Micron Technology (MU) Q2 2019 Earnings Conference Call Transcript

Micron Technology (MU -0.94%)
Q2 2019 Earnings Conference Call
March 20, 2019 4:30 p.m. ET

Contents:

• Prepared Remarks
• Questions and Answers
• Call Participants

See all our earnings call transcripts.

Prepared Remarks:

Operator

Good afternoon. My name is Latif, and I will be your conference facilitator today. At this time, I would like to welcome everyone to Micron's second-quarter 2019 financial release conference call. [Operator instructions] It is now my pleasure to turn the floor over to your host, Farhan Ahmad, head of investor relations.

You may begin your conference.

Farhan Ahmad -- Head of Investor Relations

Thank you, and welcome to Micron Technology's second fiscal-quarter 2019 financial conference call. On the call with me today are Sanjay Mehrotra, president and CEO; and Dave Zinsner, chief financial officer. Today's call will be approximately 60 minutes in length. This call, including the audio and slides, is also being webcast from our Investor Relations website at investors.micron.com.

In addition, our website contains the earnings press release filed a short while ago. Today's discussion of financial results will be presented on a non-GAAP financial basis, unless otherwise specified. A reconciliation of GAAP to non-GAAP financial measures may be found on our website, along with convertible debt and capped call dilution table. As a reminder, the prepared remarks from this call and webcast replay will be available on our website later today.

We encourage you to monitor our website at Micron.com throughout the quarter for the most current information on the company, including information on various financial conferences that we will be attending. You can follow us on Twitter, @MicronTech. As a reminder, the matters we will be discussing today include forward-looking statements. These forward-looking statements are subject to risks and uncertainties that may cause actual results to differ materially from statements made today.

We refer you to the documents we filed with the SEC, specifically, our most recent 10-K and Form 10-Q, for a discussion of risks that may affect our future results. Although we believe that the expectations reflected in the forward-looking statements are reasonable, we cannot guarantee future results, levels of activity, performance or achievements. We are under no duty to update any of the forward-looking statements after today's date to conform these statements to actual results. I'll now turn the call over to Sanjay.

Sanjay Mehrotra -- President and Chief Executive Officer

Thank you, Farhan. Good afternoon, everyone. Micron executed well in the second quarter, delivering solid results and healthy levels of profitability and free cash flow despite a challenging industry environment. We continued to strengthen our balance sheet in the quarter by increasing our cash position and total liquidity.

Although we expect industry headwinds in the near term, we continue to grow and diversify our product portfolio, improve our cost competitiveness and lay the foundation to emerge stronger, both financially and operationally from this environment. I would like to start with the review of two important pillars of Micron's strategy, improving cost competitiveness and increasing high-value solutions in our portfolio. Our strategy positions us for the tremendous opportunities ahead, while also enabling us to better navigate near-term headwinds. Strong execution against this strategy has improved our annualized profitability by over $6 billion from fiscal 2016 to fiscal 2018.

This has improved our EBITDA margin by more than 15 percentage points relative to our competitors over the same period. We expect further progress on cost reduction this fiscal year, including healthy year-over-year cost declines in both DRAM and NAND. In DRAM, our 1Y nanometer is yielding well, and we expect to increase conversion to 1Y nanometer in the second half of fiscal 2019. We are also making excellent progress on 1Z nanometer and have started sampling products utilizing this technology.

As we have said in the past, future DRAM node transitions required additional process steps and more fab cleanroom space. Consequently, in addition to the previously announced expansion of our Hiroshima facility, we are starting site preparation for the cleanroom expansion at our Taichung facility to enable the transition of existing DRAM wafer capacity to future nodes. We are still finalizing the timing but expect production output some time in calendar 2021. In NAND, we achieved meaningful production on 96-layer 3D NAND in fiscal Q2 with the fastest yield ramp of any NAND product in our history.

We are also making good progress on development of our fourth generation 3D NAND, which uses our replacement gate technology. Given the high initial capital requirements of floating gate to replacement gate conversion, we expect that our first replacement gate node will provide limited cost reduction, and hence, we are planning to deploy this node across select NAND products, with the rest of the portfolio converting later to the second node of replacement gate. This approach will optimize the ROI of our NAND capital investments as we convert our capacity. As a reminder, our replacement gate architecture will allow us to deliver performance improvements and provide us an efficient path toward scaling multiple future generations of 3D NAND.

Given the limited initial deployment at the first node of replacement gate, we expect that our NAND bit supply growth in calendar 2020 will be below industry demand levels, and we plan to utilize our cost-effective floating gate inventory positioned to meet customer requirements. Turning to high-value solutions. More than two-thirds of NAND revenues in the first half of fiscal 2019 were from high-value solutions, up from 55% in the first half of 2018. This increased mix of high-value solutions, combined with our competitive cost structure, enabled us to deliver fiscal Q2 NAND gross margin in the high 30s despite steep price declines in the industry.

In SSDs, we are making progress on transitioning to NVMe while continuing to improve our cost profile in SATA. In fiscal Q2, we began revenue shipments to a large PC OEM for our first NVMe client SSD, which features our internally designed controller and our inactive qualifications with other customers. We intend to introduce cloud and enterprise NVMe SSDs later this calendar year. In SATA, we introduced consumer and client SSDs based on 96-layer 3D NAND in fiscal Q2.

