Greetings, and welcome to Advanced Micro Devices' Fourth Quarter and Full Year 2018 Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to turn the call over to Laura Graves. Please go ahead.

Thank you, and welcome to AMD's Fourth Quarter and Fiscal Year 2018 Conference Call. By now, you should have had the opportunities to review a copy of our earnings release and slides. If you have not reviewed these documents, they can be found on the Investor Relations page of AMD's website, www.amd.com. Participants on today's conference call are: Dr. Lisa Su, our President and Chief Executive Officer; and Devinder Kumar, our Senior Vice President, Chief Financial Officer and Treasurer. This is a live call and will be replayed via webcast on our website. I would like to highlight some important dates for you. Dr. Lisa Su will present at the Goldman Sachs Technology and Internet Conference on Tuesday, February 12. Also, Ruth Cotter, Senior Vice President of HR Worldwide Marketing and Investor Relations will present at the Susquehanna Annual Technology Conference on Tuesday, March 12; and our 2019 first quarter quiet time will begin at the close of business on Friday, March 15, 2019. Today's discussion contains forward-looking statements based on the environment as we currently see it. Those statements are based on current beliefs, assumptions and expectations, speak only as of the current date and, as such, involve risks and uncertainties that could cause actual results to differ materially from our current expectations. We will refer primarily to non-GAAP financial measures during this call except for revenue, gross margin and segment operational results, which are reported on a GAAP basis. The non-GAAP financial measures referenced herein are reconciled to their most directly comparable GAAP financial measure in today's press release, which is posted on our website. Please refer to the cautionary statements in our press release for more information. You will also find detailed discussions about our risk factors in our filings with the SEC and, in particular, AMD's quarterly report on Form 10-Q for the quarter ended September 29, 2018. Now with that, I will hand the call over to Lisa. Lisa?

Thank you, Laura, and good afternoon to all those listening in today. 2018 marked another year of strong financial performance, driven by our expanded high-performance product portfolio despite near-term graphics weakness. We grew annual revenue by 23%, with Ryzen, EPYC and data center GPU product revenue growing by more than $1.2 billion for the year. Our new products gained share and significantly expanded gross margin, leading to our most profitable year since 2011. Looking at the fourth quarter, revenue of $1.42 billion increased 6% from a year ago, with approximately 65% of sales coming from our new products. We reached an important milestone in our business in the quarter as our high-margin data center CPUs and GPUs accounted for a mid-teens percentage of overall revenue. While we expect our data center revenue to be lumpy, the ramp of our data center business is beginning to contribute meaningfully to our financial results. Looking at our Computing and Graphics segment, we delivered our eighth straight quarter of year-over-year segment revenue growth. Sales of Ryzen desktop and notebook processors and data center GPUs offset lower GPU sales as the channel continued working through elevated levels of graphics inventory. Client processor unit shipments grew by more than 50% from the year-ago period. We had our highest client computing revenue in more than 4 years, and we believe we gained client CPU unit share for the fifth straight quarter. At CES, Acer, Asus, Dell, HP, Huawei, Lenovo and Samsung, all launched notebooks powered by our new second-generation Ryzen mobile processor with Radeon Vega graphics. The second-gen Ryzen mobile processor delivers more performance, enhanced features and longer battery life than any mobile processor we have ever built and is the fastest processor available for ultrathin notebooks. The industry's first AMD-based Chromebooks launched earlier this quarter from…

