Description

It is an opportune time to revisit Western Digital (WDC), which was last posted by jmxl961 in May of last year at $89.  At that time, the street, based on company guidance, estimated that Western would earn around $12 in fiscal 2018 (ending June 30).  Actual earnings for fiscal 2018 came in at almost $15 per share. In terms of risks at the time of the prior write-up, we were much more worried about accelerating HDD secular decline and potential negative outcomes from the messy dispute with Western’s flash partner Toshiba.  Neither of these turned out to be worth losing sleep over, but we were willing to live with them to gain exposure to Western’s flash business, with its very strong secular trends, at a very appealing valuation. We did indeed get exposure to strong trends, but it turned out to be a cyclical decline due to NAND over-supply.  After a weak revenue and margin guide for the rest of the calendar year, the stock now sits at $71. Even with weaker than originally anticipated flash results, Western should earn $11-12 over the next twelve months, making it the third cheapest name in the S&P 500. The stock price tells us that the market thinks the NAND cycle gets worse before it gets better.  We don’t have the clarity at this point to say that the market is wrong on that score. However, we believe that the strong secular trends are still intact and have decent visibility to $15 per share free cash flow and a $150 stock within two to three years. In addition, we suspect that there is limited trading downside from here, as it is difficult to come up with reasonable scenarios where the company earns less than $8.  Applying an 8 P/E, gets to $64 per share, or 10% downside. At that price, we would be paying little, if anything, for the NAND business, based on backing out Seagate’s (STX) valuation for Western’s HDD business.

We have arranged this write-up to begin with updates since the last writeup was posted followed by commentary seeking to dispel the main bear points as we see them.  Finally, there is additional background information on Western and its business lines for those who are new to the name. Please also see the numerous quality WDC write-ups posted here previously.

Updates

Fiscal 2018 earnings results / LT Model Update

Western reported earnings last week for its fiscal year ending June 30.  Initial earnings guidance coming into the year was >$13 for fiscal 2018. It was then raised to $13.50-13.75.  Actual earnings came in at $14.73. For the quarter, the company met its prior guidance with some impressive cost cutting but took down its margin guidance for the remainder of the calendar year and guided to an underwhelming $5.1-5.2 billion in revenue, flat with the June quarter despite going into what has historically been a seasonally strong period.  Margins were guided down almost 300 bps q/q on the back of NAND ASP declines, which we discuss in detail below. Inventories came in high, up almost $600 million y/y, especially considering a flat to declining outlook. The guidance summary for fiscal q1 is as follows:

One of the analysts asked on the call if the prior guidance of $13 per share for calendar 2018 was still good.  The company confirmed that it is still confident in exceeding that number. Thankfully, the HDD side of the business is very strong.  Gross Margins of Seagate (Western’s public HDD competitor) expanded over 200 bps q/q while guidance called for 5% sequential revenue growth.  Both Western and Seagate said they were on allocation in their HDD businesses and running at full utilization. Here are the implied results for the next two quarters and how a full year would look if we annualized those quarters:

In an apparent bone thrown to frustrated investors, management took the opportunity to update its long-term financial model, which the company has operated well above for the past year and a half.  Few seemed impressed though, as the stock made a new 52-week low.

We had previously thought that margins might be sustained above 40%.  Time will tell if the company is again being conservative. Regardless, the implied $11+ of earnings is enough to support a $71 stock once the negative sentiment eases.    

Other relevant updates from the call were that Western and Toshiba were discussing pulling capital from the flash JV to reduce capacity in an over-supplied market and that the company planned on buying $1.6 billion in stock per year as a part of a new $5 billion repurchase program.  The tone of the call was positive but the message was muddled. Reporting for the flash business has always been lacking – the company does not even publish revenue and margin for the division. With the rapid deterioration in NAND, investors are trying to find some grounding in the supply demand/macro as well as some issues specific to Western like slower bit growth and SSD execution misses.  We left the call frustrated with management but wanting to add to our position given the valuation.

Toshiba dispute resolution

WDC had been engaged in a serious dispute with Toshiba over the sale of its interest in the companies’ three flash JVs.  This dispute had prompted concern that Toshiba would block Western from participating in future NAND fabrication facilities, the output of which WDC needs to successfully operate its business.  Dating back to 1999, Western (previously SanDisk) and Toshiba have a seventeen-year successful history of working together on manufacturing 13 generations of NAND flash memory. The JVs are widely considered one of the most successful partnerships in the history of the technology industry.  The relationship between Toshiba and Western began to fray last year when Toshiba initiated a process to sell its Memory business, which holds its interest in the JV. Toshiba needed to raise capital to cover billions of losses incurred in its Westinghouse Nuclear division. Provisions in the JV agreements give Western broad consent rights in approving any change of control.  Western became very concerned when both competitors and key customers began negotiating a transaction with Toshiba without first seeking Western’s consent. The spat quickly devolved into corporate trench warfare, with both companies suing each other in multiple jurisdictions and even locking employees out of offices. WDC was part of a KKR consortium also interested in purchasing the Toshiba interest.  Certainly, there were compelling reasons for Western to fully control one of the top two NAND manufacturing companies. WDC, however, was stymied for perhaps overplaying the leverage it had in the situation and not giving Toshiba management enough room to save face in an already humiliating situation. In September, Toshiba agreed to sell its JV interest to a Bain Capital-led group, including WDC competitors Seagate and Hynix, and WDC customer, Apple.  This announcement, a clear violation of the JV agreements, put further pressure on the stock.

On December 12, 2017, Western, Toshiba and Bain announced that they had reached a settlement in their dispute.  Included in the agreement are: 1) extensions of two of the JVs to 2027 and 2029, 2) participation by Western in Fab 6, the new 3D NAND manufacturing facility planned by Toshiba, 3) agreements to build a new wafer fabrication facility in Japan, 4) renewal of joint R&D commitments, 5) establishment of IP protections in the event Toshiba sells some or all of its JV interest to a third party, especially a customer or competitor, 6) provision of Western’s consent to Toshiba transferring its JV interest.  While Western did not exit this fiasco with full ownership of the JV, we are not discouraged by the outcome. Purchasing the balance of the JV would have been a very large $18 billion acquisition, further leveraging WDC to capital-intensive manufacturing near the top of the NAND pricing cycle and adding more debt to the balance sheet. Moreover, the company will retain one of only a few seats at the top NAND manufacturing table, giving it critical influence in NAND development and strategic access to high quality components without the large capital outlay.  The settlement also eliminates the significant specter of Western losing access to NAND capacity, the risk we previously had worried about the most in this dispute.

Refinancing

On November 8, 2017, Western announced the repricing of approximately $3 billion of term loans at LIBOR + 2%, 75 bps lower than its prior term loan priced in March 2017.  The maturity remained April 29, 2023. Annual interest savings was $22 million.

On November 17, 2017, Western announced the prepayment of €874 Euro-denominated term loans.  Interest savings were $28 million + $20 million in hedging cost savings.

On November 29, 2017, Western announced increased capacity on revolver from $1 billion to $1.5 billion.  On February 27, 2018, Western further increased its revolver capacity by $750 million to $2.25 billion from $1.5 billion.

On January 29, 2018 Western issued $2.3 billion in 4.75% senior notes due 2026 and $1 billion in 1.5% convertible senior notes due 2024 ($121.91 conversion price, 40% premium at the time) to retire all of its $1.875 billion 7.375% senior notes due 2023 and tender for its $3.35 billion 10.5% senior notes due 2024.  To offset potential dilution from the coverts, Western repurchased $155 million of stock in February, with the ability to repurchase an additional $500 million.

