慧榮科技 (SIMO) 2018 Q3 法說會逐字稿

Ladies and gentlemen, thank you for standing by, and welcome to the Silicon Motion Technology Corporation Third Quarter 2018 Earnings Release Conference Call

This conference call contains forward-looking statements within the meaning of Section 27A of the Securities Act of 1933 and Section 21E of the Securities Exchange Act of 1934 as amended.

Such forward-looking statements include, without limitation, statements regarding trends in the semiconductor industry and our future results of operations, financial condition and business prospects.

Although such statements are based on our own information and information from other sources we believe to be reliable, you should not place undue reliance on them.

These statements involve risks and uncertainties, and actual market trends and our results may differ materially from those expressed or implied in these forward-looking statements for a variety of reasons.

Potential risks and uncertainties include, but are not limited to: continued competitive pressure in the semiconductor industry and the effect of such pressure on prices; unpredictable changes in technology and consumer demand for multimedia consumer electronics; the state of and any change in our relationship with our major customers; and changes in political, economic, legal and social conditions in Taiwan.

For additional discussion of these risks and uncertainties and other factors, please see the documents we file from time to time with the Securities and Exchange Commission.

We assume no obligation to update any forward-looking statements, which apply only as of the date of this conference call.

(Operator Instructions)

I must advise you that this conference is being recorded today.

And I will now hand the conference over to your first speaker Mr. Jason Tsai.

Thank you, please go ahead.

Thank you, and good morning, everyone, and welcome to Silicon Motion's Third Quarter 2018 Financial Results Conference Call and webcast.

My name is Jason Tsai, and with me here as well as Kou, our President and CEO; and Riyadh Lai, our Chief Financial Officer.

The agenda for today is as follows: Wallace will start with a review of our key business developments, Riyadh will then discuss our third quarter financial results and provide our outlook.

We'll then conclude with Q&A.

Before we get started, I'd like to remind you of our Safe Harbor policy, which is read at the start of this call.

For a comprehensible review of the risks involved in investing in our securities, please refer to our filings with the U.S. SEC.

For more details on our financial results, please refer to our press release, which is filed on Form 6-K after the close of market yesterday.

This webcast will be available for replay on our website, www.siliconmotion.com for a limited time.

To enhance investors' understanding of our ongoing economic performance, we will discuss non-GAAP information during this call.

We use non-GAAP financial measures internally to evaluate and manage our operations.

We have therefore, chosen to provide this information to enable you to perform comparisons of our operating results in a manner similar to how we analyze our own operating results.

The reconciliation of the GAAP to non-GAAP financial data can be found in our earnings release issued yesterday.

We ask that you review it in conjunction with this call.

With that, I will turn the call over to Wallace.

Thank you, Jason.

Hello, everyone, and thank you for joining us today.

I will first update you on our business, and Riyadh will review our financials and provide our outlook later on this call.

We are pleased to report third quarter revenue of $138.6 million, which is flat sequentially, but up 9% compared with the same quarter last year.

Earnings per ADS this quarter was $0.95, an increase of 68% compared with last year.

In the third quarter, while overall sales were flat sequentially, sale of our client SSD controller increased 35% sequentially.

Compared to the last year, our SSD controller sales this quarter are almost 80% higher.

Secondly, and as expected, our SSD solution declined in the third quarter due to timing of the technology transition at our major hyperscale customers.

Our eMMC controllers sales were soft, primarily due to smartphone market weakness.

Our client SSD controller sales growth continued to exceed our own expectation at the start of the year and then about 3 months ago.

We now expect our client SSD controller sales to grow 35% for the full year.

Growth has been a fair rating with the falling price of NAND, which make SSD increasingly affordable in PC and other client devices.

This quarter, sales of both our NAND flash partners and module maker was strong.

But especially, sales to our flash partners.

While NAND prices have fallen a lot since the start of the year, the price of NAND is essentially just back down to early 2016 levels.

The price of client SSD today are also roughly where they were 2.5 years ago, though moving lower swiftly.

We believe the price of the NAND and the price of SSD can fall a lot further, especially since the costs of NAND today is significantly lower than where they were 2 years ago.

Today, 64-layer 3D NAND are ready in high volume production.

And NAND maker have started initially production of 96-layer NAND.

Additionally, Intel and Micron are already in volume production of their 64-layer 4bits per cell QLC NAND.

And we expect majority of NAND maker to begin production of 96-layer QLC next year.

The NAND industry is currently in oversupply.

And the price of NAND will continue to fall rapidly.

With the NAND industry supply expected to continue to grow, we believe current oversupply conditions will continue likely through the middle of next year, and possibly, even through the second half of next year.