Thank you, Lisa, and good afternoon, everyone. 2018 was a strong year for AMD. New product introductions wrought the highest annual revenue since 2011 and a significant improvement in gross margin year-over-year. Earnings per share increased from $0.10 per share in 2017 to $0.46 per share in 2018. Full year 2018 revenue was $6.48 billion, up 23% year-on-year, driven by strong performance in the Computing and Graphics segment with significant growth in Ryzen processor sales. Although there was weakness in the graphics channel in the second half of the year, we saw strength in data center CPUs and GPUs. Gross margin was 39%, up 440 basis points from the prior year. Gross margin improvements are primarily driven by our new Ryzen, EPYC and Radeon products. Operating expenses were 29% of revenue, an improvement of 2 percentage points from the prior year. For the full year, operating income was $633 million, up $409 million from $224 million in the prior year. Net income was $514 million, up $411 million, compared to net income of $103 million in the prior year. On the balance sheet, we reduced principal debt by $171 million and improved gross leverage significantly from 4.6x a year ago to 1.9x at the end of 2018. Let me turn to the details of the fourth quarter results. Fourth quarter revenue gross margin, operating margin and earnings per share all improved year-on-year. Quarterly revenue of $1.42 billion was up 6% from a year ago. Strong sales of Ryzen and EPYC processors and data center GPUs more than offset lower channel GPU and semi-custom sales during the quarter. Fourth quarter 2018 revenue did not include any IP-related revenue, and blockchain-related GPU revenue was negligible. Gross margin was 41%, up 740 basis points from 34% a year ago. Year-on-year gross margin improvement…

Thank you, Devinder. Operator, we're ready for our first question, please?

[Operator Instructions] Our first question today is coming from Toshiya Hari from Goldman Sachs.

Lisa, I had a question on your full year 2019 outlook. You guys are guiding revenue growth kind of in the high single-digit range. Embedded in that outlook, what kind of growth rates are you assuming for your core businesses, computing, graphics, semi-custom and server CPUs? And then on the gross margin side, 41% or greater, is the improvement year-over-year primarily a function of revenue growth and mix? Or is the amendment in the WSA playing a role there as well?

Sure. Thank you for the questions. So relative to the full year guidance of up high single digits, the primary growth drivers are Ryzen and EPYC, are the 2 largest growth drivers. With data center GPU also, we're expecting that to be up. And then we would expect that semi-custom will be down. If you look at the life cycle, semi-custom will be in the seventh year of the console cycle, and so we expect semi-custom to be down approximately 20% and then we would expect consumer graphics to also be down, let's call it, double digits as we really burn off some of the channel inventory that we see in the first half of the year. So Ryzen and EPYC are the large drivers of the growth. And that's as we launch our 7-nanometer products throughout the year. And then your second question, Toshi, was?

Yes, in gross margins, 41% or greater, that's 2 percentage point or greater improvement year-over-year. Is that primarily a function of revenue growth and improvement in mix, given Ryzen and EPYC? Or is the WSA amendment providing a tailwind as well?

Yes, it's primarily due to mix. So the mix of the higher-margin new products, Ryzen, EPYC, and the data center GPUs.

Next question, please?

Our next question is coming from Matt Ramsay from Cowen and Company.

Lisa, I guess, I wanted to attack Toshi's question on the full year a little bit differently and maybe just talk about a little bit what share gain assumptions you guys are embedding for your 7-nanometer road map versus maybe some pragmatism or conservatism on the market just given the macro. I mean, Intel talked about a flat PC TAM in terms of units. And obviously, it's only, I guess, guided us to mid-single-digit growth for their sort of expansive server business for the year, and you guys are starting off down maybe 25% year-over-year in revenue for the first quarter given what's going on in the graphics business. So maybe you could just walk us through a little bit what you're thinking about the TAM growth of your 2 main growth markets for the year versus share gains you're embedding.

Yes, absolutely, Matt. Look, I think, as it relates to the market, I don't think we would have a very different view of the market as others may have stated. Our story really is a share gain story. And when you take a look at sort of the progress that we'd made in 2018 and the design wins that we had in 2018, as we go over to 2019, when we look at both Rome as well as Ryzen, we look at sort of the breadth of OEM platforms that we have, the customer engagements by workload and sort of how we see that progressing. We feel very good about the opportunity to gain share as we go through the year, particularly given how competitive the products set is. Similarly, on the PC side, we've made nice progress over the last 4, 5 quarters. On Ryzen, we started strong on desktop, and then the last couple of quarters, we've made good progress on notebook. We see a broader portfolio with our OEMs as we go into 2019, and we also see a more competitive desktop product line as well as notebook product line. So that's how we're thinking about the year. To your point about starting a little bit with lower guidance in Q1, I think that's true. When we look at Q1, particularly on a year-on-year basis, there are a lot of moving pieces. But it's primarily due to the graphics channel. And that's both the reduced demand as well as the absence in blockchain revenue on a year-over-year basis as well as the semi-custom business. But as we move forward, as we go into Q2 and beyond, we see a significant opportunity with the ramp of our new products, and that's how we see the year.