The net result of the transactions listed above was to reduce debt by gross $1 billion, term out maturities, increase liquidity and reduce annual interest expense from $808 million to $440 million ($1/share after tax).  We were originally not thrilled with the dilution from the converts or the $1 billion of transaction costs and would have preferred the company retire a portion of the debt at maturity. Clearly from here, the converts are way out-of-the-money, and we would happily be diluted at $122 today.  Based on our conversations with management and their actions, it appears that management would rather term out debt and keep excess cash to retain the flexibility to 1) execute on acquisitions and, 2) in the absence of deals, buy in shares.

Stock Repurchases

On November 9, 2017 Western announced that it had “board authorization to resume potential repurchases of its common stock under a previously approved repurchase program.”  The existing authorization, which was suspended after the Sandisk acquisition announcement, was for $5 billion, of which $2.1 billion was remaining at the time.

Western began buying shares the following quarter, with $155 million of repurchases.  Last quarter Western purchased $436 million after indicating it would buy $500 million.  The splashy buyback announcements without much follow through are another source of frustration for us.  We believe real buybacks will begin in earnest this quarter.

Now that the company resolved its JV dispute with Toshiba, the company is likely to be far more aggressive in repurchasing shares and paying dividends.  Prior to April 2015, the dividend had been doubled in a little less than 2.5 years.

Prior to the large acquisitions over the last decade, Western had a stated policy of returning half of its free cash flow to shareholders.  We expect that policy to resume now that Western has established itself as a leader in its markets and was unable to execute on buying Toshiba’s memory business.  Peer Micron recently announced a similar policy, and Seagate is notoriously very aggressive with capital returns. We could envision Western repurchasing a significant number of shares (20-30% of outstanding) over the next three years and then materially increasing the dividend.  Regardless of the specific approach taken, the current valuation calls for more aggressive capital returns in our opinion.,

Capital Structure

As of June 30, 2018, WDC had $5 billion of cash, $11.4 billion of gross debt and generated $6.3 billion of EBITDA over the last year.  Thus, WDC possesses a very comfortable 1.0x trailing Net Debt/EBITDA despite levering up to pay $14 billion to purchase SanDisk in May 2016.  On our base case EBITDA for fiscal 2019, we still expect leverage to be around 1x on net debt/EBITDA, leaving the company with significant financial flexibility to return capital and/or pursue acquisitions.  

Western had an average diluted share count of 307 million at the end of the June quarter.  With its targeted $1.5 billion in annual share repurchases, and at the current share price, Western could easily retire 7% of its shares annually.  

WDC currently pays a $0.50 quarterly dividend, which yields 2.8%.  This dividend has not been increased since April 2015 because capital was required to execute the SanDisk acquisition.

Valuation / Model

The uncertainty around the NAND cycle has created a unique opportunity to purchase this storage leader for 6x run rate earnings.  Typically, this type of valuation is placed on “melting ice cubes” with declining end markets and deteriorating financial results.  Western, however, presents us with strong long-term growth prospects – the company projects a long-term revenue growth rate of 4-8%.  Our base case model calls for revenue growth slightly below the low end of that range. We assume a decline in HDD units is offset by ASP growth.  Industry revenue growth and margins are currently well ahead of our HDD assumptions. For Flash we assume the current margin environment continues but revenues begin to grow again due to demand elasticity.  Assuming Western holds market share, the implied TAM is some $12 billion below industry estimates for 2021.

For operating costs, working off of non-GAAP assumptions, we assume some modest inflation over the next three years.  This could flex up or down depending on underlying business strength. For example Western managed costs down last quarter in the face of weak demand.  Conversely, we could see the company investing behind strength in the HDD business if that market remains stronger than we assume here. We hit them for stock comp in earnings but assume dilution of 1.5% per annum..  Interest and tax rate are as per guidance, the latter growing to 10% next year. We assume cash capex goes from 6% of revenue to 7%, within the guided long-term range. Now that the company has termed out its debt and further debt paydown is unlikely, we assume that all free cash flow goes to share repurchase at $115 per share.  Walking this model forward, we arrive at over $15 per share in 2021 free cash flow. Please note that the 2018 column below is not actual results, but is more of a representation of run rate.

If one assumes that a storage company should trade at 10-11x earnings, the market is telling us that WDC’s earnings should decline to below $8 per share due to weakness in the flash business.  Based on the discussion here, we think that is highly unlikely. As earnings grow again after the NAND correction, the company’s price/earnings ratio should rerate higher to 12-14x earnings, implying a WDC stock price above $200.  However, to be more conservative we use a 10x multiple on 2021 FCF and still get to a price target of $150.

While one may quibble with the appropriate multiple, there is no denying the fact that without stock price appreciation, Western’s valuation gets cheaper every quarter with the amount of free cash flow coming in.  Over the next few years, the company will be able to achieve a net cash position even under the most draconian assumptions. This will allow the company to pursue more aggressive capital allocation strategies, either acquiring suitable companies as it has done in the past, dramatically increasing the dividend, or repurchasing a considerable number of shares.  Most likely, some combination of the three will be in order. Regardless of how the company chooses to deploy the free cash flow, the debt paydown should accrete value to the share price. Should the valuation continue to languish for whatever reason, we think the company becomes a target for private equity. The storage space has seen many private equity deals over the years, from Seagate going private to the most recent Bain/Toshiba deal.  Western’s cash flow and leverage profile would make it an attractive LBO candidate. We also think Micron would be interested in buying Western at this low valuation. Putting DRAM and NAND together makes good strategic sense, and combining the two companies could be done in a highly accretive way due to potentially vast cost-savings at the fab and corporate levels.

Bear Case 1:  Sustained NAND Oversupply

NAND pricing has clearly rolled over for this cycle, so those skeptical of Western’s margins have been proven right so far.  Many industry participants indicate that the market is in a state of modest over supply while, on the demand side, mobility has been soft over the last two quarters.  Coming into the year, TrendForce forecasted that 2018 NAND flash prices would decline 10-20%. IDC made a grimmer (and, as it turns out, more accurate) forecast, calling for a 35% decline, though they had to take TAM numbers up when the market tightness persisted through the winter (maybe they come back down).  Since what appears to be the cyclical peak for margins in the March quarter (~4 months), commodity NAND prices have already declined 35%.

Given the elasticity of demand for NAND and the opacity of supply additions, it is very difficult to predict short-term imbalances.  Comments from various companies in the space point to bit supply and demand growing by 40-50% in 2018. Importantly, storage manufacturers like Western can have a very viable business as both pricing and cost roll down the Moore’s Law curve.  In fact, storage companies would be unwise not to underwrite sustained double-digit annual ASP declines. As the following graph shows, the trend of NAND price declines is well-established and everyone should expect it to continue. One can run countless scenarios on revenue outcomes with the key variables of bit growth, unit pricing and unit cost.