We believe client SSD demand will continue to grow rapidly as the NAND prices continue to fall.

Approximately 300 million client HDDs are still shipping annually for using PC, game consoles and other client devices all displaceable by client HDD.

This quarter, we are proud to help both Intel and Micron bring to the marketed world first client PCIe NVMe SSD using 64-layer QLC NAND.

We believe PC OEM are extremely interested in these low-cost high-performance SSD using QLC NAND.

And we believe QLC based HDD shipments will scale more meaningfully early next year.

Furthermore, you should expect to see even lower costs and higher production volume of QLC NAND, when 96-layer QLC NAND enter production next year.

This will further reduce the costs of NAND and SSD, and further accelerate the adoption of SSD in the client devices.

QLC NAND is harder to manage compared to TLC NAND, because of worsening data retention.

Lower endurance, slower read/write data rates, and other issues.

We are extremely well-positioned with our QLC controller technology, and believe QLC further extended our technology leaderships.

Additionally, our highly configurable hardware plus firmware controller platform held by our NAND flash partner quickly and cost-effectively cover -- convert NAND component into competitive value-add SSD solutions.

Today, a large percentage of our SSD controller projects with outside partners involves QLC NAND.

Now allow me to update you on the progress of our data center and enterprise NVMe and open-channel SSD controllers.

We are happy to announce that our controller has passed final testing with 2 hyperscale customers.

And SSD using these controller are scheduled to begin initial production this quarter.

However, initial shipment will be small, as SSD solutions using this controller will still need to undergo live data center application testing before they are rolled out extensively in the second half of next year.

Turning to our eMMC+

UFS controller.

Sales this quarter were down modestly.

The result, smartphone industry built weakness as well as sales lumpiness relating to smartphone OEM, transitioning from eMMC to UFS mobile-embedded memory.

We believe UFS adoption will accelerate next year.

When leading mobile chips vendor introduced mainstream application processor that supports uMCP, which should combine user controller with NAND and mobile DRAM in a single BGA packaging.

We are excited that smartphone OEMs have already extensively design flagship devices using UFS and are actively bringing many of these UFS phone to market.

Our user design activity as our flash partner continue to grow.

Already, our user controllers will account for more than 10% of our overall eMMC+

UFS sales this year.

We expect our UFS controller sales to continue to grow rapidly next year, as the market transitioning further from eMMC to UFS and uMCP and our customer design activity continues to pick up.

As part of the transition from eMMC to UFS, we are also diversifying our customer concentration from sales primarily to SK Hynix, to a more balanced customer mix, which also meaningful contribution from Micron, a Chinese flash maker and module makers.

Now let me discuss about our SSD solutions business.

Sales for this sector declined sharply.

Sale relating to our new project were delayed and will be further delayed.

We have been transitioning our Shannon data center SSD by a BGA-based controller to AC-based that include the open-channel and standard enterprise grade NVMe controller that I have previously talked about.

Our open-channel controller recently passed testing at our 2 hyperscale customer.

And our Shannon SSD are literally days away from passing final qualification before initial production.

We expect these open-channel SSD will be testing live application with the objective of production to scale in second half of next year.

Let me summarize by saying that, while overall sales this year will be lower than our original projected, caused primarily by large SSD solutions project delay.

We have made up of some of delay SSD solutions sales by our stronger-than-expected SSD controller sales.

We expect our SSD controller sales to continue growing strongly through next year.

Given our large pipeline of projects and very favorable tailwind of declining NAND price trend.

All our SSD solutions after a brief delay, we have delivered a major milestone and our SSD solutions should start scaling beginning around the beginning of next year.

Now let me turn the call over to Riyadh to discuss our financial performance and outlook.

Thank you, Wallace, and hello, everyone.

I will summarize our financial results and then provide our outlook.

Before I begin, I would like to reiterate that our comments today will focus primarily on our non-GAAP results, unless otherwise specifically noted.

A reconciliation of our GAAP to non-GAAP data is included with earnings release issued yesterday.

Our Q3 revenue was flat sequentially and in line with our guidance.

For Q4, we expect revenue to be in the range of $120.5 million to $127.5 million, down 8% to 13% sequentially.

Let me point out that while our revenue this year will be lower than expected, our gross margins are higher than expected because of stronger SSD controller sales.

And with operating expenses tightly controlled, we anticipate our full earnings to remain consistent with what we were expecting at the start of the year, and in the ballpark of analysts consensus.

I will now walk you through the performance of our key products before taking you through key elements of our P&L and the rest of our financials.

