Description

Introduction

Micron (MU) is a special-situation investment that incorporates a major transformative acquisition, a new joint venture and a potential paradigm shift in its major memory end markets: DRAM and NAND.  We think Micron has the potential to generate $2 of normalized earnings on a stand-alone basis and $3.50-$4 of earnings once the synergies from their acquisition of Elpida are fully realized.  We think the stock should trade at 10x normalized post-synergy earnings, or $35 - $40 a share, compared to current trading level of ~9.60, for a potential return in excess of 300%.  Even if we are too optimistic in our assumption of a paradigm shift in the memory market and our earnings estimates are cut in half, we are still buying a stock at 5-6x earnings.  Under this no-paradigm shift downside scenario, we still think this investment can generate returns of 75%-100% based on our current investment.

DRAM Market

Micron is a company that has never earned a reasonable return on its cost of capital.  In fact, the DRAM industry, in the aggregate, has rarely returned its cost of capital.  Twenty years ago, there were over 60 DRAM manufacturers.  These DRAM manufacturers, as well as the sovereign governments that supported many of them, fought aggressively for DRAM market share.  This created treacherous market conditions that, over time, forced many competitors out of business.  In turn, the Japanese, Korean and Taiwanese governments ensured the DRAM industry would be oversupplied for over two decades. However, due to the intense competition, many players exited the business, either as a strategic choice or because of bankruptcy, and the end result was the reduction of market participants to five by the end of 2011. These five remaining DRAM players were Samsung, Hynix, Micron, Elpida and Nanya.

Historically, DRAM was primarily used in computing, both personal and enterprise (server).    As recently as a few years ago, PCs represented 70-80% of revenues and bits sold in the DRAM market.  In the last couple of years, with the proliferation of tablets and smart phones, the DRAM landscape has changed materially, reducing the 2012 PC bits sold down to 40% of industry bits sold and, as a result of the higher price point for mobile DRAM, reduced PC DRAM revenues to 20% of the total DRAM market.  However, in 2012, as the market was expecting moderate growth in the PC market (and produced as such), commodity DRAM prices imploded.  This price correction forced Elpida to file for bankruptcy in Japan, while Inotera, a publicly listed DRAM producer controlled jointly by Micron and Nanya, ran into significant financial difficulties.

Nanya has been a perennial disaster of a business.  Massive negative cash flows have limited its ability to invest in research and development.  As a result, Nanya has been selling 50% of the DRAM produced by Inotera at negative margins and ASPs below industry average.  In January of this year, Nanya decided to exit the market, giving Micron access to 100% of the output.  The new Micron deal for the Inotera output is a fixed margin % (our best guess is 15%).  This is the equivalent of fixed 30% margins on Micron’s historical Inotera bits with the upside coming from their 40% ownership position in Inotera (which will come through in equity income).  The Micron stake in Inotera is currently worth $737mm, or roughly $.70 a share to Micron.  As a result of the Nanya exit, the DRAM market was reduced to four players in January of this year.

Micron agreed to act as the plan sponsor for the Elpida reorganization process in mid-2012.  As I will discuss later, the closing of the Elpida deal, is a matter of when, not if.  As a result, Elpida has become a rational market participant (Micron has agreed to backstop any capital funding gaps in the interim before close, but has not been required to do so at this point as we believe Elpida is now free cash flow positive).  Once this deal closes, the DRAM market will be reduced to three players.

Hynix has historically been a poor market participant.  A recent CEO change (CTO promoted) and our conversation with the company highlight the transformation in Hynix management focus from survival and market share to economic return.  The company is using EVA (Stern Stewart’s Economic Value Added framework), and is focused on generating returns on capital over the cycle in excess of their cost of capital.  This enlightened thinking encourages us that there is a “new normal” in the DRAM market, with Hynix emerging as a rational participant.  Hynix emphatically stated to us (as did Micron) that they will never add a new wafer start to the DRA M market.

Samsung is in an “arms” race with Apple for the high-end phone market.  One of Samsung’s competitive advantages is their memory vertical integration (both NAND and DRAM).  Samsung has doubled its DRAM per phone and now surpasses Apple’s DRAM per unit (2 Gb for Galaxy S4 vs. 1GB in the iPhone 5).  Samsung has every incentive in the world to keep memory prices elevated as higher prices for NAND and DRAM provide a substantial cost advantage relative to their non-vertically integrated peers.  In fact, as a result of this “arms” race, Samsung is using so much of its internal capacity for their own products that they intend to ship 50% less DRAM to third parties.  This has further tightened an already tight and consolidating market.