Western has called for a normalization in pricing and estimates long-term NAND prices will decline 15-20% annually, not dissimilar from the prior five-year period but certainly inconsistent with recent trends.  At this pace (-15-20%), given Western’s annual unit cost reduction roadmap of -15-25% annually, Western can deliver sustained solid margins (we believe in the high 40% range). It is no surprise that Western’s margins have eroded with the rapid ASP correction this year.  We estimate that Western’s flash margins declined 500 basis points q/q in the June quarter. Our read on the weak results, based on peer results and conversations with various participants, is that mobile phone demand for the quarter was weaker while supply was in line with expectations.  We think this might be due in part to over-ordering/a channel inventory correction by customers like Apple. In general terms, this is not a structural change. In the following table we have made a guess on quarterly HDD margin, primarily using STX’s margin as an input and backed into WDC’s implied flash margin:

Rolling up these three variables/assumptions – bit growth, unit cost and ASP – gets to a flash TAM that grows at an 8% CAGR through 2021.  We still think these numbers are good, if not conservative; however, visibility into the next year is quite a bit murkier. IDC thinks the flash TAM will decline another 5% to $56 billion in 2019 before beginning to grow again in 2020, but thinks that further price declines will spur new demand and device content gains (e.g., new 512GB smartphones, continued data center strength and higher P.C. attach rates).  With bit growth demand growing at 40-50% and costs shrinking at 20%, one must assume very large, sustained pricing declines to claim that revenues will materially decrease. A good rule of thumb is that pricing should decline at half the pace of bit growth. Both Gartner and IDC estimate a $60 billion industry going to $80 billion over the next three years. After the normalization period, our longer-term view is that bit growth should outpace pricing declines, leading to modestly growing revenue and margin.  This view was recently confirmed by WDC management on a conference call. The key question is where do normalized flash margins shake out? For our modeling purposes, we are assuming the middle of Western’s new long-term range. Western has been conservative in putting out this range in the past, and we expect the same for the new range. Most importantly though, the current valuation does not force us to underwrite anything higher than really the low end of the range. If it turns out to be better, great. We walk through some of the supply and demand dynamics below in this section.  In preview, we believe that demand will surprise to the upside and the supply outlook is very manageable.

Significant capex has been going into NAND supply over the last several years.  Worldwide, Gartner estimates some $18-20 billion per year from 2016-2018 is being spent on NAND capacity - up from $14 billion in 2014.  Gartner expects this number to stabilize in the $17 billion range during 2019-2021. There is no question that these are massive amounts of capital, but it is required to keep up with NAND demand growth, which is growing at a 40%+ CAGR.  Moreover, the 2016-2018 spend should be viewed as catchup, as NAND production capacity growth was only 5% and -1% in 2012 and 2013, respectively, way below end market growth. Additional cap-ex was necessary for NAND producers to transition to 3D NAND, a chip with a new architecture that allows Moore’s Law to continue.  There is no question that the 3D transition is going well, and that supply has started growing faster. In 2017, however, NAND was in short supply, and spot prices for NAND remained stubbornly firm, rising some 50% against expect price decline trends. Now that supply has caught up with current demand and prices are falling, industry participants are considering ratcheting back on capex and planning capacity growth.  As mentioned above, Western and Toshiba are discussing this currently. Any change in capital would not be visible in capacity until early calendar 2019. Samsung has made similar comments, repurposing at least one fab for DRAM that was previously slated for NAND. While not a duopoly like the HDD industry, the NAND industry is highly consolidated, with Samsung and Western/Toshiba controlling roughly 2/3 of output. Despite the current hopefully temporary over-supply, the players seem to behave rationally.  Chinese supply may be a wild card, but it is not a factor currently and is not expected to be for a few years. In addition, the Chinese fabs are multiple technology generations behind the market leaders.

The NAND shortage in 2016-2017 prevented many new NAND applications from ramping and slowed the replacement of HDDs with flash storage.  NAND is widely believed to have the most elastic demand profile of any semiconductor – demand is almost insatiable as prices fall. This compares to other components such as microprocessors and DRAM, which have far fewer new applications.  Thus, as NAND prices fall, which they inevitably will consistent with Moore’s Law, NAND content in existing end markets increases and countless new applications emerge in a virtuous cycle of additional NAND demand. The same trend occurred with HDDs and virtually every other storage/processer technology.  In the early 1980s, one gigabyte of HDD storage cost $500 thousand, making almost all applications uneconomical. Today, it costs a few pennies and is everywhere. The growth in NAND content in each subsequent iPhone model is a perfect example (see chart above right). Mobile phones have been the largest end market for NAND in recent years, accounting for as much as 40% of NAND output and largely contributed to the recent NAND shortage.  Average NAND capacity per smartphone was up almost 50% y/y in 2017. Apple launched its iPhone 8 and iPhone X last year, with the highest models holding 256GB of NAND, a far cry from the original iPhone at only 4GB. Samsung has announced its next generation of flagship phones will hold 512GB of NAND; Apple will almost certainly follow suit. As NAND prices decline, Apple and other NAND buyers can continue to increase NAND content in devices without having to raise prices.  For this reason, NAND content growth has outpaced smartphone unit growth by more than 5x (see Gartner chart above left).

We suspect investors severely underestimate the magnitude of pent-up NAND demand.  First, as mentioned above, high NAND prices slow SSD penetration in PCs. Only 30% of notebook PCs shipped today have SSDs.  Gartner projects that mobile and desktop PC units that ship with SSDs will grow to over 70% by 2021. The graph below shows that SSD growth is accelerating because of this while mobile phone growth remains steady.  Seagate estimates that incremental NAND capacity just to replace these PC HDDs would require an addition $40 billion of capex, roughly two years of run rate spending. Further contributing to the SSD growth is SSD penetration in cloud and enterprise storage.  In the third quarter of 2017, 3.7 EB of SSDs shipped for use in servers, up 150% year/year. Over the last four years, IDC estimates that SSD capacity shipped grew at a 105% CAGR but still only makes up 10% of the total enterprise storage market (HDD + SSD). This demand growth is highly elastic and will increasingly shift from HDDs when NAND pricing resumes its decline.  SSD enterprise capacity shipments overtook mission-critical (think “hot data”) HDD capacity shipments in 2016 for the first time. SSDs also became the largest user of NAND flash in the first quarter of 2017, surpassing smartphones. Due to this accelerating growth, in the next few years SSDs will become even more important to the NAND outlook. Today most people think about it as primarily as smartphone input.     

The final and potentially most significant area of new NAND demand is IoT devices, which are still relatively in their infancy.  In the not too distant future, virtually every electronic device will be connected to the internet and need to store data. When one thinks of data input today, one likely thinks of a person typing on a keyboard.  In contrast, IoT devices will be inputting data from sensors that record every imaginable aspect of the world around us. As Peter Levine of venture capital firm Andreesen Horowitz describes the step change coming from IoT data creation, “It will literally be orders of magnitude, in the exact mathematical sense.” (One order of magnitude is 10x; two orders of magnitude is 100x).  NAND is the ideal memory for IoT devices considering cost, physical size and speed. We see very little in NAND forecasts that captures the impact from IoT device proliferation, especially as significantly as Mr. Levine expects.

Taking cars as an example, cars connected to the network will send 25 GB to the cloud every hour according to Hitachi.  This compares to less than 1 GB for an hour of streaming HD video. Google estimates that a self-driving car will generate 1 GB of data per second.  All this data must be at least temporarily stored on the car, and then a curated portion of that data is stored in the cloud. Michael Huonker from Daimler AG R&D forecasts that SSD storage in cars will grow to more than 1 TB per car from almost nothing today.  For illustrative purposes, if one assumes that each car sold in the US has this amount of storage, the annual storage demand from cars alone would be about one-third of 2017 NAND output. We don’t expect this anytime soon, though Gartner forecasts that there will be 250 million connected cars on the road by 2020.  Of course, this is just one of innumerable potential applications. Considering the cloud storage implications of the data created, one can quickly envision a world where it is virtually impossible to create enough storage capacity to meet the need. The broadening of NAND end markets beyond PCs and phones to the wide variety of IoT devices likely smooths out the product and business cycle as well.

