Docoh
Loading...

NVDA NVIDIA

Participants
Colette Kress Chief Financial Officer
Harlan Sur J.P. Morgan
Call transcript
Harlan Sur

Okay. Why don’t we go ahead and get seated, because we are going to start in about a minute here.

Okay.

So, good morning everyone and Happy New Year. Welcome to J.P. Morgan’s 18th Annual Technology and Automotive Investor Forum here at the Consumer Electronics Show.

For those of you that don’t know me, my name is Harlan Sur. I am the semiconductor and semiconductor capital equipment analyst for the firm. Very pleased to have a solid day of semiconductor, automotive, automotive technology, general tech companies presenting. Also joining me here from J.P. Morgan are some of my colleagues, I've got Bill Peterson, who covers the RF semiconductor value chain; I've got my colleagues Ryan Brinkman and José Asumendi, who cover the U.S. and European automotive sectors; Paul Coster, our Applied & Emerging Technologies analyst; and Samik Chatterjee who covers our IT hardware and telecom and networking equipment sectors.

We continue to expand the companies that participate at our conference and for the first time at our CES Conference this year, we have management teams from Microchip, Qualcomm, Xilinx, Cree, Gentherm, Gentex and Lear.

So, we hope you enjoy the conference. Remember this conference is for you guys.

So ask questions. We’ve got a great line-up of companies. And with that, I am very pleased to introduce Colette Kress, the Chief Financial Officer of NVIDIA. It’s been a tradition now for the past, I think five or six years now that the NVIDIA team kicks off our investor conference, because the team is driving much of the trends that you’ll hear about today artificial intelligence and deep learning, next-generation compute architectures in automotive, next-generation gaming technology, just to name a few of the things. I’d ask Colette to kick us off with an overview of what the team is showcasing here at CES or maybe GTC – the recent GTC in Shanghai and then we’ll go ahead and kick off the Q&A.

So with that, Colette, thank you again for joining us this morning.

Colette Kress

Okay, thank you, Halan for the introduction and we are excited to be here and be your kick off and I think I am going on six to seven years actually.

So, I think we are forgetting on how many times we’ve been here. But we’ve seen a great acceleration.

As you know, we are an accelerated computing company.

Our focus is in many different markets and let me talk about some of the highlights that we have seen. Starting first with gaming, which I think is a great place to start since we are here at the Consumer Electronics Show.

Our gaming business is doing quite well.

Our introduction of RTX, which has been in the market for more than a year is off to a great start.

We have more than a couple dozen games in the market today focused on overall ray tracing. Ray tracing is what we believe the next wave of overall high-end graphics and high-end graphics that can be used for gaming but a lot of other different places.

We are here introducing at CES more laptops and notebooks focused on the overall RTX capabilities, but also our overall Max-Q.

You will probably see here we have now reached a total of more than a 140 laptops focused on gaming, focused on RTX that you will see new ones even here today. We’ve talked about our overall cloud gaming and our introduction of cloud gaming with many of our partnerships with the overall telcos and our GTC in China in December, we announced another partnership with Tencent, bringing overall cloud gaming to an important market there in overall China.

We are also here at CES announcing on a new form of our overall G-SYNC, which is exclusive with many of the overall LG monitors that are out there. It brings a higher overall hertz or about 360 hertz to overall playing games in a very competitive environment such as overall esports. We know that esports is again a very important part of what we see in terms of gaming around the world and has taken off to on great, great strides.

Additionally, focusing on not only our overall gaming business, but focusing on accelerated computing in terms of what this means in terms of the datacenter.

We have announced a great acceleration of what we are seeing right now both in the overall hyperscales, increasing their overall purchasing and a lot of that has been fueled by a very important industry of conversational AI with the underpinnings of natural language processing. Recommendation engines are also very key.

So we are off to really focus with our TensorRT, overall software and our inferencing to influence a big part of this overall growing market.

We also focused in China GTC, we focused on overall edge computing.

So our EGX overall platform which we will focus is on are probably next cloud platforms.

You’ve seen some major wins when we think about the edge focusing on companies such as Walmart, as well as the U.S. Postal Service that is using our overall platform to help in many of those areas with either warehouse management or just focus in terms of the overall supply and demand management that we see.

Moving forward in terms of a key area in terms of automotive. Over in China, we announced in terms of our DRIVE platform with DiDi. But we also announced our overall next-generation architecture Orin, which is probably 7x faster than our current overall architecture focused on Xavier. These are platforms that are both available for work with the overall automotives today and we have one of the true end-to-end platforms focused on automotive that takes everybody all the way from the datacenter and focusing on training, simulation of the data for - later for to be in terms of production inside of the car. We’ve had a lot of other great announcements that you’ll see here, but I’ll just start with there with a quick summary.