Got it, that's helpful. Devinder, maybe just a couple of little things in terms of licensing gains and IP revenue, just I got asked a few times, and I just want to clarify here that the gross margin outlook for the first quarter does not include any IP or licensing gains. And then, for the full year, are you including any IP revenue in the revenue outlook, and are we -- how should we think about the rest of the THATIC licensing gains applied to sort of timing for when those might get recognized? I know there's a lot there. But lots of questions on just the moving pieces there tonight.

Yes, I can take that. License gain essentially does not impact gross margin. That sits on a separate line altogether on the licensing gain line in the P&L. So the gross margin is not impacted by that. As far as IP revenue is concerned in Q1, there's no contemplated IP revenue, so the gross margin at 41% guide is based on the products that we have. And for the year, at this point, outside of the THATIC JV IP licensing gain that we are expecting or recording in Q1, while there might be opportunities from an IP standpoint, but nothing substantial contemplated at this point.

Our next question is coming from Vivek Arya from Bank of America Merrill Lynch.

Lisa, I'm curious as you look back at 2019 and the success you had with EPYC -- the initial success with EPYC. What was the mix of cloud versus enterprise? And then how do you think it trends in 2019 because of all the concerns around slowdown in cloud CapEx and so forth? And as part of that, if you could share with us what your market share assumptions are as we exit the year on EPYC.

Yes, sure, Vivek. As we look through 2018, we were pretty pleased with our progress on EPYC, and coming off of the fourth quarter, actually, it was a fairly strong fourth quarter for us and the fact that we doubled the number of units for our server business. And when you look at that mix, it is more cloud-weighted. So we had some large deployments that went online here in the fourth quarter, and that was positive for us. That being said though, we're making nice progress in the Enterprise and HPC side of the business too. We've had a number of wins in the quarter as well as going into 2019. So as we look into 2019, I would expect that the early Rome deployments will also be cloud-based. We'll be the first ones, but we have a strong set of enterprise platforms, and as I mentioned earlier, it's the breadth of the OEM platforms that gives us good confidence that we're going to a broader set of workloads and having broader coverage in the market. In terms of share assumptions, we'll have to see how the market and the year play out, but I think what we've said before is that, after reaching the mid-single digit market share in the fourth quarter 2018, we would expect it would take another 4 to 6 quarters to reach 10% market share. And I think we're still in that range.

Got it. And for my follow-up, there were 2 things that came out on your competitors' last few public discussions, which is, one, this concept of CPU shortages, and I'm wondering if that has had any positive or negative effect on your strategy or positioning. And then Intel has also mentioned being a little more tactical when it comes to pricing, and I'm wondering how that figures into your full year outlook? So both the impact of CPU shortages and pricing, any color would be very helpful.

Sure, Vivek. So look, on the CPU shortages, my comment is that there are some pockets of shortage, particularly at the low end. However, our focus has really been on ramping Ryzen. And if you look at the Ryzen percentage of our overall business, Ryzen, Devinder mentioned, was 80% of our client business. So we are really actually improving our mix in the client business. I think that, again, from my view, the shortages are temporary. But we look at it as really getting consistent share gain. And as we've gone through each quarter in 2018, we've seen consistent share gain, and we believe we gained share in the fourth quarter as well. So I think we're well positioned with the portfolio, and we'll certainly drive that into 2019. And then, the other?

Pricing.

Oh, pricing, yes. Relative to pricing, look, when we think about pricing and when we put together our long-term model, we have factored in that pricing and competition will be aggressive. I think we know that to be the case. What we've seen in the market is consistent with that, but we also believe that particularly as you go into the data center, the single largest factor from a buying decision standpoint is the performance and the total cost of ownership of the product, and we believe that Rome will be very, very well positioned in 2019. And so we are cognizant of the competitive environment but feel good that we have the right set of assumptions.

Our next question is coming from Mark Lipacis from Jefferies.