We are confident that between the richer product offering and solid manufacturing, WDC will skillfully navigate any potential brief periods of oversupply and adjust accordingly.  Manufacturing of NAND is consolidated among a few players; Samsung and the Toshiba/SanDisk JV hold around 70% market share together, much greater than that in leading edge NAND. Both parties have demonstrated significant discipline.  In fact, Samsung recently announced that equipment for a new fab would be for manufacturing DRAM, not NAND. The upshot is that NAND capacity spending is driven by a few disciplined players with varied products competing for capital. Private equity firm Bain Capital entering the mix by acquiring Toshiba should only increase capital discipline.  The Toshiba dispute has also likely delayed the Flash JVs own capacity additions. Additional participants who might be attracted to the high margins and growth outlook lack the IP, manufacturing experience and scale required. In addition, the capital alone is daunting – new fabs cost $10 billion. Analysts seem to assume that capital will continue to flow freely into the space.  China has indicated that it will enter the market, but this is several years away and will only be for second-tier products. According to Western, these Chinese product roadmaps are multiple product generations behind Samsung and WDC/Toshiba.

The final key consideration when discussing the prospect of sustained NAND oversupply is the fact that Western primarily sells high end SSD drives and NAND solutions, only selling a limited amount of commodity NAND.  WDC’s solution product portfolio continues to include more value add such as software, integration, and special purpose chips which fetch up to 6x the price of what commodity drives receive. NAND is actually an input cost to these drives.  Yet, the value of their end products to customers is based on market demand for specific products. Just because NAND prices decline does not necessarily mean that ASPs for Western’s products will decline too. Western is in the fortunate position of being able to focus on high margin drives that utilize its specialty software and leading manufacturing and cede low margin commodity share to other players.  The fact that WDC currently has the highest gross margins of any NAND producer highlights the value of its solutions and the efficiency of the Toshiba JV (see margin slide on p. 10). So far in this ASP correction though, Western has not been immune to pricing and margin hits: last quarter Western’s flash ASPs declined high-single-digit with a similar decline expected this quarter. We believe part of this decline is due to execution issues with Western’s SSDs.  We have a call into the company to get more color on some comments that were made on the conference call confirming these execution issues. It appears Western may have been designed out of a few applications in the quarter, leading to underwhelming bit growth and poor ASPs. All things being equal though, as Western sells more SSDs with its own controllers, the sales mix improves as do average ASP and gross margin.

Bear Point 2 – HDD is in secular decline, so WDC is over-earning.  Margins HDD earnings will crater as HDD decline accelerates.

“For capacity storage, hard drives will remain the most cost effective option.  They have a substantial cost advantage today and even in 2020 we expect that advantage to remain very significant at four to five times.” – Mark Long, WDC Chief Strategy Officer, Western Digital Analyst Day, 2016

This is the easiest point to get comfortable with in our opinion.   The basic bear argument on HDDs is that the secular decline of PCs, the move of storage to the cloud and conjunction with the growth of SSDs are quickly killing the HDD industry.  The narrative follows that HDD companies may show decent earnings today, but they have no terminal value, so paying what looks like a cheap multiple is still too expensive. The reality is actually quite different from this.  First, we acknowledge PCs are likely in permanent secular decline, and the HDD drives that go into PCs and other consumer devices today are quickly losing share to SSDs. We believe that SSDs will eventually fully replace HDDs in PCs.  Gartner forecasts that HDD units shipped will decline by 50% over the next four years. Historically, analysts measured the success of the HDD business in units sold, a vestige of an industry once driven exclusively by PC sales volumes (WDC still reports this every quarter).  Thus, as HDD unit declines accelerated with the adoption of flash drives on notebooks and desktops, most analysts deemed the business a commodity with gross margins likely headed to 20%. The view of rapid extinction for HDDs is still very prevalent. The last few years demonstrate a different outcome though.  Despite a 40% decline in units, gross margins for HDDs have increased from low twenties to low thirties, and we think could approach the mid-30s. Furthermore, revenue has only declined marginally over the last 6-7 years if you take out the artificial increase in ASPs due to the Thailand floods. How could this be?

While consumer devices and units shipped still drive the perception of the HDD market, the dollars are driven by the cloud.  It is true that the shift of storage from PCs to the cloud has accelerated the decline in PC HDDs. However, the cloud is not some magical place in the sky where the physical constraints of data storage don’t apply; data must still be stored on a device no matter where it exists.  What many ironically miss is that data stored in the cloud is more likely to be stored on HDDs than other storage medium. These companies require large capacity drives for their server farms. Despite the headwinds, overall storage maintained on HDDs will grow 30% per year on the back of the 40% growth in cloud storage.  HDDs are still the cheapest medium to store non-critical data (see graph below) and are still the workhorses of worldwide storage. Western estimates that by 2020, 75% of the exabytes in storage will exist on hard drives.

As the cloud storage companies make economic decisions in the buildout of datacenters, they consider:

1. How much it costs (total cost of ownership on a per byte basis) – HDDs are cheaper on initial purchase cost, but use more energy because of the moving parts.  Some think we are approaching the cost indifference point between HDD and SSD. This may be true for certain applications; however, costs for both HDD and SSD have tended to move down in parallel.

2. How often the data will be accessed – “hot” data, or data rapidly available on demand, is better suited to flash due to quicker access speed.  Archival data, information that must be stored but is unlikely to be accessed much, is better stored on HDD considering cost and durability. The energy cost mentioned above is significantly minimized if the data are rarely accessed.

3. How long the data will be stored – SSDs tend to wear out much faster than HDDs.  

4. How much physical space the drive will take up – some data centers are located in expensive real estate and/or areas where the potential data center footprint is limited.  SSDs are bit-for-bit smaller than HDDs. Minimizing the footprint can have real economic advantages.

The criteria above do not necessarily distill into a binary decision; many applications will be ideally suited for a combination of both HDD and SSD.  Indeed, an increasingly important business for WDC is providing the software and hardware integration to meld the two into a seamless solution. Generally speaking though, HDDs will be increasingly relegated to long-term enterprise and cloud storage applications in which cost is more important than speed, otherwise losing share to SSDs.  

Furthermore, innovation in HDDs has not ceased.  Western’s Helium drive has several advantages over other HDD products like running cooler and utilizing the media more efficiently.  Helium, being less dense than air, allows the disk to spin with less drag providing for increased precision, lower energy consumption and a longer useful life.  Western has compared the Helium drive’s operation to driving a car at 80 mph with your hand out the window and being able to keep your finger steady to 1/100th the width of a human hair.  This speaks to the proprietary nature of WDC’s products in an assumed commodity industry.  Leading edge capacity for HDDs is 10-12TBs, which is in high demand for hyperscale companies.  This market is basically a duopoly between Western and Seagate, though Western still holds over 60% market share and is clearly one product generation ahead.  Western holds a large HDD patent portfolio that should allow the company to keep its edge over Seagate. List prices for helium drives can be several hundred dollars versus $25 - $50 for a commodity notebook or PC drive.

Growth from hyperscale players / enterprise datacenters should offset much of the PC-related HDD TAM decline.   As HDD replacement on PCs tapers off, WDC actually believes they could see revenue growth for HDDs with much improved margins - see the TAM graph above calling for the legacy HDD TAM to be flat through 2021 while the capacity enterprise HDD TAM will double over the period.  End-market demand should look a lot less cyclical as well, a seldom discussed benefit from the secular decline in PCs.