In the cloud market, our custom persistent memory solution, which combines DRAM and NAND, is now fully ramped and contributed meaningfully to our cloud revenues. 3D XPoint development remains on track with customers' samples planned before calendar year-end. We believe 3D XPoint technology will be a key enabler for numerous new applications, particularly artificial intelligence and data analytics. As announced previously, in January of this year, we exercised our option to acquire Intel's interest in the IMFT facility in Lehi, Utah.

This acquisition provides us with the manufacturing capability and highly skilled talent to drive 3D XPoint to development and innovation. Now turning to end markets, I'll start with mobile. During fiscal Q2, we grew revenues and expanded gross margins year-over-year despite adverse memory and storage pricing and weakness in high-end smartphone unit sales. Our performance in mobile was propelled by growth in our managed NAND portfolio, where NAND bit shipments grew more than five times year over year.

We are also seeing strong demand for our 1Y nanometer LPDRAM due to its industry-leading capacity and best-in-class power consumption. Memory and storage content growth in smartphones continues, driven by features such as multiple cameras, machine learning, computational photography and 4K video. Last month, Samsung announced its premium Galaxy S10 Plus smartphone, featuring 12 gigabytes of DRAM and 1 terabyte of NAND. At Mobile World Congress, several companies announced exciting new forms featuring 5G connectivity and foldable screens.

These next-generation premium smartphones will typically feature 8 to 12 gigabytes of DRAM and 256 to 512 gigabytes of NAND versus four to six gigabytes of DRAM and 64 to 128 gigabytes of NAND in current-generation premium smartphones. These trends will likely cascade to lower-tier phones as well. We believe that 5G foldable phones and upcoming innovations in augmented and virtual reality will drive sustained content growth for years to come and should reignite smartphone unit sales beginning in calendar 2020. We are also excited by the opportunity that 5G is likely to create beyond mobile as it will enable true machine-to-machine communication and accelerate data creation and analysis, which are fundamental drivers for our business.

We expect 5G adoption to create increased demand for memory and storage in IoT devices, wireless infrastructure and data centers. Our embedded and networking businesses are already starting to see benefit from early 5G infrastructure investments. In the data center markets, the demand for memory has moderated this year, following exceptional growth in the last two years. The slowdown in demand is a result of ongoing customer inventory adjustments, as well as software optimizations at some cloud customers.

We expect growth to resume in the second half of calendar 2019 as we see improvement in our customers' inventory position. The new server processors that support higher memory densities are expected to be introduced in a few months, which should drive additional demand growth in the second half of calendar 2019. In fiscal Q2, we shipped high-density 1Y nanometer DDR4 server module samples to customers ahead of plan, which will position us well to benefit from this new CPU platform ramp. In graphics, we grew sales of our high-performance GDDR6 DRAM and expanded our customer base, which positions us for stronger growth in the second half of calendar 2019.

We are seeing steep customer inventory adjustments in GDDR5 and expect them to be largely completed by the middle of this calendar year. We had delivered strong quarter in automotive, with year-over-year revenue growth driven by increasing demand for ADAS and advanced in-vehicle infotainment systems. In fiscal Q2, we announced several new automotive products, including a collaboration with Qualcomm for next-generation in-vehicle infotainment and 5G communications modules. We also announced a new strategic collaboration with a leading supplier of ADAS platforms using our full portfolio of memory and storage products.

Like last year, automobile unit sales are a short-term challenge, however, we see the auto market generating robust growth for Micron over the next decade as memory and storage content continues to increase in autos, driven by advanced infotainment systems and the adoption of autonomous vehicles. In the industrial and consumer markets, we saw a decline in sales due to seasonal, macroeconomic and pricing weaknesses, as well as inventory adjustments. We had important design wins in video surveillance, point of sales and factory automation applications. At Mobile World Congress, we announced the industry's first one terabyte microSD card using QLC NAND.

In the PC market, sales declined more than 25% sequentially, driven by weaker pricing and the inventory drawdowns seen in other segments, as well as client CPU shortages. We remain focused on our cost competitiveness in this market, and over two-thirds of our PC DRAM bit shipments are now coming from our advanced 1X and 1Y nanometer technology nodes. Now turning to our DRAM industry outlook, since our last earnings call, DRAM pricing weakened more than expected. Our demand outlook for calendar 2019 has moderated, led by somewhat greater levels of customer inventory, weakening server demand at several enterprise OEM customers and worse-than-expected CPU shortages.

We believe macroeconomic uncertainty is also contributing to hesitation in buying behavior at some customers. However, as we discussed on our last earnings call, we still expect DRAM bit shipments to begin infusing in our fiscal Q3, with demand growth strengthening in the second half of calendar 2019, as most customer inventories are likely to normalize by midyear. Based on our current view, we now estimate calendar 2019 DRAM bit demand growth from our customers to be in the low- to mid-teens, with their end demand a few points above that. Further, we estimate the industry supply bit growth is tracking to mid to high teens.