Harlan Sur

No, no, thanks for the great introduction. It’s well, what a difference a year makes. Last year, at this time, we were talking about cryptal inventories. We were talking about the slowdown in the datacenter business.

You entered calendar year 2019 with your business down about 30% year-over-year.

If you hit your guidance for this quarter, you’ll actually exited calendar year 2019 at plus 30% year-over-year. Consensus says you guys up 20% in calendar year 2020, which is your fiscal 2021. Help us understand the trends and product cycles that are going to drive this fiscal year, fiscal 2021, and longer-term, how should we think about the overall growth profile for the different businesses, gaming, datacenter, proviz and automotive?

Colette Kress

In the perspective of overall gaming, we are very, very well positioned.

We have – as we always do a full lineup that allows gaming at a lot of different overall price points.

We are the provider now that is focusing on overall ray tracing.

First overall market, and been in market for a year and also even upgrading what we had in terms of our existing turning RTX cards to incorporate our overall SUPERs. We went into the overall holiday season knowing that this would be a great holiday seasons for the games that were coming out.

You will see in terms of games such as Call of Duty, and Wolfenstein also coming to market using our overall ray tracing capabilities.

So as we move forward, we are setting the way for ray tracing to take us through the next-generation of gaming, but not only focus just on the overall desktop.

As you know, we are bringing ray tracing to the overall notebooks opening up a market of probably untouched of the mobility of overall gaming at a high-end overall performance incorporating our overall Max-Q. And then lastly, we know that the world of cloud gaming is an important area still in its overall infancy. But we are the best to bring that to market with many of the different telcos and won many of the overall partnerships that can overall focus.

So we will continue to see the importance of gaming as an entertainment overall industry, as well as our platform being leading edge.

Now, when we focus in terms of on datacenter, datacenter and its opportunity, we are probably approaching a new wave of overall AI.

If you went back to the early 2012 timeframe, that was a timeframe that people were bringing to market the use of GPUs for overall image detection, image categorization and actually quite excited about it, because it did changed a lot of key apps. What we have seen as early as the overall summer time is the evolving of natural language processing with Google and its blog in terms of its overall BERT model, in showing people how they are using a technique to better understand how search commands are created and how better to solve this is an important expansion of both natural language processing not only for the overall hyperscales, but for startups, for consumer internet companies, but also many of the overall enterprises.

You’ve seen now a wave of focus on a conversational AI and we are only just beginning. This is something that probably takes a 100x more computing performance than what we have seen in overall history. It’s a focus of not only on the training, but also on the overall inferencing side.

You’ve seen our growth in terms of inferencing. Inferencing is now solidly within the double-digits as a percentage of our overall datacenter business. But you also have a case where we are shipping more overall inferencing cards than what we are seeing in terms of overall training.

So we have a great path going forward, not only with the conversational AI, but the recommendation engines broadly overall using GPUs.

Harlan Sur

So, let’s stick with the datacenter. It’s roughly a quarter of your overall revenues.

You will be exiting this year as you mentioned with the business trending back to kind of 30% year-over-year growth rates after a three quarter pause by cloud, hyperscale, enterprise customers. We had anticipated this business coming back in the second half of 2019 as your cloud and hyperscale customers started their next wave of the spending cycle. But with the NVIDIA team as you mentioned, it’s more than just spending reacceleration what is just new wave of AI. There are new applications like natural language, conversational AI, you’ve got the ramp of your inferencing business, just the emerging business in edge computing. Can you just talk about all three of these dynamics and the sustainability of 20%, 30% type year-over-year growth rates to calendar year 2020, as well as beyond that?

Colette Kress

Sure.

So what we had talked about with our Q3 results is, we did start to see the improvement in terms of the overall purchasing of the overall hyperscales.

As we had focused earlier on the year, many of our engineering engagements we knew were intact and we knew we would get to the point post the overall digestion to focus on overall new compute. We see the hiring continue in many of these areas, hiring in terms of the engineers of the hyperscales.

You see the overall AI papers and you see a tremendous amount of start-ups focusing on AI.

So shoring off in the second half of the year, we have started to see the comeback of the overall hyperscales.