In the last few months commodity DRAM spot prices (2 GB) have increased from $0.82 to $1.77 (Bloomberg DRAM84 index) while contract prices have improved from a low of $0.80 to the current $1.44 (Bloomberg DRAM87 index).  Over time these prices converge, so it would be fair to assume that these prices meet in the middle at some point.  For all our analysis, we assume the current contract price of $1.44.  For frame of reference, the current spot and contract prices were last seen in mid-2011, and are far from multi year highs.  Our industry channel checks and most sell-side channel checks expect contract prices to rise to the $1.60 range in the back half of this year.  We would view that move, if sustainable, as upside to our numbers.

While demand patterns for DRAM can be volatile, the consolidation of the industry to three players is meaningful not only in that an oligopoly has been formed, but also because the remaining players all have the technology to convert excess DRAM capacity to higher margin NAND production.  Elpida, and Nanya (Inotera capacity) had not invested in NAND technology and were major industry participants.  As a result, as DRAM demand declined, they were forced to continue to sell their product into an oversupplied market, depressing industry margins.  In this new oligopoly, Micron is converting Elpida capacity to NAND and can convert continue converting this supply, to ensure that the DRAM supply side is balanced.  This behavior would be expected today, as there is substantial economic incentive to convert DRAM capacity to NAND production (1,000 bps of margin expansion). In addition, Samsung, Hynix and Micron are converting PC DRAM capacity to mobile capacity, given better demand end markets and higher ASPs.

NAND Market

The NAND market seems to be a couple years ahead of DRAM in hitting the Moore’s law conundrum, where the current technologies for shrinking memory bits onto smaller wafers has hit its technological limit.  As a result, for the next few years, as market participants develop new technologies to further expand NAND capacity, NAND capacity growth and industry capital expenditures will decline substantially.  This should provide existing players with the ability to realize enhanced margin and pricing leverage.   We will not spend too much time on the NAND market, as there is a significant amount of good sell-side research on the subject (see SNDK).  It is widely believed that the NAND market will be under-supplied for the next couple of years.

The NAND market has only four real players, Sandisk (with Toshiba), Micron (with Intel), Samsung and Hynix.  Although volatile in short spurts, this consolidated marketplace has allowed for meaningfully positive NAND economics historically.  On a stand-alone basis NAND represents approximately 40% of Micron revenues and, including Elpida, will represent roughly 25% of consolidated total sales.

Elpida Closing Process

In mid-2012 Micron agreed to purchase Elpida out of bankruptcy.  The purchase price resulted in a return to unsecured creditors of 13-17% of the par value of claims outstanding.  The deal has received bankruptcy court approval and Micron can close today, if it so chooses.  However, based on our understanding of Japanese bankruptcy, creditors have a right, if they opposed the bankruptcy plan, to appeal to an appellate court.  In the history ofJapan, no deal has been overturned as a result of an appeals process.   The process is designed to adjudicate matters of law, not fairness of transactions to creditors.  In fact, inJapan, social issues weigh heavily on the Court’s decision making process.  It is our understanding that Micron sealed the Elpida purchase by agreeing not to fire any Elpida employees inJapanfor three years.  Recently, two creditors filed appeals to the appellate court.  We understand the two creditors to be US-based investors (who care about recoveries on their claims).  These creditors are playing the “nuisance” game (i.e.,  pay us something to walk away so you can close the Elpida transaction today as opposed to waiting 3-6 months as the appeals are heard in Japan.)  We do not believe there is a substantive matter of law that needs to be adjudicated and believe their end game here is to get paid to go away and for a reasonable price, Micron is incentivized to do this.  We think this is the most likely outcome and expect the Elpida deal to close somewhere between May and August.  If the creditor “nuisance” asking price is too high, the appeals process will play out and the Elpida deal will close sometime between September and year end. 