Given the lower DRAM demand outlook from our customers, we have decided to idle approximately 5% of our DRAM wafer starts. This action will bring our production levels close to our view of DRAM industry bit demand growth for calendar 2019. We will continue to monitor the market and take appropriate actions to ensure that our bit supply growth in calendar 2019 remains closely aligned with demand. Looking beyond our fiscal 2019, we expect bit demand growth to accelerate as mobile and server demand improves.

In particular, we expect robust DRAM bit demand growth in fiscal 2020 bouncing back from a weak fiscal 2019. NAND markets remain oversupplied from the acceleration in bit growth, driven by the industry transition to 64-layered 3D NAND. Although fiscal Q2 pricing came in below our expectations, we are optimistic that demand elasticity and seasonal trends will support improving demand growth in the second half of the calendar year. We expect that calendar 2019 NAND bit demand growth is likely to be in the mid-30s percent range, with industry supply growing in the high 30s, and we are targeting our bit shipments to grow close to the growth rate of industry bit demand.

We have been managing our NAND bit supply growth prudently, including adjusting our capital planning and wafer volumes. We are reducing our total NAND wafer starts by approximately 5%, mostly through reductions on our legacy nodes. Given these changes in DRAM and NAND industry conditions, we have reduced our capex for fiscal year '19 and are evaluating our capex for fiscal year '20. We are taking prudent actions to address the current market conditions while executing well on our long-term strategic objectives.

I will now turn it over to Dave to provide financial results of our fiscal second quarter and guidance for the third quarter.

Dave Zinsner -- Chief Financial Officer

Thanks, Sanjay. Micron's fiscal second-quarter results were within our guidance as we executed well in a period of somewhat weaker-than-expected market conditions. In the quarter, we returned capital to shareholders, continued to improve our cost structure and strengthened our balance sheet through our first issuance of investment-grade debt. Most of the proceeds from this offering were used to redeem a large portion of our outstanding convertible notes, reducing our fully diluted share count.

Total fiscal second-quarter revenue was $5.8 billion, down 21% from the prior year and down 26% from the fiscal first quarter. Revenue reflected worse-than-expected pricing trends in DRAM and NAND. DRAM revenue was down 28% year over year and 30% sequentially from the fiscal first quarter and represented 64% of total company revenue in the fiscal second quarter. DRAM ASPs declined in the low-20% range compared to the prior quarter while shipment quantities were down in the low double digits.

Year over year, DRAM bit shipments were down mid-single digits. NAND revenue declined 2% year over year and 18% from the prior quarter. NAND revenue represented 30% of total company revenue in the fiscal second quarter. Our overall NAND ASP declined in the mid-20% range, while shipment quantities increased in the upper single-digit percentage range compared to the prior quarter.

NAND bit shipments came in stronger than our expectation due to timing of demand from a large customer. Now turning to our revenue trends by business unit, revenue for the compute and networking business unit was $2.4 billion, down 35% year over year and 34% from the prior quarter. The sequential decline was driven by pricing across major market segments, as well as volume reductions, particularly in graphics, cloud and PCs. Revenue for the mobile business unit, or MBU, was $1.6 billion, up 3% year over year and down 27% from the record fiscal first quarter.

Lower volume due to weak seasonality and pricing contributed to the quarter-over-quarter revenue decline. Despite the current market conditions, MBU revenue grew year over year due to strong growth in our managed NAND products. The embedded business unit revenue of $800 million was down 4% compared to the prior year and down 14% from the record fiscal first quarter. The automotive business was only down slightly from record fiscal first quarter despite weakness in automobile sales.

Other embedded revenue declined quarter over quarter due to weaker pricing and lower DRAM volumes. Finally, the storage business unit fiscal second quarter revenue was $1 billion, down 19% year over year and down 11% quarter over quarter. The sequential decline was driven by lower SSD revenue, partially offset by increased component revenue. The consolidated gross margin for the fiscal second quarter was 50%, compared to 58% in the prior year and 59% in the fiscal first quarter.

Fiscal second-quarter results included ongoing impact from 3D XPoint underutilization cost, which approximated 160 basis points. Pricing came in weaker than expected in both DRAM and NAND, but strong execution on cost reductions resulted in gross margins within our guidance range. Fiscal second-quarter operating expenses of $818 million came in at the high end of our guided range. Opex included $37 million in onetime tool impairment costs that were not anticipated in our guidance.

Excluding this impairment, opex would have come in at the low end of our guidance. We remain focused on controlling our expenses while investing in future products and technologies. We delivered solid profitability in the fiscal second quarter with operating income of $2.1 billion, representing 36% of revenue. This margin is down 13 percentage points year over year and also down 13 percentage points from the fiscal first quarter.

The tax rate for the fiscal second quarter was approximately 8.3%, and we now expect our fiscal 2019 tax rate to be approximately 10.5%. Non-GAAP earnings per share in the fiscal second quarter was $1.71, down from $2.82 in the year-ago quarter and down from $2.97 in the prior quarter. Turning to cash flows and capital spending, in the fiscal second quarter, we generated $3.4 billion in cash from operations, representing 59% of revenue. Capital spending, net of third-party contributions, was approximately $2.4 billion, relatively flat compared with the prior quarter.