Great to hit that mid-single digit milestone exiting the year. Lisa, as you start -- I just wanted to confirm, you expect to ship -- start shipping Rome midyear. And as you do, what's the biggest risk for Rome? Is it execution on delivering the product? Or is it the sales the process, convincing customers to take EPYC 2 in volume?

Yes, so I think the -- with Rome, our biggest opportunity/risk is adoption rate. I think, from a competitive standpoint, the product is very solid. Everything going through our development labs looks very good as well as our customer engagement. And so this is just about getting customers to production as fast as possible. We do expect though that the adoption rate for Rome will be faster than the adoption rate for Naples. And the reason for that is we are in a socket-compatible infrastructure. And so customers who don't necessarily need the newest features of the platform can actually use the same motherboard and system that they currently have with Naples and drop in Rome. And so I think that will help us accelerate some of the adoption, but right now, it's about helping customers in their environment. We are widely sampling Rome, and there is a lot of work to be done. But we feel good about the trajectory.

Great. And a follow-up, if I may, for Devinder. So inventories, it looks like they were up about $100 million, going into a pretty healthy seasonal quarter that's declining. What should -- how should we read that, Devinder? And was there just some opportunity to get some wafers more cheaply? Or is there -- are you trying to stockpile product in front of launches, or -- and when can we expect the inventory to be a source of cash on the cash flow statement?

Yes, Matt (sic) [ Mark ]. So the inventory growth is primarily driven by our new products and in preparation for our 7-nanometer next-generation product launches later this year, as you just mentioned. We overall expect to manage the inventory in line with revenue but also in support of the new products that will be launching and ramping throughout the year. Lisa mentioned a few of the launches at CES and also in the script, and we prepared to support that from an inventory standpoint. And we'll see how the year progresses and then manage it from there.

Thank you, Mark. Next question please, operator?

Our next question comes from Stacy Rasgon from Bernstein Research.

Around the Q1 guide, obviously, you mentioned the GPU weakness, but you also mentioned seasonality across the businesses. Is my assumption correct that, that means everything is basically down sequentially, CPUs, GPUs, servers? Can you tell us, is that true? And if so, can you give us an idea of the magnitude, like at least what is down, I guess, most versus least, sequentially, Ryzen versus GPU versus EPYC?

Sure, Stacy. So the largest driver, as I've said, is the GPU channel situation on a year-over-year basis. On a sequential basis, typical seasonality for, let's call it, PCs and GPUs and data center is probably around 10% or so. We are forecasting -- or we believe that that will be a little bit -- down a little bit more than that. And in terms of the ranking, I would say GPUs are down a bit more than the other 2 segments. That's the way we currently see it.

Got it. And for my follow-up, I had a question just on, general, EESC. So its revenues are flat year-over-year. The server revenues are up a bunch and it's still losing money. And I get it, there's some incremental investments and everything, but I guess, how do we think about the profitability of that segment? How big does EPYC or Rome with the server business need be to get before it's sustainably profitable, especially as semi-custom is going to continue to decline, you said, 20% in 2019?

Yes, I think the key, Stacy, is going to be ramping the server business in 2019. You are right that year-over-year we look at it, it's lost little bit of money. But fundamentally, it's investing in the go-to-market programs and launching the new products, investing in the ongoing engineering work that's needed to make sure the customers bring out the product on time. And semi-custom, given the fact that it's in the seventh year of the cycle, we expect it to be down from 2018 to 2019. But we do expect to go ahead and ramp our server business, and that will definitely benefit the EESC profitability in 2019.

Do you think it's profitable for the full year?

Too early to tell. I won't -- I don't want to predict profitability at the segment level, but we'll see how it evolves over the year.

Our next question today is coming from Ross Seymore from Deutsche Bank.

Lisa, a potentially different way to ask the same question on the full year 2019 guide. Late last year, at a few different events, you talked about the server business being second-half-weighted. If I go through the remainder of this year, it seems like you have to grow the better part of 20% plus sequentially in each quarter to get there. I know it's not going to be perfectly linear like that. But I guess, my overall question is, if the server side of the business is back-half-weighted, what sorts of businesses are helping the middle portion of the year? Is there some snapback on the GPU side? Is there something going on in the Ryzen side? Just trying to get somewhat of the shape of the year off of a base that's admittedly pretty low as your first quarter starting point.