Now, the return of the hyperscales is not just focused on overall internal use. The overall need for the overall cloud computing, for those that don't have the capabilities to build out their older infrastructure. It allows that opportunity for them to begin their work in the cloud, they may remain in the overall cloud. But we're seeing both of those.

We are seeing in terms of our overall inferencing platform, RT4 to be used worldwide and it is actually in almost all of the different datacenter locations in terms of within AWS and I think with the other hyperscales, we will slowly start to see that also happen. But we have broad-based adoption of using the overall GPUs for inferencing. This is an important moment of something that we knew we were working on an incumbent of CPU or a traditional type of overall inferencing that a CPU is just fine. What we are seeing in terms of the future of inferencing is the complexity, the complexity that natural language processing requires, the complexity of the amount of data that needs to be mined over a short amount of period of time and a return in a couple hundred milliseconds will require a GPU or a high performance overall processor to do that.

So we are well positioned, positioned not only on the training, but also on the inferencing side.

Harlan Sur

You know, the first wave of AI adoption with your strong product cycle was image classification and recognition in classification and this drove three strong consecutive years of growth for your datacenter business.

Given, as you mentioned, the order of magnitude, compute and training complexity required for some of these natural language, conversational AI models, and on top of that, the requirements for real-time inferencing. Does this conversational AI have the potential to, again, drive a multi-year, kind of strong growth cycle for NVIDIA’s datacenter business?

Colette Kress

We do believe conversational AI is a next very large wave of overall growth that we can see going forward. We're in the early stages of conversational AI, both understanding the use cases for it, but also the complexity that it still has to add more things to it.

For example, we are really focused right now on a single language.

We are focused either on overall English or in some of the overall international. But there is still a large piece to focus in terms of that translation in that overall processing.

So, looking at conversational AI and the use cases, it’s not just capable for overall hyperscales.

You see an opportunity of moving this to the enterprises everything in terms of focusing on their callcenter, focus on that first dialogue that you have in terms of communication and using AI to streamline the overall work that they do.

Harlan Sur

On the inferencing side, you have had big success with your T4 inferencing platform. Inferencing is now double-digits percentage of your overall datacenter business.

Now that you’ve got time to understand this market and looking out over the next few years, do you think that your inferencing business can approach or even exceed the size of your training business over the next kind of three to five years?

Colette Kress

So, inferencing, we do believe is probably a larger market than what we have in terms of the other pieces.

So, we are continuing to concentrate on this. Inferencing in the overall datacenter is important, but remember, also inferencing at the Edge.

Harlan Sur

Right.

Colette Kress

Inferencing Edge computing, but also Edge in terms of autonomous devices will be very key.

You are seeing us focus not only on the processor for those different capabilities, but also the software stack.

One of our focus areas in the last six months, the last nine months is unique overall software packages that really address the industry’s overall applications that are out there.

You see us focus in terms of CLARA. CLARA is our software capabilities and focusing on overall healthcare. Metropolis, Metropolis is our overall software platform to deal with in terms of the places of the world that you maybe not just focused on cities, but anything inside of a building or otherwise.

You see us focus on Isaac. Isaac is focused on our overall robotics platform and thinking about the autonomous piece.

You can also of course think about our overall DRIVE platform.

Our DRIVE platform is focused in terms of automotive. Each and every single one of them is important to influence the acceleration not just in the datacenter, but in the Edge and what is required to help that is the connection with the overall application. And that is our work that we have really been focusing on.

Harlan Sur

You rolled out another vector of the product platform which you mentioned several times, it’s your EGX platform. Can you just talk about this? It seems like it’s more of a data analytics platform that’s GPU accelerated versus an AI-based platform. But this type of analytics has wide applicability across a number of different verticals. Can you maybe just help us compare contrast your Edge computing platform versus your cloud and datacenter platforms?

Colette Kress

Sure. It is a form of thinking about what is necessary at the overall Edge to focus in terms of a cloud that doesn’t have the capabilities to allow for it to go all the way back to the cloud in terms of that overall functioning.

So you are looking for a form of computing that either is needed in the milliseconds to respond and/or incorporates overall data that is confidential in nature that you don’t want back in terms of the overall cloud.

So these are some of the things that we may see in terms of healthcare, in terms of devices, scanning devices or otherwise that you could do the overall scanning and have the results in instantaneous amounts of time rather than what we have seen in the past of that entire industry.

In terms of what we see already in use of end store capabilities and what you see inside of the store is in terms of both checkout.

You could look at the U.S. Postal Service in the same manner. How quickly can I move the overall mail through by using overall AI capabilities than their overall manual function.