We have lowered our capex targets for fiscal 2019 to approximately $9 billion from our prior guidance range of $9 billion to $9.5 billion as we manage through the current environment while maintaining investment in our strategic priorities. As we mentioned before, a significant portion of our capex this year is going toward cleanroom construction and assembly and test operations, which do not contribute to our bit supply growth. Fab equipment spend in fiscal 2019 is down from last year and is mostly targeted toward migrating 20-nanometer DRAM and 32-layer 3D NAND to more advanced nodes, with no new wafer capacity additions. The investment in these technology transitions provides compelling cost reduction and a very attractive ROI.

As we have demonstrated over the past two quarters, we remain nimble on capex based on business conditions. In the fiscal second quarter, our adjusted free cash flow, defined as cash flow from operations less net capex, was approximately $1 billion, compared to $2.2 billion in the year-ago quarter and $2.3 billion in the fiscal first quarter. We deployed approximately 70% of the quarter's free cash flow toward our share repurchase program. We bought back approximately $700 million of stock in the fiscal second quarter, representing 21 million shares.

Through the first half of fiscal 2019, we repurchased 63 million shares for $2.5 billion, utilizing 76% of our free cash flow in fiscal first half. We continue to do share repurchases as an attractive use of capital and remain committed to deploying at least 50% of our free cash flow on an annual basis toward repurchases under our current $10 billion authorization. In addition to the buybacks, we also eliminated our 2043 converts subsequent to the quarter close, which at current share prices, effectively lowered our fully diluted share count by approximately 9 million shares. Note that the underlying share count for these converts was 35 million shares, which could have resulted in greater dilution with increases in our share price.

Inventory ended the quarter at $4.4 billion, increasing from $3.9 billion at the end of the fiscal first quarter. Our fiscal second quarter days of inventory were 134 days, compared to 107 days in the fiscal first quarter. The actions that we have announced today regarding supply reductions, combined with improving customer demand, will begin to address our higher inventory levels. We ended the fiscal second quarter in a strong liquidity position with net cash of $3 billion and total liquidity of nearly $12 billion.

Total cash increased to $9.2 billion, largely as a result of our $1.8 billion debt issuance. Our debt position increased by $2.1 billion to $6.2 billion in the quarter, primarily due do the debt offering and recognition of premium associated with the Series G convertible note redemption. Subsequent to quarter end, we used $1.4 billion of cash to complete the convertible note redemption, which reduced our outstanding GAAP debt balance by approximately $1.1 billion. On January 14, we exercised our call option to acquire IMFT, our joint venture with Intel.

We expect to close the transaction in either late fiscal 2019 or in the first half of fiscal 2020. This near $1.5 billion transaction will also retire the $1 billion of joint venture-related debt on Micron's balance sheet. Since we already consolidate IMFT's results in our financial statements, we do not expect a material impact to our near-term results.Now turning to our market outlook. DRAM and NAND markets are working through supply and demand imbalances.

Our visibility remains low, and the near-term environment remains challenging. While there have been capex reductions across the industry, they haven't yet impacted output growth due to lead times. We expect our DRAM bit shipments to grow sequentially during the fiscal third quarter and at much higher rates in the fiscal fourth quarter. In NAND, we expect a modest sequential decline in our bit shipments in the fiscal third quarter due to timing of shipments and expect growth to resume in the fiscal fourth quarter.

The output reductions that we announced today for DRAM and NAND, plus the $1.5 billion of capex reductions that we have announced year to date, will also help reduce our supply of DRAM and NAND in the second half of this calendar year. With that in mind, our non-GAAP guidance for the fiscal third quarter is as follows: We expect revenue to be in the range of $4.8 billion, plus or minus $200 million; gross margin to be in the range of 37% to 40%; and operating expenses at approximately $785 million, plus or minus $25 million. Based on a share count of approximately 1.14 billion fully diluted shares, we expect EPS to be $0.85, plus or minus $0.10. In closing, Micron continues to execute on our key initiatives, increasing our mix of high-value solutions, improving our cost profile and investing in new and innovative products.

Our solid financial footing, strong liquidity and substantially reduced leverage means that we are well-positioned to invest in our future despite near-term market conditions. I'll now turn the call over to Sanjay for some concluding remarks.

Sanjay Mehrotra -- President and Chief Executive Officer

Thank you, Dave. In response to near-term industry conditions, we are taking decisive action to reduce our supply growth to be consistent with industry demand. At the same time, we continue to invest and execute against our strategic priorities to reduce costs and increase the mix of high-value solutions in our portfolio. The long-term demand trends for Micron remain very healthy, with tremendous growth opportunities across multiple markets.

We continue to believe that the memory industry is structurally stronger with more diversified demand drivers and moderating supply growth capability. With our strong balance sheet and improved product portfolio and operating model, Micron is better positioned than ever before to win and deliver long-term value for shareholders. We will now open for questions. 

Questions and Answers:

Operator

[Operator instructions] Our first question comes from the line of Ambrish Srivastava of BMO.

