Micron's Transformation Demands A Valuation Rerating

Apr. 06, 2021 2:15 AM ETMicron Technology, Inc. (MU)537 Comments

3.21K Followers

Summary

In the second quarter F’21 earnings just released, management offered a positive and measured outlook, yet underneath lay a terabyte of optimism.
Their guide for the next quarter (May ‘21) offered a profit margin that was raised substantially; yet prices for digital memory have only just begun an ascent!
The overall outlook depicts a steady rise in DRAM demand emanating from a diverse range of industries, thereby reducing industry-specific risk.
Coupled with a disciplined oligopoly of three names, DRAM supply is likely to fall short of demand for the next three years. The result: a steady rise in DRAM pricing.
A market rerating is now due, to reflect the shedding of its cyclical stripes to a secular growth story in the transformed landscape of digital memory. Strong Buy.

Synopsis

This article outlines Micron's (NASDAQ:MU) transformation over two decades. It has survived the boom-to-bust cycles in the memory industry that has left a mere handful of players standing. Micron proved it had shed some of its cyclical stripes by remaining profitable in every single quarter during the 2019 downturn. This went halfway to prove Micron is a member of a lucrative Oligopoly in both the DRAM and NAND markets.

Now Micron is about to demonstrate its earnings leverage. Fuelled by a sustainable rise in demand for digital memory derived from a wider range of end-markets, while supply expansion will remain muted, Micron's earnings will exceed current street estimates. It might take 12 weeks or 12 months for the market to rerate Micron's valuation from a deep cyclical to an oligopolist in an industry enjoying a secular rise in demand. But rerate Micron it shall.

Q2 Fiscal '21 Results

Last week, Sanjay Mehrotra, President and Chief Executive Officer opened the Micron Earnings Call for Q2 F21 thus:

Micron delivered strong FQ2 results above our original projections driven by solid execution and higher-than-expected demand across multiple end markets. The DRAM market is in severe shortage, and the NAND market is showing signs of stabilization in the near term. The execution from the Micron team and these strengthened conditions enabled us to set revenue records…

Later, the CEO’s outlook summary also reflected Micron’s certainty for this fiscal year (my italics), but his conclusion is restrained, highlighting the risks and caveats,

Now turning to our outlook. DRAM prices have started to strengthen, and we expect the market to remain undersupplied this calendar year. In addition, NAND conditions are stabilising. These improving market conditions, combined with our significantly stronger competitive position, set us up to generate stellar financial results in the second half of the fiscal and calendar year.

While demand is strong across both the DRAM and NAND markets, our supply is now constrained as our inventories are very lean, particularly in DRAM. This restricts our ability to serve potential upside to demand. On the cost side, we are facing additional headwinds due to foreign exchange rates and drought mitigation impacting our Taiwan operations, and as result our FQ3 DRAM costs could be sequentially up. We are also assuming that there is no impact to our production output due to the Taiwan drought.

A careful reader will note the near-term optimism (‘stellar financial results in the second half of the fiscal and calendar year’). Micron has great visibility of sales for the rest of ‘21; these are largely signed contractual commitments that encompass higher prices – hence the raised guidance for the next quarter and near-term optimism.

Yet one cannot but notice the prudence beyond, for the second paragraph makes a sharp shift, to a focus on risks and caveats.

First, I must stress that I concur with Micron’s caution. But in order to fully appreciate Micron’s prudence and the recommendation of this article (see title), one must first digest two decades of the industry’s history…

The tortuous tale - laced with twists and turns - is covered (along with corroborating references) in my article Micron: The New Paradigm Unfolds (dated August ’20), but here’s a summary:

Two Decades of Feast to Famine

The period (1997-2017) depicts an industry rife with innovation and a vicious struggle for survival by numerous suppliers, each vying to secure a place in the future. No doubt, the stellar growth in demand of both NAND and DRAM portended rich harvests for survivors of the digital gunfight. However, the cost to play was a big R&D budget to remain "technologically current" in an industry evolving at a frenetic pace.

Quantum leaps in all aspects of memory design and manufacturing permitted massive cost improvements every year… the rate of technological progress was so rapid that if a supplier couldn't get its inventory out the door in a given quarter, it was virtually worthless in the next, because some other supplier that had kept pace with the digital beat offered a superior product at a lower price. As a result, the pace of innovation rendered inventory obsolete in a matter of weeks if it wasn't sold, leading to a dramatic culling and 'survival of the fittest.'

Natural selection was vicious: few survived. For NAND, since 2001, a dozen suppliers that represented one-third of the market have vanished, leaving just five players that represent 94% of supply (about $40 Bn pa of NAND memory). The consolidation in DRAM has been even more extreme, where three players have emerged to supply 97% of the global DRAM ($60bn pa), displacing a Motley crew who represented about 25% of the market in 2008.

Micron’s history is peppered with quarterly losses which coincided with an uptick in inventory. As seen in my blogpost Micron Inventory Days And Margins (quarterly data records since 2000), there was inevitably a collapse in profit margins (either in that quarter when it was marked down, or the subsequent one when it was disposed of at fire-sale prices) in any quarter when inventory days approached 100 days,

Using the parlance of an Economist, this era (1997-2017) was characterised by rapid downward shifts in the supply curve induced by innovation.

Supply Curve Stabilises into an Oligopoly of Disciplined Players

But then, another law began to emerge in the digital jungle: the law of diminishing productivity. In the early stages of a product lifecycle, the cost improvements were massive, but later, constraints in manufacturing, design, and even the natural laws of material physics led to smaller and less frequent improvements.