In Q3, sales of our SSD controllers grew approximately 35% sequentially and nearly 80% year-over-year.

In Q4, we expect our SSD controller sales to be stable sequentially, but up 50% year-over-year.

For full year 2018, we expect our client SSD controller sales to grow 35%, much stronger than the 20% we were expecting at the start of the year, and better than the 30% growth expected 3 months ago.

We expect strong SSD controller sales to continue through next year.

In Q3, sales of our eMMC+UFS controllers declined over 5% sequentially and stable year-over-year.

In Q4, we expect our eMMC+

UFS controller sales to also decline modestly due to weak smartphone market demand and our transition from eMMC to UFS.

We believe our eMMC+

UFS controller sales next year will likely be flat to down, depending on the speed at which we can scale our UFS controller sales relative to declining eMMC usage in a mature zero growth smartphone market.

In Q3, sales of our SSD solutions declined approximately 40% sequentially.

In Q4, we expect our SSD solutions sales to decline further and rebound in the middle of next year when our new Shannon SSDs are expected to ramp up scale.

Moving to gross margins.

Gross margin increased from 47.5% in Q2 to 51% in Q3 due to strong sales growth of higher gross margin SSD controllers and a decline in lower gross margin SSD solution sales.

We expect our Q4 gross margins to remain stable in the range of 50% to 52% because higher gross margin SSD controller sales as a percentage of sales mix will increase, and blended gross margin of our other products will decline a bit.

Operating expenses increased from $30.2 million in Q2 to $33.1 million in Q3, a more normalized level of spending compared to a year ago.

As discussed last quarter, Q2 operating expenses were light because of lower R&D project tapeouts and other expenses.

Our operating margin increased from 25.7% in Q2 to 27.1% in Q3 because of higher gross margin, partially offset by higher operating expenses.

For Q4, we expect operating margin in the range of 25% to 27%.

Total headcount at the end of Q3 increased to 1,292, which is 19 more than at the end of Q2.

Our effective tax rate in Q3 was 12%, higher than the 10% in the previous quarter.

While our long-term model tax rate remains at 15%, recently, we have been benefiting from a series of short term and onetime tax gains.

In Q4, our effective tax rate should fall to 11% due to a onetime R&D tax credit.

In Q3, EPS was $0.95, higher than $0.92 last quarter, and significantly better than the $0.57 a year ago.

In Q3, stock-based compensation in our operating expense, which we exclude from our non-GAAP results was $4.5 million, higher than the $0.7 million in the previous quarter due to the seasonal timing of RSU awards.

For Q4, we expect stock-based compensation to increase to between $12 million and $13 million, again, the result of seasonal timing of RSU awards similar to prior years.

We had $308.2 million of cash, cash equivalents and short-term investments at the end of Q3.

$56 million less than in the previous quarter and $25 million less than a year ago.

Cash flow from operations generated $30 million in cash in Q3.

In Q3, we had $63 million of CapEx, $4 million from routine purchase of software and design tools and $59 million for the purchase of land for our future office building in Hsinchu.

Let me spend some time discussing our rationale for the purchase of land for our future office building.

For quite some time, we have been evaluating the consolidation of office spaces in Hsinchu, some of which we own and others we rent.

Especially, since our engineering headcount has grown and continues to grow, though recently, at a more measured rate.

Additionally, our currently owned and lease office space does not allow for meaningful headcount expansion in the future as space is close to being fully utilized and availability of additional existing office space nearby is very limited.

Also, we believe consolidating and engineering resources in one site improves productivity and communications.

Recently, we located an empty lot next to the Hsinchu high-speed rail station, which we can use to construct our own office building.

Operating from that site will half the travel time between our Hsinchu and Taipei offices, through the use of the high-speed rail.

This 66,000 square foot piece of land was purchased this quarter for $59 million.

On this land, we are planning a building with 360,000 square foot of office space.

This building will be sufficient to held our current 600 employees in Hsinchu as well as future hires.

We anticipate construction to begin late next year and construction to take 3 years at an estimated cost of approximately $70 million.

Total land plus construction costs will be approximately $130 million.

Since the market for build-to-lease financing is not well developed in Taiwan, we need to sell finance and manage the construction of our office building.

Once the building is complete, we plan to enter into sale and leaseback arrangements with the finance company.

Since we are a established semiconductor company, we believe a sale and leaseback arrangement is consistent with our capital-light business model.

Additionally, once we move into our building, we anticipate selling the office space which we currently own that has a current market value of between $30 million and $35 million.