Sure, Ross. So yes, I think it is fair to say that our quarterly progression will have a lot to do with our product launches. So you would expect that server will be more second-half-weighted than the first half. In general, our business is more second-half-weighted, though, if you think about the consumer portions of our business. But think about it as we would expect sequential growth in PCs. We would certainly expect sequential growth in data center, although stronger in the second half. We would -- we're also -- as we see the GPU business right now, we see the first quarter as the low point in the business, with the channels getting -- improving as we go into the second quarter. And we have additional product launches there as well. So that's the way we would see the portfolio. And semi-custom, although it's lower on a year-over-year basis, we would expect it to also increase as we go from second quarter into third quarter as well.

Perfect. And a question on the GPU side, your big competitor in that market obviously is having some current issues and mentioned that the new product at the high end wasn't selling through. Are you at all concerned that the competitive landscape in that market, whether it's because China demand being weaker, as they cited, or them having to be a little bit more aggressive on pricing would do anything to your ability to penetrate the market with either existing or new GPU products as the year progresses?

Yes, so, look, when we look at the GPU market, and let's separate sort of gaming and data center, I think, on the gaming side, from what we are seeing, we did see sellout increase in Q4 versus Q3. So gamers are still buying GPUs. They may be more discerning about price points, and so I can imagine that there might be a bit more softness at the high end versus in the mid-range. But we believe that we have a good understanding of what's happening in the gaming side of the business. And it will be driven -- our gaming growth will be driven by new products. And we would see that as we go through this year with our Radeon VII launch as well as our Navi launches on the gaming side. On the data center side, we're making good progress in GPU data center, obviously from a low base, but the GPUs in the data center are very workload-dependent. There are some workloads that actually we do very well in things like cloud gaming and virtualization, and we have some early HPC wins. And so we see this as sort of customer acquisition redeployment. And so those are the sources of growth on the PC side for us.

Our next question today is coming from Aaron Rakers from Wells Fargo.

Just 2 questions, if I can, building on that last one. In your prepared remarks, you noted that you've reached the milestone of which GPUs as well as data center CPUs reached a mid-teens percentage of your total revenue. I'm curious as you kind of contemplated your full year guidance, where do you think that mix of business can go? And is there a point in time where we actually -- you foresee us kind of getting granularity on how big the data center GPU business is trending?

Yes, Aaron. So I think the mid-teens percentage revenue for us in Q4 was -- was a good milestone. I mean, I think that's a meaningful percentage of our revenue. And it was a contributor to our gross margin as we exited 2018. I think, as we go into 2019, again, we will continue to give you visibility into where the data center growth is. To be fair, my expectation is that the CPU side of the data center business will grow faster than the GPU side, just given what we see in terms of overall deployments. But I think both businesses will be meaningful for us and our key part of the growth story for us in 2019, and we'll continue to give color on how they develop as the year develops.

Okay, that's very helpful. And just real quickly on the semi-custom side. As you kind of contemplated the guidance as well through the course this year and that 20% decline for the full year, is your assumption that Q1 kind of marks the bottom and we can grow seasonally off of that level? Or should we think about something differently from a progression through the quarters in '19?

Yes. Aaron, yes, that's the correct way to think about it. So Q1 will be a low, Q2 will be higher, Q3 will be higher, and then Q4 will come down again. That should be the seasonality of semi-custom.

Our next question is coming from John Pitzer from Crédit Suisse.

Lisa, notwithstanding that the macro is extremely uncertain, making any predictions difficult. I'm just kind of curious, going back to Ross' question, how do you see sort of the cadence of growth sequentially throughout the year? Because as he pointed out, if you kind of divided it up equally by quarter, you're growing north of 20% in Q2, Q3, Q4, half-on-half growth, it would be somewhere around 50%. If you assume more seasonal, and seasonal is a little bit difficult to get to because of the change of accounting, I think you can get half-on-half growth as high as 75%. So as you think about the full year growth, is this something that's more like 50% half-on-half or 75% half-on-half?