HDD profitability should remain very steady or only decline modestly, and this business should continue to generate meaningful cash flows for years to come.  Cash flow will be stable on declining units because average selling prices (ASPs) of HDDs should increase 50% as the mix shifts dramatically to higher margin enterprise and cloud drives (like the helium drives mentioned above).  Our base case calls for Western’s HDD revenue and margin to remain flat over the next three years, at which point it will be less than 40% of total gross profit, though still generating $4-5 per share in earnings. This may seem like an aggressive assumption considering the negative perception of HDD’s outlook.  However, at least over the next year, HDD looks more like a growth business than one in secular decline. From Seagate’s last quarterly conference call (5/1/18):

So if you remember the history of the nearline markets in the last five years, there's been periods where we interpreted it as seasonality, but there were periods of feast and famine, if you will. The market – generally, the exabytes were increasing at about a 35% CAGR, but there were some markets sometimes where the market was much faster than that and other times where it slowed down.  We're starting to see that abate and the demand be a lot more predictable, longer lead time visibility. And some of that's because the build-out of major cloud installations pulling really hard on high-capacity drives is not as feast or famine, anymore. I think behind that, and this is something that is still developing a little bit, are the more legacy technologies for on-prem or smaller data centers where the capacity points aren't necessarily the highest capacity points, so we'll say 8 terabytes and below, but the demand is quite strong and quite diverse, geographically.  So, from my perspective, that strength actually picked up middle of last year and feels strong through the back half of this calendar year as well. Looking out into calendar year 2019, I think it may be a little bit premature. We'll see the size of the back half of this year first, but it's certainly feeling like the install patterns that most of the cloud service providers and some of these big on-prem companies are in the middle of are still very healthy. And so, I wouldn't be surprised at all if it does extend for another year.

At some point, likely well beyond our multi-year investment horizon, SSDs may become cost competitive for data center storage, ushering in an accelerated terminal decline for HDDs.  Having half of Western’s business in flash is somewhat of a hedge against this eventuality, as the two storage media are substitutes (though they can clearly move up or down together too).  We have heard some express concerns about what HDD margins look like once decline sets in and decided to look at historical decremental margins to illustrate (see table below). This is a high fixed cost business, but no one is adding capacity.  The HDD market is essentially a duopoly between WDC and STX, each with 40-45% market share.  Both companies have managed the businesses for cash for many years, spending enough capex to keep facilities operational at around current drive capacity but not heavily investing.  Western announced on its last conference call that it would be closing an HD facility in Kuala Lumpur, to stay competitive with Seagate in having only three HDD sites.  In terms of HDD facilities capex, Western estimates 4-5% of revenue ($400-500mm).  HDD capex for STX, which has a slightly smaller but substantially similar HDD business, has been running at around $400 million, which makes sense.  As compared to memory fabs, HDD facilities/equipment have a longer useful life and lower maintenance capex requirements.  Most of the operating cost is labor, which both WDC and STX have gotten really good at ratcheting up and down with demand.  To quantify decremental margin, we looked at two periods – 1) the global financial crisis in 2008-2009, and 2) 2015-2016, which can be characterized as the acceleration of PC secular decline.  Seagate is more illustrative in the latter period because WDC had acquired Sandisk by 2016, at which point we lost the ability to track its HDD margins.  It makes sense that HDD decremental margins were better in 2015-2016 as the companies had more time/visibility to react and adjust costs, whereas the global financial crisis was a relatively quick demand shock.  Both companies were still looking at HDD as a secular growth business at that point (i.e., not managing for cash).  Based on our understanding of the cost structure, we would have intuitively predicted a forward-looking decremental margin closer to the 2015-2016 period (30-40%).  Clearly, it could be worse than this over the short-term if there is a GFC-type demand shock.  

At the point of accelerated secular decline, Western can pull back from and eventually eliminate its estimated $600 million of HDD research & development.  Both Seagate and Western already have ongoing significant cost reduction programs and are active in eliminating unneeded production capacity. Finally, Western still has additional Hitachi synergies to realize, which should accrue additional profitability to the HDD business.  

For better or worse though, the market does not seem to be overly concerned with the HDD outlook though.  Seagate, essentially a pure play on HDD, gives us a near perfect market comparable for Western’s HDD business as the two have equal market share, similar revenue/earnings and share count Seagate, essentially a pure-play HHD company, trades at 7-8x EBITDA vs. WDC at 4.4x.  As a quick-and-dirty analysis if you take out $54 per share for the WDC’s HDD business, it leaves around $16 for the flash business.  Western’s flash business earns around $6 per share assigning half of the total company overhead, interest expense and income tax. Thus, we are implicitly buying the flash business for less than 3x earnings, even at these lower margin levels. Given the NAND business has better long-term growth prospects, we think it should trade at a large premium to the HDD business at 11x. The math is similar looking at the two on an EV/EBITDA basis, which is definitely a cleaner way to do it. Seagate also owns a small piece of Ripple, a crypto-currency business, which we have not valued here.  Even valuing the Ripple interest at $3 billion, the highest estimate we’ve seen, one arrives at the same conclusion: the market is mispricing WDC’s flash business. One could argue that Seagate is overvalued, though based on our case for WDC’s HDD business, we think Seagate is reasonably valued. Western’s HDD-only business traded as high as $115 per share in 2014 before acquiring SanDisk. It is also interesting to note that the implied value of WDC’s flash business is at a significant discount to what Western paid for SanDisk in late 2015 prior to the current semiconductor up cycle ($14 billion) and half of what the Bain consortium paid for Toshiba’s memory business ($18 billion).  As such, we feel we are buying both underlying business lines with a considerable margin of safety.

Industry Background and History

Western Digital broadly participates in the data storage industry.  There are four basic types of storage platforms in use today – magnetic tape, optical (CD/DVD), hard disk drive and solid-state drive (flash).  Tape and optical hold negligible market share in terms of sales today despite representing sizeable amounts of the existing base of stored data.  In this report, we will focus on hard disk drives and solid-state drives, which are Western’s primary business lines.

Hard Disk Drives (HDDs)

HDDs have been the most prevalent medium for the data storage industry over the last fifty years.  HDDs were invented by IBM in 1956 and use a mechanical arm with a read/write head (like a needle) to store data on a rotating piece of magnetic media.  These devices store significant amounts of data cost-effectively. HDDs are not semiconductors or transistors like SSDs (discussed below), which have no moving parts.  HDDs are mechanical devices, and as such, are slower and use more power than other storage solutions even after decades of improvement in speed, size, capacity and reliability.  Nevertheless, costing around $0.04/gigabyte, HDDs hold a significant cost advantage per unit of storage over SSDs (about $0.25/gigabyte). HDDs are ideal for long-term storage / archiving of data that may not need to be accessed quickly or regularly.  Examples of such data include photo libraries, CCTV camera archives, historical financial data like old bank statements and backup files.

As PCs began to supplant workstation and mainframe computers as the main tool for enterprise computing in the early 1980s, data storage based on cheaper commodity HDDs also became more widely adopted and replaced proprietary disk drive storage solutions utilized by mainframes and workstations.  As PCs proliferated over the next two decades, HDDs became the foremost storage solution. At the beginning of that period, there were dozens of emerging disk drive and disk drive component companies. As highlighted in the 1997 management book The Innovator’s Dilemma, the industry became synonymous with fierce competition and nimble, cost-effective manufacturing in Asia.  These companies would hire tens of thousands of workers to produce and ship tens of millions of disk drives in a matter of weeks during strong seasonal periods such as back-to-school and Christmas.  The fiercely competitive environment initiated a period of consolidation that lasted twenty years, with only the three most vertically integrated players surviving – WDC, Seagate Technology (STX), and Toshiba.    