In August, we paid $10.8 million for dividend to shareholders, the fourth $0.30 per ADS quarterly installment of our annual $1.20 per ADS dividend that was announced in October of last year.

Recently, on October 26, our Board declared a new $1.20 per ADS annual dividend to be paid in quarterly installments.

Please note that while the company's profitability, cash flow and long-term outlook continues to be strong, the Board has decided to maintain the same annual dividend as last year, given the current volatile global economic conditions and uncertainties.

Our Board will revisit our annual dividend in January 2019, with timing to coincide with our annual operating -- our operational planning, so we can better align decisions about our dividend with business, earnings and cash flow goals and objectives.

This concludes our prepared remarks.

We will now open the call to your questions.

(Operator Instructions) Our first question comes from Anthony Stoss from Craig-Hallum.

A 3-part question.

I just wanted to confirm, Wallace, that you mentioned the Shannon Systems part that's been delaying the project, that it will be qualified in a few days.

And secondly, Riyadh, maybe you can highlight the amount of revenues that you're expecting in your Q4 guide from Shannon Systems.

And lastly, given where the stock is trading, I think it's below where you guys had highlighted in the past of potential share buyback, no commentary whatsoever on share buybacks, I'd love to hear your view on that.

Yes, the Shannon SSD solutions, we're in the final stage, I think in the qualification, what brings you the past 3 to 4 months very intensely qualification.

And I will go to the live occasion through a data center.

We believe we'll start to ship small volume this year and that this we will bring to different business units, and they will start to ramp up from probably late Q2 and middle of next year.

Tony, let me also address your second and third question.

We don't break into granular details specifically relating to our Shannon revenue.

But what I can say is, our SSD solutions sales declined quite sharply in Q3, declined approximately 40% in Q3, and we expect to continue to decline sharply in Q4.

To your other question about share repurchase.

Before I move to share repurchase, let me also say that our SSD solutions, after declining to lower levels in Q4 will stable -- will be stable going into next year and should stabilize at the lower levels in Q1 and Q2 before starting to rebound in Q3 of next year as our new Shannon SSD move into production and ramp in scale.

So moving on to your third question about our share repurchase.

As you know, our Board has authorized share repurchases -- repurchase programs in the past.

But you should expect that our primary way of retiring capital to shareholders to be through our dividend.

We will also utilize our cash flows to explore M&A opportunities as well as small tuck-in acquisitions that could make sense to extend our expertise and leadership in the storage industry.

If I may, a follow-up, again, related to Shannon Systems, then.

It's been quite a few periods probably over the last year missed execution on the Shannon Systems side, I know the reasons or the rationale behind the acquisition was to get you entry into China.

Is there anything you can do to tighten up the management of that operation or give us a sense that this isn't going to be an ongoing surprise every other quarter?

No, I think the open-channel is also very new in China market.

We work closely with our partner.

So the fine-tuning performance not only we rely on our own firmware, our own driver, also depend on the hyperscale, their server and their host software fine-tuning.

So this is really very major milestone for both companies, naturally, with our partner together to reach that performance.

And this will be, well, after we finish, I think say we can see the very, very strong momentum scalability to provide new revenue growth.

We have got a reassurance from our partner, the hyperscalers are leading in China, they really want to move to open-channel, that's their long-term plan.

And there is no -- really no hesitation when the product is ready, they're going to ramp up quickly.

But Tony, let me also add to what Wallace had said.

Since you've got the full explanation of -- to your question.

Most of our SSD solutions technical issues have been fixed.

This relates to the project that we've been talking about.

We still need to further tune the performance of our products for the hyperscalers host server and application, but we feel good about entering production in mid of 2019, and we expect our Shannon business to return to growth again next year as these programs ramp.

Our next question comes from the line of Karl Ackerman from Cowen.

I had 2, please.

As sort of a follow-on to this -- the Shannon Systems business, I was hoping you could help us frame the opportunity and visibility for your Ferri solutions office and industrial markets.

Does that rebound at a faster rate than your Shannon system business?

And I have a follow-up, please.

Yes, the traction we are seeing with our Ferri product is accelerating.

And our grasp of customer and end-market continue to grow rapidly.

Ferri will continue to be an important growth driver for SSD solution business in long-term.

Ferri addressed the market that needed robust industrial or commercial grade SSD, as these product tend to have long qualification cycle, but sale to this OEM tend to last for several years as well.

As in we are confident that we'll continue to see strong demand for FerriSSD for automotive and surveillance system and networking and enterprise server boost storage.

The reason we slowed down a little bit this year is due to NAND transition from 2D NAND to 3D TLC NAND, that qualification take longer for end customer.