So this is something that we believe the Edge platform will be an enormous part of the type of compute that will happen in the years to come. We’ll see every single device focusing on a overall ability to accelerate and we believe our platform is very well positioned for that.

Harlan Sur

Great.

Before I move on to the gaming, I just wanted to see if there are any questions in the audience. If there are questions, I would just ask for you to wait for the microphone.

Okay, well why don’t we move over to the gaming side of your business? 50% of your overall business, you entered this year with the business down 40%.

You are exiting it with it being up 50% year-over-year. But Q4, from a sequential basis is going to be down, primarily on notebooks and your console business, just normal seasonal trends. But desktop is driving a strong upgrade cycle with your RTX SUPER platforms. The ramp of your new GTX SUPER mainstream platforms and a strong lineup to go along with that of blockbuster games that support your ray tracing technology. Sticking with desktop for a moment and talking with your Board level partners, game developers, retailers, what’s your sense on the demand pool this holiday season? And as you look into the lineup of games for 2020, how do you see the demand trends for this coming year?

Colette Kress

Sure.

So, we entered into the holiday season. The holiday season starts as early as November. Holiday season for example, starts with Singles’ Day on 11/11. A very important start of that overall holiday and we're actually not done.

We are waiting now for Chinese New Year which is usually at that tail end of our overall holiday season. But each of the overall points that you have probably seen through discussions with the overall channel and the partners is it's actually going quite well. It’s going quite well and through our overall expectations that we had planned for the quarter, as we do have a full lineup available.

We have a lineup that now focuses on ray tracing and bringing ray tracing down into the upper $200 of a retail price for you to overall be able to appreciate the overall ray tracing capabilities. But this also allows us to influence all types of overall sales, not just focus on desktop.

As you know, we have focused in terms of on the notebooks.

Our notebooks over the last several years have continued to outgrow what we have seen in terms of the desktop.

Our overall notebook business is approaching approximately one-third of our overall gaming business as a whole.

So an important piece, and you'll see a significant amount of that out here at CES. The overall OEMs have focused on this notebook as an unique niche for them to attract the overall gamers. The gamers that aren’t necessarily interested in building their own overall desktops, but now can have that high performance in their overall notebook design.

Our overall ray tracing even from a desktop standpoint has now accumulated to be more than 60% of our overall sales on the desktop and is making a good path in terms of with the notebooks as well.

Harlan Sur

Got it. And then, - so let's talk about – let’s talk about the cloud gaming segment of the market.

So, NVIDIA has its GeForce Now cloud gaming platform.

You’ve announced some partnerships with regional global telco service providers. At the recent GTC in China, NVIDIA and Tencent announced the partnership where you will be supporting Tencent, the largest gaming company in China, largest gaming company globally by providing them with your GPU technology to power their cloud gaming platform, can you give us more insights into the partnership? Will you be supplying your GeForce Now architecture? Or are you just going to be supporting them more with your GPU products? And when do you expect this service to kind of go mainstream?

Colette Kress

So, right now, our relationship with Tencent has always been there as a continuation of not only focusing on gaming in China, but also focusing in terms of in the datacenter. This was an opportunity for us to align with a very important partner in overall China. Tencent, here to endorse the growth of overall cloud gaming even though we are still in the early stages. They have right now overall pods focused on building out into their overall datacenter and will come to market soon in terms of their testing. Testing the overall first pass of some of the software that is necessary, and will come out later in terms of where they are in that overall relationship ready to go overall live. But we are excited about this as an important partnership, a partnership that allows us to both respect the overall ecosystem as they continue to address the overall gaming market. It’s a win-win for both Tencent and NVIDIA in cloud gaming.

Harlan Sur

And actually, that's a good segue, because Tencent, like I said is the largest gaming company in the world but they are also the largest gaming company in China. And I always bring up China in the conference calls, because it is a largest gaming market in the world, 700 million gamers in China, obviously, the home of Tencent and many of your key datacenter customers are in China, as well. I remember, when you started off calendar year 2019, Tencent specifically called out weak demand trends in China, but China demand, especially gaming seems to have come back strong.

So, help us understand what’s driving the China demand?

Colette Kress

Yes, I think a year ago, probably in a different space in China did have probably some slower quarters than we had expected. But where we stand right now, things are going quite well in overall China. China is an important part of our overall gaming business, but also very important part of our datacenter business as you inferred. But what we see there is the adoption of our full stack.