In the early 2000s, there were five main PC makers – Dell, HP, Compaq, Gateway and IBM, which were the dominant buyers of HDDs, representing around 80% of WDC’s revenue at the time.  The storage business was notoriously cyclical, with Western’s and Seagate’s financial results wildly swinging based on PC product cycles. PCs are now less than 20% of WDC revenue and even a lower percentage of profit.  Mobile devices like smartphones and tablets, which primarily use flash for storage, have replaced PCs as the primary computing form factor and driven PCs into secular decline. In the remaining PC business, hard disk drives are losing share to flash for reasons we discuss below, further lessening the HDD industry’s dependence on them.  Lastly and perhaps most impactful, data is increasingly moving to remote storage in networked, large scale server farms, a trend referred to as cloud storage. Because of these trends, the HDD end market today is much more diversified and includes TV set-top boxes, gaming consoles, removable drives, automobiles and hyperscale infrastructure/data centers.  Hyperscale infrastructure players are companies like Amazon, Microsoft and Google, which offer cloud based computing platforms (discussed below).

Flash / Solid-State Drives (SSDs)

In the early 1980s, Toshiba invented flash memory, which utilized semiconductor technology to store information without a charge.  Semiconductors are, of course, the essential components of most electric circuits. They conduct electricity under some conditions but not others, making them a good medium for the control of electrical current.  This allows the storing of 1s and 0s, which the infinite combinations thereof allow for infinite storage and processing of data in electronic form. Semiconductors can generally be categorized into two types: those that store data – memory chips – and those that process data – processor chips.  There are multiple types of both categories of chips. Memory chips are usually broken down into volatile, those that lose their content without power, and non-volatile, those that retain their content without power. Hard disk drives are a non-semiconductor, non-volatile memory. DRAM (Dynamic Random Access Memory) is the most prevalent type of volatile memory.  Except for a few niche product types, NAND (Negative-AND, referring to the type of logic utilized) represents the preponderance of non-volatile semiconductor memory. NAND is a faster version of flash memory often used to accelerate the performance of I/O (input/output) intensive applications. The terms SSD, Flash, and NAND are often used interchangeably, sometimes referring to the type of device and sometimes to the underlying technology.  In contrast to HDDs, SSDs have no moving parts, use half the power, generate no heat or vibration and do not need to boot up. They also have much faster I/O rates, are lighter and more durable (you can drop an SSD and it will still function because it has no spinning drive). The primary disadvantage of SSD is cost. SSDs can cost around 5-6x HDDs or $0.25 per gigabyte depending on drive size. One other potential disadvantage for some applications is the limited write life for SSDs; many wear out after 100,000 writes and then need to be replaced.

Digital cameras began to meaningfully replace film in the early 1990s, substantially accelerating the utilization of NAND and making it the primary type of flash memory.  Flash memory cost improved throughout the 1990s and early 2000s while mobile phones concurrently began to require multimedia (pictures and video). NAND was a critical component to the success of the smart phone, enabling many of its multimedia applications.  As a result, flash became the default storage medium for cameras, cell phones and other mobile devices, taking share from other storage types. Today’s high-end iPhones ship with 256 GB of NAND memory, and Apple only references the amount of memory when advertising its phones’ features, not the fact that it is NAND.  Apple also doesn’t even reference the amount of DRAM, as was convention with PCs. This highlights that NAND has become the universal storage memory for mobile devices. More importantly, flash memory continues to gain wider adoption in enterprise and cloud storage devices, driving all of the growth in the Computing category in the graph below.  

Yet, manufacturing and selling NAND flash was widely considered a cyclical commodity business.  Anecdotally, a member of our team attended a presentation in 2003 by NAND manufacturer, SanDisk, at a Bear Stearns technology conference where he was the sole attendee.  DRAM initially was viewed as far more important and profitable than NAND in the 1990s, as it was the key semiconductor utilized in PCs other than the Intel microprocessor.  Micron Technology (MU), the leading provider of DRAM, traded to a $50 billion market capitalization during this period. Its success attracted foreign subsidized Korean competitors Samsung and Hynix and other Japanese and European competitors, which along with the technology bubble bursting, crushed DRAM industry profitability literally until the last few years.  In contrast, NAND manufacturing never received the focus and investment that DRAM did. Nonetheless, NAND innovation continued unabated, and as its costs per bit, durability, and access speed improved, it was used in a wider set of applications and in increasing amounts. DRAM, although seeing a bit of a renaissance recently, should not grow as fast as NAND since it always requires a charge to retain its information and is more expensive – 30x more than NAND in 2016.  DRAM’s one key advantage is its faster access speed. Thus, data are always stored in DRAM when about to be processed, but the preponderance of data requires being stored permanently. For smartphones and other mobile devices, it is permanently stored in NAND.

Data Growth and The Cloud

“Between the dawn of civilization and 2003, we only created five exabytes; now we’re creating that amount every two days.  By 2020, that figure is predicted to sit at 53 zettabytes (53 trillion gigabytes) – an increase of 50 times.” – Hal Varian, Google Chief Economist

Data creation is growing at an exponential rate, growing 10x in the past five years.  This growth is fueled by:

1. Digitization of photos and video – streaming movies on Netflix, uploading YouTube videos, and even uploading your personal photos and videos to cloud storage databases and social media applications like Facebook.  Taking YouTube as an example, the first YouTube video was uploaded in April 2005. Today, 5 billion videos are watched every day, and 300 hours of video are uploaded every minute. The rate of uploads has increased 6x in the last six years.  The storage impact of these uploads is magnified by increasingly higher video resolution technologies. For example, an Ultra HD movie requires 4x the data of an HD movie. Multiple camera, 3D and 360-degree filming add further multiples to the data created for video.  Today, more than half of all internet traffic is related to video. If virtual/augmented reality takes off, it will likely be the next step change in video data creation.

2. Big Data processing – Much of the data that will be generated in the coming years will be created and analyzed by a machine: machine learning.  Examples are mapping the human genome, searching billions of online flight schedules for the best deal and high frequency trading. IBM’s artificial intelligence program Watson can process 200 million pages of medical records in three seconds, but only if that data is digitized, stored and accessible.  The rise of Big Data has made every industry a technology intensive industry where seemingly every piece of data needs to be stored and managed, leading to a 6x increase in the amount of data deemed valuable. In an example of the “power of 1%”, Western estimates that if oil companies can reduce costs by just 1% by effectively mining for inefficiency the data they already generate, $90 billion of value would be created over 15 years.

3. Internet of Things (IoT) – This game changer began when our phones became powerful computers that we carry at all times.  Instead of only consuming and creating data at home or on our office computers, we now do so wherever we go. Smart phone sales are more than 10x larger than PC sales.  Enabled by 5G (the next generation cellular network which provides a 10x increase in speed, 10x reduction in latency, and 10x improvement in connection density), cars, medical devices, appliances, etc. will be increasingly connected to an instantaneous network all the time.  Western estimates that some 20 billion connected devices will be active by 2020, with the number of devices per user doubling. Many of these are new markets that will rapidly grow from a zero base in the next few years.