We have high confidence we're going to rebound very strongly for 2019.

I appreciate that.

As a follow-up, how are you thinking about the gross margin equation -- excuse me, how are you thinking about the gross margin equation as the open-channel -- as your custom SSD business and Ferri solutions business advances from here?

I mean, how sustainable is this uplifting gross margins?

Particularly, in light of NAND ASP deceleration that should extend into the first half of '19, why wouldn't margins remain into the 50s?

Our margins for our SSD solutions, whether they are Ferri or Shannon have consistently, in the past, been below corporate average.

And we expect them to continue to be below corporate average because by definition, a big part of the build material for these sort of SSD solutions are coming from the NAND components.

NAND is a -- accounts for a very large part of the build materials.

And as a result, the gross margin for these products are below corporate average.

And so you should expect the gross margin for these products to remain below corporate average going forward.

But what you should be aware of is, we blend lower gross margin products against our higher gross margin products, higher gross margin products being are controllers in general.

And we've been blending down in order to drive earnings accretion when -- where it's possible.

Our next question comes from the line of Rajvindra Gill from Needham & Company.

The open-channel opportunity at your 2 hyperscaler customers, I'm wondering if you could characterize the importance of this transition to open-channel, number one.

And number two, how that will affect your business in terms of higher ASPs for your controllers?

And any kind of sense in terms of the revenue opportunity for Shannon when the open-channel opportunities start to come to fruition?

So let me explain about our product plan and game plan for open-channel.

Open-channel, we codeveloped our hyperscaler partner in China with the 2 leading hyperscaler.

But if you take the NAND per open-channel as it allow the cloud maker, they can monitor their health of the NAND components.

So in the past, the cloud maker can only treat SSD solution to product as a one black box.

Now they can go to the open-channel to handshake boot driver to the host level, they can monitor each NAND component independently, and then, can repartition NAND components into multiple functions and reduce the latency.

So they can really tell then to decide which media move to where, when they can move around the data.

And because you know the 3D NAND, the single die is from 512 gigabit move to 1 terabit.

And we're going to see 2 terabit single die in 2020.

So when you put a die into one BGA, that's about 1 terabyte or 2 terabyte.

And we don't need to mention the SSD solutions could even move to 16 terabyte -- sorry, 2 terabyte.

So yes, it's really expensive if cloud maker can only view SSD as one black box.

They like to see more down to a component level.

So that need tremendous different software handshake and housekeeping and maintenance between the controller side and versus solution move to the system layer with the host developer.

That will help them to save the money and also can more efficiently manage their data.

Because they now the data the best, then the controller, then the SSD solutions.

Raji, let me also add, our open-channel is part of our broader data center plus enterprise SSD controller strategy.

We have the opportunity of driving economics directly relating to our controllers and also, then, participating in the economics of the entire SSD module that we're codeveloping with the hyperscalers, the products that they need to go into their data center.

At the same time, by working with the hyperscaler, we're able to greatly tune our SSD control that we can then offer as a standalone to customers who don't need the solution from us, where we can just sell the controller and generate the economics as a controller merchant supplier.

That's helpful.

On the eMMC UFS business down modestly and kind of flat to down, next year forecast.

Trying to couch the risk there.

And how you're thinking about the acceleration of UFS and the impact to your business with SK Hynix and possibly offsetting that with Micron and the Chinese flash maker, do we expect that -- those kind of share shift in either direction, either positive or negative, to kind of accelerate next year?

And when you're providing kind of the flat to slightly down guidance, what assumptions are you making in terms of share shift with respect to SK Hynix, with respect to Micron, Chinese flash makers and module makers, et cetera?

I think the primary eMMC business, we're doing with Hynix is the eMMC.

The main point of UFS we're doing today is with Micron.

We do see UFS transition now become stronger due to Qualcomm promote the UFS solution from flagship premium lines and really transition to mainstream, the Snapdragon 600 family.

So we believe uMCP really will become main driver to move the mainstream smartphone adopt UFS controllers.

But in the same time, we also want the other China flash maker to start eMMC as well as the UFS simultaneously.

And we do see China several module maker this year will start to ramp up eMMC as well prepare UFS for next year.

So we see, we're probably going to maintain the more balanced position next year.

We see Micron UFS traction is accelerating and UFS sales are growing quickly this year.

But they are of small base.

We believe UFS will account for more than 10% this year, it will be more than double next year for our eMMC UFS combination.

And we expect our eMMC+

UFS sales to be flat to down next year.

It all depends, eMMC, how fast they transition into the UFS.

But we believe around the ballpark, we should see around a flat.