Ambrish Srivastava -- BMO Capital Markets -- Analsyt

I just wanted to focus on the inventory situation. Sanjay, you talked about hyper-scale inventories a couple of times and that they are tying up also, trying to see what the balance sheet inventory would do. So a, just update on where are we in terms of weeks of inventory. And then my sort of quick follow-up, Dave, what's the flex capex now left that you could moderate down if bit conditions were to get worse?

Dave Zinsner -- Chief Financial Officer

OK. So first, let me talk about our inventories. So as we noted, our inventories are now at 134 days. A couple of dynamics going on.

So in DRAM, as we talked about last quarter, we expected to carry a higher level of inventory in the first half of the calendar year as our customers that have built up inventory start to digest through that. And then on top of that, what we did was we reduced our capex. And so the outcome of that reduction in capex should start to have an effect in the back half of our calendar year in terms of bit growth, so that will start to help manage inventory as well. And then on top of that, this quarter, we talked about further reductions in capex down to the approximately $9 billion, and we talked about taking some underutilization, about 5% of underutilization to kind of further reduce the output.

So that should help bring DRAM inventories back more in line. So we would expect that back half of the calendar year, we'd start to see DRAM be in a better place. On the NAND front, obviously, we're at elevated levels of inventory as well. Now that is somewhat by design.

As Sanjay noted in his remarks, we do have the transition from floating gate to replacement gate in NAND going on in calendar 2020, and so the bid output is not going to be very robust. And so the idea was to carry higher levels of inventory into 2020 such that that would help support the demand from our customers. Having said that, part of the capex reduction does impact NAND as well. And also, we did adjust our wafer output by 5% and NAND as well to also try to manage inventory.

So I think the inventory, we'll start to see improve in the back half of the calendar year and then into 2020 as we start to work through the NAND portion of the inventory. As far as capex goes, really, the way we're managing this is what we said when we talked about it at the Analyst Day. We're committed to our model of low 30s as a percent of sales for capex. Obviously, there will be years where that is above the average, and there'll be years that that will be below the average.

I think the last two years, fiscal 2017 and 2018, we were below that low-30s model. So it's not surprising to be a bit above that low-30s level for these couple of years. Plus on top of that, we had a bit of pent-up demand to get some of the cleanroom space staged to be able to manage these node transitions, and so we have a number of different cleanroom build-out initiatives going on this year, which is maybe elevating a little bit from kind of our normal run rate. But over the long term, you would see us be in this low-30s percent level.

And as I think Sanjay mentioned in his remarks, I think I mentioned in my remarks, we're flexible on capex. We modulate it based on the economic circumstances.

Ambrish Srivastava -- BMO Capital Markets -- Analsyt

And the hyper-scale?

Sanjay Mehrotra -- President and Chief Executive Officer

With respect to hyper-scale, as I said in my remarks that we do expect that inventory will work through the system at our customers during the first half of the year. And we certainly are seeing evidence of inventory consumption by our customers. And we do expect that that will largely be completed by mid this year. And we do expect in DRAM that growth will accelerate both sequentially and year over year for the rest of the calendar year.

And second half of the calendar year, with respect to demand improvement, demand will improve in the second half of the calendar year compared to the first half of the calendar year. And of course, effect of CPU shortages, those will continue to improve in situation over the course of the year, and we expect that also to contribute toward improvement in demand in the second half of the calendar year compared to first year.

Operator

Our next question comes from the line of John Pitzer, Credit Suisse.

John Pitzer -- Credit Suisse -- Analyst

Sanjay, I guess, I just want a little bit more detail behind the decision to actually shutter some DRAM capacity in the current quarter. Last quarter, you talked about how the flattening cost curve actually increased the useful life of holding inventory, which seemed to make sense to me. There seems to be a little bit of a reversion from that. And just given that you're not a big enough piece of the DRAM market to, I think, impact overall supply demand, it really kind of depends upon what the guys in Korea are doing.

I'm just kind of curious as to why you came to this decision to shutter some capacity.

Sanjay Mehrotra -- President and Chief Executive Officer

So I think as Dave had previously mentioned in one of the earnings calls, that about 150 days of inventory is something that we are comfortable with. And when we really look at our demand at this time and look at our production output, we think it is important for us to bring our supply to be in line with the expected demand for the year. And it provides us obvious benefits in terms of managing our inventories more effectively, managing our cash flow effectively as well and really managing our business in a healthy fashion. Of course, our inventory is at good cost.

There is no obsolescence issue with this inventory at all, but we really considered it prudent management of our business to bring our supply growth in line with our demand expectations.

Dave Zinsner -- Chief Financial Officer

And I'd just add that the underutilization costs that we'll incur are not going to kind of layer into our cost per units. They will be taken as period costs within the quarter, and it's implied in our 37% to 40% gross margin guidance for the third fiscal quarter. Sorry to interrupt you, John.