There will undoubtedly be numerous applications that generate and use data that we cannot yet envision in today’s data-centric world.  Certainly though, some of the big drivers in the next five years will be continued growth of cloud applications such as social media, autonomous driving, machine learning, virtual reality/augmented reality devices and ubiquitous sensors generating data for IoT devices, to name a few.  It is difficult for one to grasp the size of this data explosion and the resulting storage needs. One zettabyte equates to approximately 250 billion DVDs or 1,000 datacenters. To store the 53 zettabytes of estimated data created in 2020, you would have to cover all of New York City with data centers.  Many believe there will ultimately be trillions of IoT devices creating so much data that we could not conceivably store it all. Not all data created is actually stored; however, a meaningful portion of this data is stored for some period, locally on a device or remotely on a cloud storage platform. In addition, stored data is often backed up in a separate location, creating a multiplier effect on storage needs.  Cisco Systems expects data usage to increase at a 47% annualized rate from 2016 to 2021. Storage demand is growing even faster; Western estimates the amount of stored data in the world to quadruple between 2016 and 2020. The additional secular trend of data moving to the cloud has created a need for large scale data infrastructure buildout to store and manage that data. Western estimates that by 2020, 90% of all stored data will be in cloud servers.  That storage requires high-end, largescale components and related solutions to manage it, which is exactly Western’s business.

Why Western?

Western traces its roots to being a leader in HDD storage solutions.  Along with HDD rival Seagate, Western opportunistically consolidated the HDD industry over the last two decades.  In WDC’s case, its acquisitions were, seemingly without fail, well-timed and executed at attractive prices. Western and Seagate each retain 40-45% of the HDD market, and Toshiba holds approximately 10-15%.  WDC not only did a good job excelling within the competitive HDD industry, it successfully expanded its nascent NAND flash and enterprise businesses. The company correctly understood the transition to flash occurring in all areas of storage, especially in the enterprise and cloud.  Several years ago, Western started to complete varied modest acquisitions to enter complementary parts of the flash business, resisting pressure from Wall Street to use its large net cash balance for capital returns to shareholders. Western had almost half of its market cap in overseas cash at the time and would have faced large repatriation taxes if it chose to pursue dividends and buybacks with that cash.  In its $4 billion acquisition of Hitachi Global Storage Technologies (HGST) in 2012, WDC acquired a leading enterprise provider of storage, which had a profitable flash business for WDC to ramp. The Hitachi deal solidified its enterprise HDD presence, as well, where it previously lacked significant products and distribution (WDC sold almost exclusively into consumer PC end markets prior). Again, WDC made a prescient strategic decision to diversify away from PC HDDs.  Seagate, on the other hand, plowed its available cash back into repurchases and now finds itself at a strategic disadvantage and with a strained balance sheet.

With its $14 billion acquisition of SanDisk in May 2016, WDC went all-in on the future of flash and acquired the preeminent mobile storage company.  The SanDisk purchase was especially savvy, done in a down NAND cycle with cash Western patiently hoarded, again resisting pressure from Wall Street to return capital.  Overnight, Western became a leader in NAND. SanDisk, with its superior brand and implementation of flash (SanDisk originally was the leader in NAND for digital cameras), partnered in 2000 with Toshiba, the creator and most important manufacturer of flash.  These two companies, through their three Flash Ventures JVs, represent the second largest producer of NAND, slightly behind Samsung, producing 2 million flash products per day. Samsung and the JV each have approximately 35% of the NAND market. The JV secures NAND supply for WDC’s branded SSD products.  They also control virtually the entire high-end 3D qualified output such as that which ships in the iPhone X. The strategic importance of this JV cannot be overstated: Western essentially controls at cost half the NAND output of Samsung.  NAND is currently in short supply and likely will remain tight through much of 2018.

As a result of these strategic acquisitions, Western today stands out as the low cost, vertically integrated manufacturer of storage devices and solutions, leaving it best positioned to benefit from the exponential growth in data storage demand.  Western also possesses the best flash supply chain, packaging NAND into the most diverse product line with unparalleled portfolio breath and market reach according to Gartner.

Western operates in many secularly growing markets.  The company categorizes its end markets as 1) client devices – PC, mobile phone and consumer electronics HDDs and SSDs, embedded chips and wafer sales, 2) client solutions – branded HDD and flash removables (think thumb drives), and 3) Data Center Devices and Solutions – high-performance, high-capacity enterprise HDDs and SSDs, data center software and solutions.

By 2020, Western estimates the total addressable market (TAM) for its business lines will approach $95 billion.  We estimate that Western has approximately a 32% share of a $63 billion global storage component market today (FY 2017).  WDC expects the core component business market opportunity to reach $72 billion by 2020 and believes that data center solutions represent an additional $23 billion opportunity in that time frame, an over 34% TAM increase in less than three years (11% CAGR).  There are puts and takes within this TAM, which we discuss in following sections, with the biggest declines coming from client HDD and the biggest gains coming from data center devices and data center solutions. With the explosion of big data analytics, increased real-time access to more data, and multimedia and cloud applications, Western’s devices for enterprises and the cloud are key components to building out data infrastructure.  In selling storage and memory components to the largest technology companies in the world, Western believes that it has developed valuable IP and know-how in building and implementing storage architecture, effectively managing data and creating value-adding outputs from it. Western has begun to “move up the stack” and offer data center systems, platforms, software and complete solutions. These solutions include managing the storage mix (HDDs vs SSDs), private vs. public cloud storage, archiving, compliance, applications, analytics, security, etc.  This new Data Center Solutions business should be WDC’s fastest growing segment in the future and is an important step for the company. First, these solutions carry the company’s highest margins and provide further scale on an already vertically integrated company. They also put Western in the position of being a trusted partner to the hyperscale players and cloud enterprise customers instead of just a component supplier. Data center components are already about one-third of Western’s revenue. The resulting business should be less cyclical, as it does not ebb and flow with consumer product cycles, and deserves a higher valuation multiple.  The dichotomy between the perception that Western’s underlying businesses are in secular decline and the reality that exponential data growth needs more storage capacity and solutions is quite striking. While there will certainly be cycles in the storage industry, this secular trend will overwhelm any cyclicality over a multi-year period.

Western’s product breadth provides several advantages as SSDs and HDDs compete for the same storage TAM.  First, it provides a natural hedge. If NAND prices fall quickly, as many project, HDDs should lose share, but WDC’s SSD-related businesses should more than offset the HDD declines.  The reverse holds true, as well. Current WDC expectations are for the growth of flash to exceed the decline in HDD by more than four times. Second, WDC sits in an unparalleled position in terms of understanding manufacturing, the storage ecosystem and its customer needs, which should assist them in successfully building out the data center solutions business.  This product and customer insight should be improved with the company’s understanding and input to leading-edge NAND manufacturing. The results support these observations. Western Digital has built a $20+ billion storage company with an enviable customer base of enterprises and leading internet companies. Western also possesses the highest NAND margins in the industry according to Gartner.  This margin profile is indicative of a broad-based product solutions provider with leading manufacturing and IP, not a commodity hardware company. This reality is not appreciated by the Street. All the above has translated into superior return on capital for Western, which we expect to continue.

With its industry leading intellectual property, Western is poised to continue to innovate and create the next wave of storage technologies.  The company’s portfolio of 13,500 patents was ranked 4th in the world by IEEE, behind only Apple, Google and Qualcomm.  In fact, NAND rival Samsung pays Western hundreds of millions of dollars per year in IP licensing fees to be able to use SanDisk patents in its NAND applications.  Intel and Micron also pay Western for licenses on their 3D NAND product XPoint. In the coming years, Western plans to commercialize Storage Class Memory (SCM), a non-volatile memory that will be fast like DRAM but cost competitive with NAND.  This technology will provide customers a middle ground in the performance and cost trade-off they must make today.