Our full stack not just focused on in terms of our ray tracing, but we also sell cards into that market that are important just for overall gaming as a whole but allows them to have both the new architecture even if they may not be overall ray tracing. It’s an important piece when you think about the overall iCafe market there. The iCafe market is one of its own not necessarily something that’s always worldwide, but important part of overall China.

We continue to see them looking for the overall upgrade opportunities of our platform and we continue to hold a significant market share in China against many of our peers.

Harlan Sur

Before I move on to autos, is there any questions in the audience on either Gaming or Datacenter segments of the business? Okay, why don’t we switch to Auto? Auto is roughly 5%, 6% of your business but with a long tailwind and expected to grow strongly as the market transitions more and more towards autonomous driving.

You guys announced strong success with your Xavier SoC platform and as you mentioned at GTC in Shanghai, you announced your Orin platform. Help us understand the differences between Orin platform, help us understand the difference between Xavier, Orin and maybe discuss the timelines on when we should start to see first production volumes of your AV platforms with your major auto OEM partners?

Colette Kress

Great.

So, we announced our next overall architecture for the SoC and focusing on a big piece of that on our overall automotive partnerships.

So, Orin is a probably about 7x speed overall compute capacity in terms of what we had with Xavier. And Xavier was probably top of line against any overall competitive view of it even from a standpoint of their PowerPoint’s that they may have had of something in the future. We know the importance of the overall data collection, the data processing that is going to be necessary inside of the AV cars as we move forward.

So, we see this as something that we’ll have in the hands of our overall auto customers in the couple years ahead, but an important piece in terms of finishing out that full platform.

Our platform view and focus in terms of AV cars is not just focused in terms of on the production, it’s focused, as well as back in terms of the datacenter.

You’ve seen us focus on creating the overall ability for them to overall test their platforms with both simulated data and/or live data. There is a significant amount of training the overall data for these overall AV cars.

We are seeing what we’ll call is a fork in the road of the type of AV that we are working on with many of our overall OEMs focusing on AV that will focus on high-level of the two moving to level three, but also on the other side, focusing on robo-taxis or focusing on level four and overall above. We can do these two things separately with level two plus working on high-end passenger cars and looking at level four with removing the overall driver in what we’ll find is confined overall scenarios.

Now what we are seeing in the overall market for automotive is it’s taking them a little bit longer. The investment that they need to focus on this is something that’s going to draw them out, probably closer to the 2022 timeframe versus a little bit earlier than what we have had previously communicated.

So, we are still working with them.

We are probably one of the only ones with a full platform today that they can build upon, test upon of something that would actually be inside of the production car.

Harlan Sur

Great. Any updates for us on the proposed acquisition of Mellanox? I know you are waiting for China approval.

Colette Kress

So, with our Q3 results, we had announced that we believe that the closure of our Mellanox acquisition would probably fall into the first couple areas of 2020 and even since then, we have reached overall regulatory approval with the overall EU.

With the EU, with unconditional and nothing holding us back in terms of when we actually close.

So, yes, we are waiting on China. And again, we do believe this will close in the first part of 2020.

Harlan Sur

First – so, half of 2020? Is that…

Colette Kress

The first part of 2020, we have to see.

Harlan Sur

First part of 2020.

Okay, last question on the financials. There is a big step-up in gross margins last quarter, 64%. The kind of, back to the gross margin levels during the first half of 2018, part of that is normalization of the business, you guys are past the inventory overhang. The higher margin datacenter business is starting to inflect and gaming continues to perform very, very strongly into the holiday.

So, how do you – how should we think about the sustainability of the 64% plus, sort of gross margin profile going forward?

Colette Kress

Yes, so, our gross margin has always been influenced probably by the biggest piece being our overall mix. The mix starts even with our overall gaming business and with inside our overall gaming business.

We have many high-end overall gaming, overall cards and platforms that influence just the overall growth of gross margin.

We continue to see an opportunity for people to both upgrade and upgrade choosing a higher-end overall card. But the overall ray tracing capability also allows us to appreciate higher gross margins. Then you discuss in terms of what datacenter and our pro biz does to our gross margin as well, now that we’ve seen a return in terms of the growth.

Harlan Sur

Yes.

Colette Kress

Both from a training and/or an inferencing standpoint that can, as well overall influence our overall gross margins.

As we move forward, we will continue to concentrate and see that growth likely improve in the future. But it will probably change a bit quarter-by-quarter with the mix.

Harlan Sur

Great. Well, with that, we are out of time. Colette, thank you as always for participating.

Colette Kress

Thank you so much. I appreciate it.

Harlan Sur

I appreciate. Yes, thank you.