John Pitzer -- Credit Suisse -- Analyst

That's OK. So Sanjay, you explained why in fiscal 2020, you will under-grow the industry in NAND bit as you move to fourth generation. I'm wondering if you can help us understand what the cost implication of that is. Will your cost curve in NAND be negatively impacted? Or should you be able to benefit in 2020 from continued mix toward 96-layer on the technology side and just product mix to more manage NAND? How do we think about costs in fiscal-year 2020?

Sanjay Mehrotra -- President and Chief Executive Officer

Our cost position will be good in fiscal-year 2020 because today with our both 64-layer, as well as 96-layer technologies, we absolutely have been outperforming the industry in terms of our -- at the die level, the cost on a per gigabyte basis. And obviously, as you know, we have discussed this many times in the past, those benefits come from CMOS Under the Array architecture, which Micron has pioneered, and we currently are in our third-generation using that. So that gives us a very significant cost advantage, and that will carry over well, giving us competitive mix of NAND production in 2020 in our fiscal-year 2020 to meet the market demand. And keep in mind, the fourth generation of our 3D NAND, which will be the first node in replacement gate technology, even with greater number of layers, will actually have greater capital intensity associated with it because of transition from floating gate to replacement gate that requires more tools, more specific unique tools to implement replacement gate.

And that's why we are really converting portion of our portfolio to that first-generation replacement gate technology. And remainder of the portfolio will convert from the third-generation floating gate to the second node of replacement gate later on. And obviously, that will provide us a bigger benefit when we do that. The second node of replacement gate will get higher layers later compared to the first node.

So that will give us a benefit of cost, as well as ROI and it will really be the best way to manage this transition for us. So in summary, even with increasing the number of layers in the current and first generation of replacement gate, we feel very good about the cost competitiveness with the mix of floating gate and replacement gate technologies that we'll be achieving next year.

Operator

Our next question comes from the line of CJ Muse of Evercore.

CJ Muse -- Evercore ISI -- Analsyt

I guess, a two-part question. You used the word idling regarding 5% for DRAM and then reducing your NAND wafer starts by 5%. I guess, two questions to that. Are you trying to send us a different signal in terms of what you're doing for both? And then could you provide a little clarity on how we should think about the impact to your cost structure as you slow down wafers? And then when you start to turn back on, how should we think about the impact?

Sanjay Mehrotra -- President and Chief Executive Officer

So bottom line is we are reducing our output of DRAM and NAND. In DRAM, the reduction in supply growth is a result of this idling capacity. Idling wafer starts is bigger impact. And in NAND, as we reduce our wafer starts, particularly related to the legacy nodes, the impact on the supply bit growth is smaller because that wafer start reduction that we mentioned is primarily targeted toward legacy nodes.

So again, the objective of these is to bring out or to bring down our supply growth toward our estimations of demand, both in DRAM and NAND from the point of view of 2019, as well as 2020 time frame. And with respect to cost, as I said before, we have continued to outperform in terms of year-over-year cost reduction capability in DRAM, and we feel very good about our ability to manage with this slowdown in DRAM advantage in a competitive fashion with, of course, over the course of the next few quarters. And in NAND, I just commented on our cost position there. And I'll just add to the previous comments on the cost position that, of course, next year, we will also continue to get the benefit of our 96-layer transition from 64-layer to 96-layer, the ongoing transition during the course of next year, which will also be helping us with cost.

So I believe our cost in DRAM, as well as NAND will be competitive. And I just want to point out that the benefits of an output reduction go beyond just aspects of cost. They, of course include -- those benefits include inventory management and cash flow management.

Operator

Our next question comes from the line of Harlan Sur of JP Morgan.

Harlan Sur -- J.P. Morgan -- Analyst

So back to a question on inventory. So when you first started seeing your customer demand inventory-related weakness, I know it takes time for the actual supply growth output of your manufacturing infrastructure to kind of trend toward your target. So initially, you augment that manufacturing ramp-down with some inventory accumulation. It appears that you executed to that in fiscal Q2.

But now that you're moderating output lower again, again, it appears that it's going to take some more time. So do you anticipate inventories actually moving higher here in the mid quarter before it starts to downshift in the back half of the calendar year?

Dave Zinsner -- Chief Financial Officer

I think inventory might be a little bit elevated in the May quarter relative to where we are today. Obviously, as you point out, it does take some time. I mean, most of the capex reductions, we took in the first fiscal quarter, so those take a couple of quarters, but we should start experiencing that in the back half of the calendar year. And the underutilization, you get probably a quarter before it starts impacting you.

So I think we'll be in a pretty good shape on DRAM, as I said as we kind of exit the calendar year. NAND will take a little bit longer because of the reasons we cited about keeping elevated levels of inventory into 2020.

Harlan Sur -- J.P. Morgan -- Analyst

Great. And then my follow-up question, Sanjay, much of your industry outlook is based on a view of second half improvements in virtually all of your target markets, PC client, cloud, mobile, etc. Given the value chain and lead time, should we anticipate that your customer second half recovery should start to drive an inflection in your NAND and DRAM business in fiscal Q4?