While the Hynix portion decline, Micron portion increase, our module maker increase, our China -- the other flash maker increase, so all beneficial to SMI.

That's helpful.

Just a follow-up, though.

Clearly, as you just said, Qualcomm is promoting UFS for mainstream phones, and so your business with SK Hynix has primarily been in mainstream OEMs.

So if they adopt UFS and if SK Hynix is using an internal UFS solution, could that have a greater impact next year compared to your ability to offset that with Micron and others?

We see that like MTK, their mainstream still remain eMCP.

So Qualcomm although they promote UFS more, but they also support eMCP.

So it's really up to end-customers to decide which standard they want to choose.

Our next question comes from the line of Suji Desilva from Roth Capital.

As you talked about the SSD market looking into '19, can the year-over-year growth in the '19 market for SSDs accelerate or remain consistent with '18 growth?

And what quarter is in 2019 does the 96-layer NAND become available to help benefit you guys?

We believe SSD, client SSD growth will be continued through 2019.

And our gross rate definitely will be above the industrial average growth.

And because through the current designing pipeline with NAND OEM, we are confident we can maintain the growth rate through 2019.

96-layer TLC, as the majority NAND maker just start enter small production.

But through the first half of 2019, all the 6 NAND maker will enter full production for 96-layer.

For QLC, we're going to see -- 96-layer QLC, we're going to see at least 4 to 5 maker enter production by Q2 next year.

So we believe this is going to help and further reduce the costs for client SSD.

And the value line, we believe by end of 2019 majority value line SSD will be all used QLC NAND but with 96-layer.

Okay.

That's helpful.

And then on the enterprise SSD used by hyperscalers not SSD solutions, but the enterprise SSD, kind of -- that adoption increasing in the hyperscalers more generally, you talked about right now, they're kind of field-testing it to see versus lab testing.

What -- where -- how many quarters in field do you think the hyperscalers need to have your enterprise SSDs running before they feel comfortable ramping up volume there?

How many quarters does that process take?

I think it would take about at least about 2-quarter.

But so far I think, we have been what we said -- 6-server provider in each of the hyperscaler provider.

And we are almost there, just -- it all depends on business allocation for which unit, which BU use, what will have high power to use open-channel solutions.

Suji, let me also add, we expect to complete testing of our solution.

I mean, let me just go back a second, our controllers have already passed testing qualification.

The next step is the actual SSD solutions.

The SSD solutions using our new controllers are literally days away from passing the qualification before they go into initial low volume production, where these SSD solutions need to be tested in live applications, in live application, which will then have to go through further tuning of our driver software as a well as the host device software before they're able to go into a high-volume production of both the SSD module itself as well as our controllers used in those SSDs.

And timing of this, we believe should happen somewhere around the mid of next year.

Our next question comes from the line of Mike Crawford from Riley.

Can you elaborate more on the progress of your SM2270 controllers with a PCIe Gen 3 channels that also bring you up into the enterprise space in addition to what the work you're doing with these 2 hyperscalers?

Yes, our 2270 is our first enterprise controller to support open-channel as well as the NVMe standard.

We are in good progress to engage with a U.S. major supplier.

I think we are engage with the several.

And I think the first production will be around late Q1 or early Q2 next year.

And we also have the second generation of the 2270 coming in Q3 next year because we get tremendous feedback from U.S. major leaders and we had a certain security key value into the controller.

And that will expand further, and we gave got even better.

Because all the cloud makers in U.S. they are not really follow the standard and no -- there is no such call standard open-channel.

But they all have a similar concept like open-channel.

So that give us a tremendous opportunity to support, customization to meet their internal needs.

And we are exploring these concepts in our experience with the 2 hyperscalers in China and transferring into the U.S. market to selling our enterprise controller into China either in server side as well as in cost side.

Our next question comes from the line of Gokul Hariharan from JPMorgan.

So couple of questions from my side.

First of all, on the SSD solutions, I think you also ship a lot of non open-channel stuff also into your hyperscale customers, could you talk about what is happening from a procurement view, in addition, to the delay of qualification on some of your new SSD solutions?

Is it just that the spending overall on storage have been pushed out?

Or is it very specific to this model-related delay?

And I have one another.

So, for standard enterprise SSD solutions in this year, because we have to make a business decision, how many unit we want to sell, because our gross margin in that particular segment is not compelling.

The open-channel gave us much better opportunity to sell value add to the customer.

That's why we work with the 2 hyperscaler in the very beginning as open-channel SSD solution barely give us much better margin and better profit and moving forward.

However, to say that, we are also in the transition of our standard NVMe solution for already the enterprise customers and the e-commerce customers.

