move into ‘12, that’s why it looked that this way, but this is the real -- but we don’t mind that because the cash is in the bank.

So we put more than $182 million into the bank in 2012. So we are still cash flow positive but much less, but just because of the fact of the $50 million of the Romley delay, $50 million in inventory in 2012 instead of ‘13.

And if you look at this number at the quarterly graph you see this hump here, but we continue to grow and we will able to continue to grow into 2014 moving forward. So we have this hump of $50 million Romley, $50 million inventory, but it's a good growth and will continue to grow into ’14.

And again, what explains is exactly this graph of data growing exponentially, more applications, more devices, they need more pipes. We have about $306 million in the bank and it's growing, and we probably use the cash for acquisitions when we need. It’s not something we need now, but that's what our intention is.

Our long-term model is to be in the mid to, or the low to mid 20s on the bottom line, mid to high 60s in the gross margins and so on. We’re not there. We were there in 2012, way above our long-term model. But because of the topline going down, we did not reduce OpEx, we actually increase because to develop the 100 gigabit per second requires a lot of investment, but in the future we’ll get back to the long-term model.

So basically, we're in a very strategic and important area of the data center of computing and storage. We are growing and all of those applications markets we are playing -- I think, we now with silicon photonics, we have all of the building blocks and we can actually play where we integrate them and we have a very nice financial in terms of gross margins. We need to increase our topline, so that we meet our long-term financial models which we’ll do in the future. Basically, I think we are going to continue growing in the coming years.

So with that, I think we have seven minutes for questions or we can go and drink coffee. Yes, you don’t prefer coffee? Go ahead.

Question-and-Answer Session

Unidentified Analyst

(Inaudible)

Eyal Waldma

What kind of return on capital? So, I’ll give you a rule of thumb I don't have that number. We usually do a project where we think we are going to do 10X our investment in the coming three to five years. So let say, we invest $12 million or $15 million in specific project, we expect to do in around $60 million, $70 million, sometimes even more in the coming three years, sometimes even $120 million that’s our expectation.

Unidentified Analyst

(Inaudible)

Eyal Waldma

Less revenues. That takes 66% or 68% gross margin, I would say about -- let's take 25% earning out of that. Yes.

Unidentified Analyst

Two questions. The first on Intel, they bought QLogic's InfiniBand business couple of years ago and they don’t seem to be investing. What do you think their goal is?

Eyal Waldma

Intel bought four companies to compete with us. They bought NetEffect to do RDMA, Ethernet. They bought Fulcrum to do Ethernet switching. They bought QLogic's InfiniBand and then they bought Cray interconnect, so they bought four companies. What Intel obviously wants to do is to provide the best interconnect for the data center, both for compute and storage. So they are going to come out with something.

We expect something half or semi proprietary Ethernet whatever, which I don't think people will like and adopt because it's not open, it’s almost standard. And they are going to try and intercept the 100 gigabit per second, which is going to be very challenging, put on top of that they need to put all of their offload engines, integrate the software zones, so its going to be challenging.

So we expect them to come with 100 gigabit per second interconnect in the 2015, ‘16 timeframe. We believe we are going to be ahead of the game by two to six quarters and because everything for us really working and out there, Oracle, Teradata, all those guys software, its going to be easier for them to adopt our future roadmap than to move to Intel.

Unidentified Analyst

Yes. Second question on Ethernet versus InfiniBand, is there a point where Ethernet becomes more popular for some of the 2.0 and hyperscale cloud platforms or is this, a kind of thing where IBs going to be around for years and years.

Eyal Waldman

Well, Ethernet today is more popular in the Web 2.0 and cloud. But InfiniBand is taking more and more market share and the reason is its better, even at the 100 gigabit per second, it is easier for us that we do both Ethernet and InfiniBand to do InfiniBand than to do Ethernet, it’s much easier.

We can have higher density when we do just InfiniBand. So an InfiniBand is taking away the fabric channel markets, so the storage is going to go InfiniBand, because there, backend, guess what, you can do with Ethernet, you can do with InfiniBand, with InfiniBand the same speeds and feeds you get 40% more utilization, if lower cost you will use InfiniBand as the backend and it will flow into the best.

So, yeah, InfiniBand is here to stay, InfiniBand is going to grow and take more market share, I think more cloud will use InfiniBand, definitely storage will use InfiniBand, we’re seeing some of the Web 2.0 go back and play with InfiniBand, some of them deploy InfiniBand and more applications. So I think InfiniBand is here to stay and take market share. Yes.

Unidentified Analyst

Could you just explain a bit more about that pre-buy like two customers and why would they, why did they do that, what did they get out of it?

Eyal Waldman

Oh! They didn’t want to do that. It was a mistake. But basically what happened here in, you will see the nice growth from 88 to 133 to 156, there was a big Web 2.0 guy that bought Q2, Q3, $25 million of the group and say they are going to continue buying and continue deploying, what happened in end of Q3, early Q4 today. We’re not buying anymore, something is wrong here. We either overbought and so on. So they did over buy. The OEM that provide them also overbought and some of this now being deployed in 2013. So it wasn't on purpose.

Unidentified Analyst

Yeah. Okay. That’s the second questions. So given what you know, when are they back in the market?

Eyal Waldman

Well, we think, kind of it’s behind us. We think they have depicted their inventories. Okay. Okay, guys. Thank you very much. Have a good conference.

This entry passed through the Full-Text RSS service — if this is your content and you're reading it on someone else's site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.

from SeekingAlpha.com: Home Page http://seekingalpha.com/article/1874071-mellanoxs-ceo-presents-at-nasdaq-annual-conference-transcript?source=feed