Management

As mentioned above, Western management has been very adept at acquiring and integrating companies, evolving from a consumer HDD component supplier to a leading storage solution provider.  Management has also instilled a culture of innovation at Western. In the last several years, WDC began offering advanced technologies (the most important one being Helium Drives) in several of its HDD product lines that were widely questioned.  Again, WDC turned out to be visionary, and the entire industry scrambled to catch up. Today over 70% of WDC’s HDD enterprise exabytes shipped stem from Helium drives. In addition, as previously mentioned, Western possesses the most important storage and memory patent portfolio in the industry and continues to issue hundreds of new patents per year.  Operationally, management has also proven itself quite capable. In 2011, a flood in Thailand virtually destroyed WDC’s principal HDD manufacturing facilities. Experts and Seagate management opined that it would take WDC 12-18 months to ship again in volume. The company stunned the Street and was shipping material volumes in four months and fully recovered in less than twelve months.  Earlier this year, the Flash JVs were thought to be behind Samsung in the transition of manufacturing facilities to 3D NAND, the next generation in NAND technology. WDC management insisted that the transition was going well and that Samsung needed to move to 3D first since it was less competitive at 2D. Western management was indeed correct, and now the Flash Ventures JV is believed to be on par if not more competitive.  WDC will be producing 75% of its NAND on 3D by yearend.

Risks

The following are the principal risks to our thesis, many of which we have already discussed above.  These are generally in order of our perceived significance; the list is not intended to be comprehensive.

Recession / Cyclicality

A significant global recession would cause demand for storage solutions to decline while supply, especially of NAND products, could continue to grow.  It remains to be seen how much cyclicality will be in the larger secular trend of increased data storage. We strongly believe that end markets, especially for WDC’s mission critical, high margin drives will be much less cyclical than previous PC-driven cycles.  We do not dismiss this risk though, and our downside case contemplates a sizeable recession in which customers slow purchases of consumer products and hyperscale storage. While we think the stock is already priced for a very large cyclical downcycle, we do worry about the short-term pricing of NAND, which could depress sentiment further even if it is inconsequential to Western’s long-term outlook.  In the broader tech sector, we also are in what we believe to be the late innings of a strong upcycle. Pervasive commentary that essentially can be distilled down to “this time it’s different” also makes us a bit queasy. Unlike many other tech stocks though, Western is clearly not priced for a continuation of the cycle.

HDD Secular Decline Accelerates

HDD still represents a significant profit center for Western, earning $56 per share.  As comfortable as we may be with Western’s outlook, if HDD secular decline accelerates at a much faster pace than we and others anticipate, results could suffer.  Data centers are the only real growth end market for HDD. HDD revenue has undoubtedly benefitted from the tight NAND market in the form of a slowing of SSDs replacing hard drives.  If SSD replacement of HDDs resumes at faster pace due to lower NAND pricing, HDD results could underperform our expectations. Furthermore, NAND price declines could hurt pricing for SSDs at the same time the HDD business gets hit.  Also, WDC does generate revenue from selling pure NAND products such as those that go into iPhones. The company is likely over-earning here as NAND prices have spiked. Yet, pricing declines only drive more demand for NAND memory which drives more demand for other value-added products like cloud storage.  Also, WDC has a natural hedge for these risks in that HDD declines can only come about by faster adoption of SSDs. WDC’s stock price reflects irrational concerns regarding NAND pricing, HDD secular declines, and the Toshiba/SanDisk dispute. Thus, almost all risks we can envision appear largely discounted.  

Mobile Phone Leverage

A large portion of SSD end-market demand comes from smartphones (some estimate 45%).  If this end market corrects, WDC’s results could be negatively impacted. Chinese-branded smartphone demand, at 17% of total NAND capacity shipped (45% of smartphone capacity), has recently hit a soft patch, declining 16% in October and November of last year.  However, the biggest driver of consumer product NAND growth going forward is non-phone IoT devices.

Solutions Business Fails to Ramp

Most of Western’s TAM growth outlook comes from emerging data center solutions.  This is a relatively new business that may require additional acquisitions and/or further technology development.  We see execution risk as the primary concern with this opportunity. Mitigating this risk is the fact that it is already a multi-hundred million dollar revenue business, so Western is not starting from scratch.  The company also has a strong go-to-market track record for new products/solutions. Still, if Western doesn’t execute, growth will be underwhelming. However, the company would still be reasonably valued in that case based on the current share price.  For conservatism, we have built no Solutions growth into our Base Case below, and as such, the large opportunity is all upside.

Price Wars

The SSD and HDD markets are highly consolidated.  The resulting oligopolistic market structure should provide for some price stickiness and practical coordination among the competitors such that returns on capital can remain reasonable and attractive.  If the market participants choose to be overly competitive, pricing and earnings could suffer. WDC also has many levers to pull to support strong margins and profitability such as additional synergies from its mergers, interest savings, new products, and capital allocation.  We are quite confident that WDC will be perceived as a leading storage solutions provider with solid growth prospects and strong profitability over the next 18 months.

JV Conflict Flares Up Again

WDC derives the bulk of its NAND supply from its JV with Toshiba.  It seems very unlikely that this supply becomes cut off after the recent settlement announcement.  Still, if personalities or culture get in the way, irrational outcomes could occur. Without access to NAND supply, WDC would be forced to purchase NAND supply in the worldwide market.  This would undoubtedly cause significant pressure on gross margins.

Technology Shift

If an emerging storage technology takes hold, demand could rapidly decline for NAND-based products.  The majority of capital being deployed for emerging products is for a technology jointly developed by Micron and Intel called 3D XPoint.  It is at least a decade away from representing any threat to NAND. This technology is in its infancy and is primarily being utilized for new markets such as Artificial Intelligence.  It is very expensive and represents a greater threat to DRAM, but even that is minimal for five years. 3D Xpoint’s cost is estimated by Gartner at $3-5 per GB versus NAND’s cost in the $0.25 per GB range.

Agency Risk

We think Western management runs a good business, but its record on acquisitions and capital allocation is mixed.  The Sandisk deal was highly strategic and done near the bottom of the NAND cycle. However, we wonder whether long-term investors are better off on a per share basis given the dilution and expensive debt required to complete the transaction.  Seagate has vastly better total returns over the last decade. Here is a sampling of Western’s underperformance over various periods:

1-yr: -100%

3-yr:  -31%

5-yr: -17%

10-yr: -59%

As mentioned above, we have been frustrated with lots of buyback announcements but not much actual buyback.  This is one data point of many in the category of frustrating messaging. In contrast, If STX Chairman Steve Luczo says he is going to buy shares, you can take it to the bank.  Western CEO Steve Milligan also only owns $20 million of WDC stock despite being paid $30 million in 2017. This compares to the other Steve at STX holding $100 million worth of stock and being paid $3 million last year.

To us, Western still seems more focused on finding its next acquisition than closing any perceived valuation gap.  We are glad Western management has recently talked about the valuation disconnect in the stock and hope that the repo becomes much less dainty.  Management’s compensation is highly geared toward stock performance, so that should help.

Catalyst

Catalysts

Analyst day on December 4th  

Share repurchases

M&A

NAND prices stop falling 10% per month