Because the NAND -- the key is how to secure a lower cost 3D NAND.

And now we are in the process to make commercial deal.

And we believe next year, we might be able to grow standard NVMe as well as open-channel.

But our primary focus is still open-channel because that gave us better margin and better future for our long term growth.

Okay, great.

As you look into next year, do you feel that most of your open-channel controller sales is going to be bundled with SSD solutions?

Or do you anticipate some open-channel controller sales on a stand-alone basis as well?

So in China, most of our open-channel controller were being the solution base.

Because there are a very few customers will use the controller base.

However, in U.S., there will be more controller base.

Okay.

So next year it's fair to assume most of it is going to be solution-based, given China seems to be still...

That's correct.

Okay.

One last follow-up.

I just wanted to confirm, you mentioned next year you're expecting the mix of UFS within the mobile solutions is going to be doubled up this year 20-plus-percent?

Or do you think that it's going to be much higher next year?

It will be at least double in our own pool, our eMMC+

UFS.

This year it will be 10%.

Next year, it will be minimum double.

Our next question comes from the line of Charlie Chan from Morgan Stanley.

So my first question, really, is about that enterprise SSD delay, do you see there is any demand-related issue, meaning associated to the recent CapEx costs of the customers?

As you know, we have heard about the overall CapEx spending slowdown by Chinese hyperscalers.

Our Shannon business is much smaller scale.

So we see no impact by this.

As you know the sale of the Shannon SSD can be lumpy as the timing of SSD procurement can fluctuate from time to time.

And we have successfully passed testing qualification for our new lease generation with the customer.

And we have confidence next year we're going to ramp by mid of next year.

Okay.

And next question is regarding the SSD penetration.

Because according to our industry survey, it seems like the emerging markets PC users, as they all want a big storage -- for example, 1 terabytes hard disk -- because they don't really have the access to cloud drive, et cetera.

So in your assumption or in your interaction with your customers, do you think that you need like 1 terabyte SSD or 512 gigabytes SSD for the penetration in the low end segment to happen?

Well, Charlie, with SSD pricing continue to fall sharply, the price of 1 terabyte drive in a year's time or 2 years is going to be significantly cheaper than what they are today.

Prices will continue to come down.

Price today for SSDs have reached a level, have only just reached a level that we saw in 2016, before NAND prices starts to increase.

With OEM pricing for SSDs becoming increasingly competitive relative to HDDs, and also, the higher density SSD coming down.

We're going to see meaningful step up in adoption and interest.

SSDs continue to be cheaper and cheaper, and they'll become -- will become cheaper and cheaper with the low-cost QLC, and the 96-layer version of the QLC.

So we are seeing more and more OEMs designing SSD into PCs, game consoles and client devices where they are for developed markets or emerging markets.

Now Charlie, I think, the 1 terabyte for PC OEM it won't be mainstream or value line next year.

It probably have to wait for 2020 or even further.

As in next year, 256 gigabyte will become mainstream or come to value line.

But 1 terabyte still so because still cost sensitive but most of the PC OEMs are happy even with 256 gigabyte for the main value line.

Okay.

Yes, so if 256 gigabyte is going to be acceptable by those low end segments.

And I think after 40%, 50% SSD price drop year-to-date.

Can I assume that next year the penetration of SSD in notebook to be like 80%, 90%?

I think commercial line is already about 70%, 80%, but consumer line is still now -- it's still about just the 20% range today.

Okay.

So yes, I'm just asking, if 256 gigabyte density is not good enough, what would be the penetration for consumer segment next year?

Because of PCIe.

As the consumer line mainly really is the Intel promotion, the transition of SATA to PCIe.

For HDD you don't have a PCIe, only the SSD have a PCIe.

So HDD is going to stay with SATA.

That's why the transition when moving to PCIe and cost become more competitive, they're going to have more PC OEM transition or even consumer line from HDD to PCIe.

In the same time, and even in some desktop, we see the transition from HDD to SSD due to the PCIe demand.

Okay.

And lastly, on that eMMC UFS product line.

So I hear you that there are several dynamics, right, for example, your customers market share and demand, et cetera, right.

But can we get kind of an anchor now?

What was the insource ratio of Hynix eMMC controller IC today?

So from that point we can gauge the potential decline for data eMMC business?

So what is the insourcing ratio today?

Well, Hynix continues to use our eMMC controllers.

We still have a lot of sales to Hynix.

And as we grow our sales to Micron and to the Chinese module markets, we -- a part of our strategy is to have a more diversified customer base.

So, Charlie, let me give you another angle so you can do it analytically yourself.