Yes, thanks for the question, John. It might be a little bit early for us to go there. But what I would say is the following. When we look at the year-on-year compares from 2019 to 2018, Q2 2018 was still a very strong quarter for us because we still had a significant amount of blockchain sales in Q2 2018. So I expect that we will grow sequentially into Q2, but I think that will still be, let's call it, a tougher year-on-year compare. But then when we get into the second half of the year, we expect to be fully ramped on the entire 7-nanometer portfolio. And so we would see a heavier weighting in the second half. So we see sequential growth into second quarter but more heavily weighted into the second half.

That's really helpful. And then, Devinder, on the profitability front, if you kind of run the numbers from Q1 guidance to kind of what you need to be exiting Q4 at to hit the full year number, you could see revenue up well over $1 billion, Q1 to Q4. I'm a little bit surprised that despite that, the gross margin guidance isn't a little bit better. I know you guided the full year to greater than 41% with Q1 at 41%, but I'm kind of curious, if you look at a Q4 run rate that's north of $1 billion higher than the Q1 run rate, how should we think about gross margin and operating margin exiting the year?

I think, John, if you look back the last many quarters, I think 7 quarters year-on-year, we've improved gross margin on the strength of the newer products. We ended the year quite well in 2018 and achieved gross margin in the fourth quarter that was nice. And now we start the year with 41%, and as we start the year, it's early guiding greater than 41%. I think you should expect that as the products ramp, in particular with the 7-nanometer product that Lisa referenced, our margins will continue to improve, and we'll see where we end up at by the end of the year. But we do expect that, year-on-year, margin goes up from where we ended in 2018 at 39% and greater than 41% for 2019.

Our next question is coming from Mitch Steves from RBC Capital Markets.

I have 2, quickly. First, on the C&G side, I realized that there's going to be a lot of kind of sequential difficulties there, but when I think about the long-term growth rate, I know you guys have kind of talked to kind of high singles to double-digit growth. Is that still essentially where you think that business can grow at long term? And then secondly, in terms of the next EPYC 2, is there any reason why it would underperform an Intel 10-nanometer in a testing environment? And if so, why would that be?

Let me take the second question, and then maybe, Devinder, you can take the first question. So as it relates to our Rome product, when Rome was originally planned, we had planned that it would be competing against a 10-nanometer product. So that was the -- that was our expectation when we started. I think we've shown some of the generational performance advantages in terms of double the performance on a socket basis, 4x performance on a floating-point basis. And the other thing I would say is that, with our second-generation Rome, the customer set is now more used to our architecture, and so some of the architectural and software improvements that we've also spent quite a bit of time on over the last 4 to 6 quarters are coming into play. So I think we feel very good about the competitive positioning of Rome. And the other thing to keep in mind is we are deep in development of our next-generation beyond Rome. So the Zen 3 product portfolio is deep in development as well. So that's -- our goal is to ensure that we have a consistent cadence of very competitive products.

Yes, and then as far as your question on the segment growth is concerned, relative to high-single-digits growth on a company level, in both of the segments, there are some factors that come into play. In Computing and Graphics, obviously, we feel good about the data center side of GPU. But on a compare basis, when you look at 2018 to 2019, we do have the impact of the blockchain and some memory sales from 2018 that don't happen in 2019 from a standpoint of where the market is. And then on the EESC side, semi-custom being down obviously is a factor when you compare year-on-year per segment. But good growth on the data center piece of it, in particular, but with the 7-nanometer Rome product launching in the back half -- I mean, ramping in the back half of 2019. And that's the way you kind of look at it. It is still early in the year from an overall standpoint. And where we ended up in 2018, we're projecting high single-digit growth at the company level for 2019 over 2018.

Our next question is coming from Blayne Curtis from Barclays.

Lisa, I just want to ask here, you laid it out a little bit, but in terms of EPYC 2, same socket, but -- and you have sampled it early. But in terms of getting over the finish line and getting these shipped, I guess, the customer still has to wait for production silicon. And then -- and typically, people talk about 6 to 12 months so they can kind of test it and make sure it does everything with the production silicon. So can you just walk us through the timeline because, obviously, you're baking in a pretty steep ramp in the second half?