Sanjay Mehrotra -- President and Chief Executive Officer

We are not providing fiscal Q4 guidance here. But as we have said, that second half of this calendar year, we do expect that bit growth in DRAM to continue on a quarter-over-quarter basis, both sequentially, as well as on a year-over-year basis. And that is already -- we are guiding to bit growth in DRAM for our F Q3, and that should continue in F Q4 and for the rest of the year as well. And then NAND, there is, of course, seasonality in the second half of the calendar year.

And of course, we expect the effect of elasticity also to help improve the demand. And the trends of increasing average capacities in smartphones, both for DRAM and NAND, continues to be intact. When you look at the results that some of the cloud service providers have reported, those results continued to be solid, pointing to the demand and the growth in cloud services. And of course, memory and storage is an essential part of all the computing paradigms of the future that are increasingly being based on the IML as well.

All of that bodes well for the long-term demand trends as well. But yes, I mean, in all of our end markets, we do see generally improvement in demand in the second half versus the first half. And of course, all of this is barring macroeconomic uncertainty. In case there is any further deterioration in macroeconomic climate, then we can have different impact in this outlook.

But that's what based on what we know today, that's our assessment of demand expectations.

Operator

Our next question comes from Timothy Arcuri of UBS.

Timothy Arcuri -- UBS -- Analyst

Sanjay, I wanted to go back just to the question I kind of asked before. So in DRAM, you're still well above cash cost, yet you're idling capacity. So obviously, demand is a lot worse than you expected. But you sound pretty optimistic that it's going to come back in the back half of the year.

So why would you reduce starts now if you're so optimistic demand coming back in the back of the year? I get that it has to do with inventory, but I'm just trying to reconcile those two comments because it seems like you're losing money on the table.

Sanjay Mehrotra -- President and Chief Executive Officer

So demand has been weak so far this year, and we expect demand environment to be weak in the first half as well. So when we talk about our total DRAM demand growth expectation, we are talking about the full year basis, right? So we need to be -- and when we look at our supply growth, of course, we are looking at that, too, on a full-year basis. So we want to bring our supply in line with our demand expectations. And we believe that this improving demand environment, second half versus first half that I referred to earlier, we will be able to capture the benefit of that with our supply capabilities that we have mentioned and also help, over time, continue to drive improvement in our inventory position as well.

Timothy Arcuri -- UBS -- Analyst

OK. And then just a quick follow-up, Dave. I just wanted to go back to the question on cost. You've been bringing DRAM cost down kind of like low to mid-teens year over year and NAND has been kind of in the mid-20s year over year.

So it seems like some of these actions have to ultimately impact your ability to cost down vis-a-vis that level. Am I missing something there? Because it sounds like you're saying that there's no impact, so can you just again reconcile those for me?

Dave Zinsner -- Chief Financial Officer

Yes. We do expect healthy cost declines in both DRAM and NAND this year. We are still making technology transitions next year. And so we would obviously expect, particularly in DRAM, that cost will be good.

As we said, replacement gate might not be where we want it to be, in the very first generation. But one of the advantages of carrying the inventory we're carrying, 64-layer and 96-layer inventory in the next year is that it has a very good cost structure, and so we should have a good cost structure in 2020 on NAND as well. The other thing, on utilization, I just want to make sure it's clear what period cost means is we will take that expense kind of as it comes in the quarters in which we take the underutilization. It won't roll up into inventory.

It won't be counted as part of the cost. So obviously, it will have an impact on margins. And as I said, it's implied in the guidance of the margins we gave for the fiscal third quarter.

Sanjay Mehrotra -- President and Chief Executive Officer

And on the end, I just want to point out again that 96-layer is still early in our production ramp, and it will continue to ramp during the course of next year as we bring out our first node of replacement gate. So 96-layer will also continue to be a strong driver of our cost reduction capability in 2020.

Operator

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

Karl Ackerman -- Cowen and Company -- Analyst

I wanted just kind of go back to bit growth. I understand that idling your wafer starts slows your bit supply, but do you expect any shifts in your portfolio mix toward certain end markets or shift toward new technology nodes that should impact your second half bit supply?

Sanjay Mehrotra -- President and Chief Executive Officer

So I guess, we really pay a lot of attention to this in terms of managing the mix of our production. We do that on an ongoing basis anyhow in production. And when we implement this output cut, both in NAND, as well as DRAM, obviously, we make sure that we are not impacting any aspect of our ability to fulfill our demand. So as we said, I mean, we have high levels of inventory currently, both in NAND and DRAM, and it gives us an opportunity to bring our production to be in line with demand.

And when we say that, we mean that we will be able to address the demand in all of our end markets. We are not going to compromise our production cuts so that we can go in terms of meeting our customers' requirements in any of the end market segments. So we believe this is a very prudent way for us to manage our business.

Karl Ackerman -- Cowen and Company -- Analyst

I appreciate that. I guess perhaps a clarification on a question then. The Street would say that NAND demand elasticity has not yet kicked in. So could you just perhaps shed some light on design wins, maybe not just in NAND but also in DRAM that you've seen across your product portfolio that would anchor your view of a second-half recovery?