As you can see this, the NAND transition from 2D NAND to 3D NAND, because 3D NAND become more affordable, more costs effected, and that thing become bigger.

For eMMC in the past, they really needed a 2D NAND, because they have so many different density, especially for the low end.

When you go to a higher density, you have to transition to 3D NAND.

That's why that motivation move the whole market transitioning from eMMC, uMCB to UFS due to the 3D NAND transition.

Because that become more cost-effective than legacy 2D NAND.

So how soon, how fast in moving depend on the price to transition.

That's why we see the fraction model this year, meaning 120 gigabyte or above, next year, you're going to see even 256 gigabyte.

And we then -- we already received 512 gigabyte UFS today, right?

So that momentum in transition due to the costs driven.

And you see more momentum and we're going to see more eMMC transition to UFS next year and the year beyond.

So I think that's also because of the industrial changing from 2D NAND to 3D NAND, and driving more attractiveness to the mobile phone adopt the UFS, because they use the 3D NAND.

Okay.

Understood.

And lastly, last earning call you mentioned that some opportunity at China memory consumer side, like YMTC.

But I think yesterday, there was a section from U.S. established shipment to some memory makers in China.

So do you think that your opportunity as a China customers will get delayed, given that U.S., China trade tension?

Charlie, we cannot comment the China customer.

And by the way, the China memory customer, we didn't say is the YMTC either.

So you might mix it together.

But there will be -- could be some customer, who have contract with the NAND maker to assure, that's why, we do see our eMMC UFS also growing from that particular China customer.

Charlie, let me also add.

The sanctions was against a DRAM company not a NAND company.

And we can't -- we don't know whether these sort of sanctions will spread more widely or whether it's more company specific, as what we saw yesterday.

Our next question comes from the line of Donnie Teng from Nomura Securities.

My first question is regarding to the fourth quarter sales growth outlook for your different business units.

I wonder if you could give me a more detailed guidance on that.

Donnie, for our fourth quarter, we expect our SSD controllers to be flat sequentially.

We expect our SSD solutions to continue to come down.

We expect our UFS, eMMC to be flat to down a bit.

And those are essentially our primary products.

Okay, got it.

And my second question is regarding to the enterprise SSD and open-channel SSD opportunity.

I know it's a whole new product and a whole new technology and also, Wallace mentioned about our controller has passed the qualification.

But overall system probably still takes time to pass through the whole qualification process, but would you elaborate more about what kind of process we will need in order to get all customers qualification?

And if we are looking into next year, what kind of sales contribution we will have open-channel SSD opportunities?

And if those assumptions are based on that, our open-channel SSD controller are all bundled with Shannon system products?

Yes, in China -- first of all, in China, majority is leased solution sale.

There is almost no customer would take enterprise controller only, right.

It's quite different than the U.S. Now we have been working with our customer for more than 1 year, but our solution with the software ready just about 4 months ago.

So we have been through very intensive qualification with customers and with the server provider, right?

Each of the hyperscale customer have 5 different 6 server provider We have need to go through at least 300 to 500 server and go through their very intensive test, right?

We are in the final stage controller and firmware sale has been verified where our hyperscale customers need to fine tune the performance regarding how to meet the latency?

How to meet all the heavy traffic?

How to meet all the different application, different workload, different media?

So this is full layer decision.

For our hyperscale customers, we have a winged design.

Now their decision is how they can transfer different business units to provide us bigger chunk of our business for 2019.

Now we are preparing those for more intensive fine-tuning iteration with our customer.

So I think we are in the final stage.

And we have been verbally committed by our customer, they are very exciting to see the open-channel solution in the final stage, they really want to move into mass production.

So we are in the trial, in the live data center test and make sure that we'll go through all the commercial in the different business unit.

For financial service or online shopping, the work is all different.

So this is how we're working with our customers.

How about the -- do you have any preliminary view on the sales contribution start from mid next year?

And if I can -- I am just wondering if we can assume that.

Currently the open-channel SSD controller should will be bundled with Shannon products next year?

Donnie, we're expecting our overall SSD solutions to grow next year.

And this growth -- a big part of the driver will be the ramp and scaling of our Shannon SSDs beginning middle of next year.

I'll now hand the call back to today's presenter, please continue.

I would like to thank all of you for joining us today, and your continue interest in Silicon Motion.

We will be attending several investor conferences in Asia, in U.S. in the fourth quarter.

Details of these events will be available on our website.

Thank you, and goodbye for now.

Thank you.

Ladies and gentlemen, that does conclude our conference for today.

Thank you for participating.

You may all disconnect.