Yes, so we started sampling EPYC actually last year, so the second half of last year, to some of the top cloud vendors as well as our OEM partners. We've had a significant amount of development with our ODM partners as well. So I think there is a good amount of overlap between the customer development cycle and our development cycle. Now as you said, production silicon is very key, and ensuring that the final specifications are locked in is work that we still have to do here in the first half of this year. But I think we feel good about it. We feel good about the platforms, the customer engagements, the current progress of development and so that is very much our plan from an execution standpoint.

I just want to ask on the GPU side. It looks like your ASP was up substantially with the data center. I was just kind of curious, as you think about the mix of data center, do you think -- how do you think about the trajectory of that business? I mean, is it -- is that portion lumpy as well? Or is this kind of the go-forward kind of level after that pretty sharp initial shipment in December?

Yes, so on the data center side, we do expect that the data center GPU business will be a growth driver for us in 2019. It will be a little bit lumpy so I wouldn't say it's a straight line. But I would say, on a year-over-year basis, it's a important growth driver for us. And as it relates to the overall data center GPU ASPs affecting overall ASPs, I think that's true because the gaming ASPs are -- I'm sorry, the gaming business was lower than expected given the channel inventory situation. So I think it's also just weighted in terms of the units.

Thank you, Blayne. Operator, we have time for 2 more questions, please.

Our next question is coming from Joe Moore from Morgan Stanley.

It seems like the GPU data center business is doing pretty well. Can you give us some color, if you look at mid-teens exposure overall between the 2 data center businesses, CPU and GPU, can you give us some qualitative sense for how much of that is GPU at this point and what are the main applications? Is it mostly virtualization? What other applications are you seeing there?

Sure, Joe. So for the mid-teens percentage of revenue in Q4, between CPU and GPU, it's actually roughly similar. And then in terms of the workloads, the GPU data center side, we do very well in cloud gaming, we do very well with our virtualization solutions, and we have some early traction in HPC.

Great. That's helpful. And I guess, just in terms of the business segmentation, I think you alluded to this earlier, but it seems like the conversation typically is around microprocessor, graphics and semi-custom as sort of 3 different opportunities. Have you thought about moving the segmentation to more align with that? And I realize that's easy to talk about and hard to implement, but what's your thought in terms of, over time, sort of migrating that more in line with the way you guys talk about the business?

Yes, so we look at the business in all different cuts. As you can imagine, we look at it by segment, our CG and EESC. CG is very much a PC-driven segment and EESC with enterprise and semi-custom. And we are also, based on some of the feedback, trying to be a little bit more granular in terms of how the various businesses fall underneath that. So I think as the data center business becomes a larger piece of our business, we'll continue to look for how do we get -- give you guys more color and more transparency in how that is doing.

Our next question is coming from Vijay Rakesh from Mizuho.

Just in terms of the GPU gaming side, you mentioned higher inventory levels. I was wondering if you can give us some number on where normal inventory GPU levels are and where are they running currently.

Yes, so when we look at the GPU channel inventory, we do believe our channel partners reduced inventory in Q4, and they will reduce inventories in Q1. We still think that Q1 levels are elevated, and there will be some spillover into Q2 where we will see some elevated inventory levels. We'll have to see how the sell-through really plays out. But my expectation is that, in Q2, we'll have sort of improved channel inventory levels and we will return to a sequential growth in the gaming side of our business.

Got it. And on the 7-nanometer side, is the 7-nanometer GPU and CPU, as those ramp in the back half, are they accretive to this 41% gross margin levels?

Yes, so the 7-nanometer GPUs on the data center side started ramping in Q4 so that was part of the data center GPU revenue that we talked about. The 7-nanometer CPU will ramp middle of the year, and the 7-nanometer CPU and GPU are both above corporate average margins so you would expect that they're above the 41 points and would be sort of accretive to margin as that mix improves.

We've reached the end of our question-and-answer session. I'd like to turn the floor back to over to management for any further or closing comments.

Thank you very much, operator. We appreciate everyone joining us here today. We're quite proud of the accomplishment in 2018 and look forward to seeing you all on the road in 2019. Have a great afternoon.

Thank you. That does conclude today's teleconference. You may disconnect your lines at this time, and have a wonderful day. We thank you for your participation today.