Sanjay Mehrotra -- President and Chief Executive Officer

With respect to NAND elasticity, I mean, certainly, the average capacities in notebook, PCs, as well as the attach rates of SSDs to notebooks does continue to go up, so that is, of course, a price-elastic market as well. And the attractive price points of NAND, of course, provide benefits of greater HDD replacement, both in the data center market, as well as in the client market. Mobile definitely continues to be pointing to elasticity with respect to NAND. At Mobile World Congress, you saw smartphones, foldable phones and 5G phones being announced with 512 gigabyte and terabyte of storage capacities in them.

And again, applications are also advancing toward features that require more and more storage. For example, 5G will have greater connectivity, as well as low latency, will of course, be driving applications that will need more DRAM, as well as more NAND. So end-market applications continue to do very well, and Micron itself, with respect to our product portfolio, we are very much focused on continuing to expand our product offerings. And we have also on the SSD side, we are expanding our portfolio with NVMe solutions.

We have just brought our first NVMe client drive into the market, and later this year, we will have NVMe for enterprise and cloud applications as well. So this is -- and we announced that we got a qualification win and just began shipment to a customer with our NVMe -- first-generation NVMe client driver as well. So these are all the things that we are absolutely focused on in terms of a road map toward expanding our opportunities. And of course, we have made a lot of improvement with respect to some of these solutions over the course of last couple of years.

And there's lots of execution that is remaining ahead of us with respect to expanding our SSD opportunities, for example, as well as growing our mobile opportunities on the NAND side, where we have gained meaningful share over the course of last few quarters, but there is a lot more opportunity and a lot more road map execution that is required for us here.

Operator

Our next question comes from the line of Blayne Curtis of Barclays.

Blayne Curtis -- Barclays -- Analyst

I was curious on the customer timing, if you can elaborate. Was that just timing between quarters or was that more of a strategic buy that was multi-quarters in nature? And then I was just curious if you can elaborate on the data center you said back in the second half. It seems like you're not seeing a pickup in orders yet. And kind of what gives you confidence that you're going to see that starting Q3?

Sanjay Mehrotra -- President and Chief Executive Officer

So with respect to the timing of demand in the second quarter, that was, as we mentioned, it was our large customer's timing of specific fulfillment of their demand, and that was related to NAND. And with respect to the second part of your question, as I said before, that we do see evidence of inventory consumption by our customers, and that gives us confidence that this inventory will largely work itself through the system in the first half with improvements in demand in the second half. And we have also, certainly in recent weeks, seen some new orders from certain customers that previously had high inventories. Of course, too soon to use this as any kind of a slope in terms of demand buildup from here on.

But bottom line is that ultimately, the end-market demand drivers continue to be strong. The consumption of DRAM by our customers in the end markets, particularly cloud markets, continues to be healthy. It's just that those customers are using up their inventory to meet that demand. And over the course of next few months, by mid this year, that inventory will be returning to normalized levels to a large extent.

Then that will provide for opportunities for greater demand in the second half of the year compared to the first half.

Operator

Our next question comes from Joe Moore of Morgan Stanley.

Joe Moore -- Morgan Stanley -- Analyst

Just curious on the inventory side for the May quarter, and sort of Harlan's question, to the extent that you built 27 days of inventory, this last quarter, it seemed like a fair amount of this production went on to the balance sheet. Wouldn't you need to have bits and pieces? You only had them in the first half a little bit. And I guess, I just wanted some computation around that.

Dave Zinsner -- Chief Financial Officer

You've got a little noise to that. OK, let me see. I'll take a crack at it, Joe. There's some background noise.

So yes, as you cite, inventories were up. We do expect DRAM to grow sequentially in the third fiscal quarter. That combined with -- are starting to see some benefit. So those should be the combination of the reduction utilization and the capex activities.

Our capex reductions that we've already put in place will start to adjust the bit output growth. So we will start to see some benefit from that in the fiscal third quarter. As I've said we do expect inventories to be up again next quarter. But then, as Sanjay cited, a lot of the customers that have built up inventory will start to be back to normal and they'll start to place heavy orders on us again in the third and fourth calendar year.

We'll start to see these reductions in the DRAM inventory levels.

Operator

[Operator signoff]

Duration: 61 minutes

Call Participants:

Farhan Ahmad -- Head of Investor Relations

Sanjay Mehrotra -- President and Chief Executive Officer

Dave Zinsner -- Chief Financial Officer

Ambrish Srivastava -- BMO Capital Markets -- Analsyt

John Pitzer -- Credit Suisse -- Analyst

CJ Muse -- Evercore ISI -- Analsyt

Harlan Sur -- J.P. Morgan -- Analyst

Timothy Arcuri -- UBS -- Analyst

Karl Ackerman -- Cowen and Company -- Analyst

Blayne Curtis -- Barclays -- Analyst

Joe Moore -- Morgan Stanley -- Analyst

More MU analysis

This article is a transcript of this conference call produced for The Motley Fool. While we strive for our Foolish Best, there may be errors, omissions, or inaccuracies in this transcript. As with all our articles, The Motley Fool does not assume any responsibility for your use of this content, and we strongly encourage you to do your own research, including listening to the call yourself and reading the company's SEC filings. Please see our Terms and Conditions for additional details, including our Obligatory Capitalized Disclaimers of Liability.