Earnings - Standalone

We think Micron, on a stand-alone basis, will earn $.40-$.50  per quarter, starting in the August quarter.  This takes into account the current DRAM spot price of $1.44 and flat NAND prices.  It also incorporates $50mm of run-rate equity income from Inotera (Intotera reported a profitable month in  March in a Taiwanese regulatory filing).  As most of Micron’s DRAM and NAND sales are on contract, moves in spot/contract prices lag actual results by 4-6 months.  As a result, the most recent price increases in March and April won’t fully impact Micron operating results until the August quarter.  Many sell-side and buy- side analysts mistakenly look at pricing information that Micron provides in their earnings release as “guidance”.  In reality, this metric is just quarter-to-date pricing based on realized prices on contracts already signed months earlier.  Furthermore, this “guidance” is not impacted by the lag impact of contract pricing or by forward moves in contract or spot prices.  As a result, earnings this quarter and next will far surpass street expectations.  We anticipate earnings in the current quarter (excluding one-time items) will be $.20-$.30 vs. street expectations of roughly break-even results.  With an earnings run-rate of $1.60 - $2.00 pre-Elpida, and with Capex guidance at $1.6-1.9Bn vs. depreciation at $2.2Bn, free cash flow per share should normalize at $1.90 – $2.50.

We think Micron is materially undervalued on a stand-alone basis.  At 10x free cash flow, Micron stand-alone may be worth $19-$25 per share, or more than 100% upside from the current quote (why 10x?  No good reason.  We just assume that at 10x free cash flow, there will be further upside in the stock for others to enjoy). 

Note:  NOLs and future cash generation.  Our numbers assume nominal federal and state taxes as a result of close to $4Bn of federal and state NOLS.  If we are correct in our numbers, these tax attributes will be used up in 2-2.5 years.  We think our normalized earnings and free cash flows will still be maintained without these tax attributes as we assume no use of free cash flow generation which we believe will be used to reduce share count (and interest expense) through convertible bond repurchases. 

Elpida Accretion

There are five components to our Elpida accretion analysis:  the core business, purchase accounting adjustments, the move in the yen, cost synergies and DRAM to NAND margin enhancements (offset by increased interest expense).  The aggregate of these adjustments will accrete to Micron stand-alone earnings by approximately $1.50 per share (untaxed).   At 10x earnings, the Elpida deal could be worth $15/share to Micron (excluding the benefits to Micron core from incorporating that capacity into a rational entity).  Incorporating our stand-alone valuation from above, we believe that Micron including consolidated Elpida could be worth $34 - $40 a share in 6-18 months.

 Elpida is run-rating close to $1.2B of depreciation and amortization per year (per Micron’s 8-K pro forma financials).  As a result of the Elpida purchase price (bargain basement), Micron will write down the fixed assets to account for 1) the agreed upon transaction price, 2) a reduction in the actual price paid as a result of the rapid decline in the value of the Yen (purchase price was agreed to in Yen terms) and 3) the interim cash generation at Elpida that Micron will have access to once the deal closes.  As a result, we expect depreciation to decline by close to $700mm per year, or 65-70 cents per share. 

Our current estimates are that Elpida’s current operations, pre-depreciation adjustments 2) and 3), are $100-300mm EBIT positive, or $.09 - $.27 per Micron share. 

According to management, every 1 Yen move relative to the Dollar is worth $7-8mm per quarter in net income, or $28 – $32mm per year.  Since the Elpida deal was agreed to, the Yen has declined by roughly 20 Yen to the Dollar.  This results in incremental earnings of $480 -$640mm, or $.44-$.58 per share.

There are three components to our synergy estimates; 1) back-end testing, 2) SG&A and 3) R&D.  We have been told that back-end testing savings can be up to $400mm, but for conservatism, we are using $150mm.  Since Micron can’t reduce headcount inJapan(as a result of the sponsor agreement), we assume all SG&A and R&D savings will come from Micron expense reductions.  Our estimates are for total overhead cost cuts of $125mm, mostly from R&D.  Synergies are a further $.25-$.30 in earnings accretion.

We assume 10% of Elpida capacity will be converted to NAND at a 1,000 bps improvement to margins.  This will result in a further $40mm in earnings enhancement, or $.04 per share.

Interest expense, including the imputed cost of the future payouts to Elpida creditors, will be $275mm per year, or -.14 per share, before any additional converts are repurchased through FCF generation.

Conclusion

Despite the increase in Micron stock price from the beginning of this year, it is currently at similar levels to early 2012 (when investors were getting excited by the prospects of an Elpida consolidation).  As value investors, we hate buying stocks up, but when the opportunity exists for double, triple or quadruple returns on our initial investment, we hold our collective noses and dive in.  We think Micron is such an opportunity.

Catalyst