The memory players estimate the current rate of improvement in cost/gigabyte is 5% pa, whereas 40% quantum leaps (both in cost and storage density) were the norm a decade ago. Furthermore, the capital investment necessary for that marginal improvement has soared, as the complexity of squeezing yet another kilobyte of data into a nanometre of wafer that's already three stories high has deepened.

In both DRAM and NAND, there are now just a handful of players, finally in a position to capitalise on the enormous legacy investment, both in intellectual property and cutting-edge manufacturing capacity. Having survived the loss-ridden past, and given the enormous capital cost of a new plant, each player is far more cautious of embarking on a greenfield plant.

Longer Shelf Life of Inventory

Another critical issue is the longer life of inventory as we enter the age of a slower rate of marginal improvement.

The marked change in the shelf-life of inventory is evidenced in Micron’s results. Investors were highly concerned that the slump in demand in 2019 (covered in an article later) would lead to an inventory impairment and operating losses. (see May 2019 Transcript where the word ‘inventory’ features 44 times!) This showed investors were reluctant to acknowledge the corollary of a slower rate of cost improvements is a longer inventory shelf-life.

Yet actual results have borne this out. As seen in the graph below, Micron’s inventory exceeded 100 days in every quarter from November ’18 to November ’20, (Peak:145 days May ’19) and yet - in stark contrast to the past – the operating margin never went negative.

The Chinese Threat

A growing concern in the memory industry, especially given the rising acrimony in US-China trade relations, has been the threat of new Chinese competition. After all, China’s government has made an explicit objective to be more self-sufficient in semiconductors, as elaborated in the Digital Silk Road Initiative. One would be naive not to heed their ambition.

However, significant hurdles do exist. It would be very difficult for a new player to enter the digital memory industry. First global patents have been tested and upheld. Micron successfully sued a Taiwanese company and its Chinese affiliate for IP theft last year. It’s worth noting the semiconductor industry has amassed significant Intellectual Property over the decades, and as technology improved in steps, a current product/process may well encompass a succession of patents accumulated over time. Micron ranked 19th in the world for patents in 2020 with 1535, doubling from just three years ago.

Also, there is significant technical know-how in achieving satisfactory yields from a fab, garnered by experience over an evolving technology roadmap over two decades. An entrant is obliged to master one manufacturing node before transitioning to the next.

No doubt, Chinese players will enter the memory industry, but their presence will not be disruptive for the foreseeable future. To quote two examples, the Chinese leader in DRAM (unequivocal support from government) China’s ChangXin Memory Technology (CXMT) is shipping its first 19nm DRAM line, (whereas Micron is working on 5nm); in NAND, China’s Yangtze Memory Technologies (YMTC) recently entered the 3D NAND market with a 64-layer device, but has remained incapable of scaling production.

The Nature of Memory Supply Curves

There are a host of reasons why DRAM and NAND supply have become more predictable: naturally fewer players means greater scope to forecast industry supply, but there are others.

DRAM Market Share

DRAM MARKET SHARE

Source: Statista

NAND Market Share

NAND MARKET SHARE

Source: Statista

There is the huge fixed cost required for the construction of a fabrication cleanroom – only then can silicon wafers be produced. As a result, supply is expanded in a step function. But the giant capital commitment will not be made before there is sufficient demand.

Once built, the DRAM or NAND supplier will increase throughput by a steady rise in productivity from a wafer, by transitioning through nodes (DRAM) or increasing the layers (NAND) in a nanometre of wafer.

In Micron’s case, after the fab space built in F ’19, bit output has steadily risen as the company progressed on a pre-defined technology roadmap.

MICRON TECHNOLOGY ROADMAP

Source: Micron Q2 F21 Slides

There is also the issue of lead times. Even if orders were placed today, the capital equipment required (silicon wafer extrusion, lithographic/laser/UV cutting equipment, etc., produced by a niche of global suppliers) would only be delivered in a minimum of a year.

As a case in point, SK Hynix just last week announced it intends to build a giant new fab cluster where each fab facility will cost $25 Bn, but even though the project is about to begin, commercial production will only commence in 2025

South Korea authorities this week gave SK Hynix a green light to build a new, ($106.35 billion) fab complex. The fab cluster will be primarily used to build DRAM for PCs, mobile devices, and servers, using process technologies that rely on extreme ultraviolet lithography (EUV). Keeping in mind that we are dealing with EUV fabs, it is not surprising that a huge 200,000-Wafer/month plant with EUV tools will cost SK Hynix north of $25 billion. The first fab in the complex will go online in 2025.

The combination of the above has the end result of a more predictable supply curve. None of the remaining DRAM/NAND players would embark on the monumental cost of a new fab before a close look at the other suppliers’ expansion plans, and before a detailed appraisal of future demand.

This is precisely why Micron telegraphs its expansion plans to the industry.

Outlook: Micron and Industry

From the conference call slides last week:

Long-term DRAM & NAND Micron bit supply growth CAGR in line with industry demand growth CAGR

Long-term DRAM industry bit demand growth CAGR of mid-to-high teens

Long-term NAND bit demand growth CAGR of approximately 30%

Micron intends to maintain its market share, growing supply in line with demand. The message to the other suppliers has been clearly broadcast:

And you, you other suppliers take note, if you chose to upset the apple cart, you do so at our mutual peril.

A Reflection on 2018/2019

Having made the case for a marked improvement in the fortunes of the memory suppliers, the obvious question would be,

But then why did both DRAM and NAND prices collapse in 2019?

The feast-to-famine of 18/19 was caused by a shift in the demand curve. The supply curve has indeed shifted as described, but it was the demand curve.

In summary, enterprises felt an urgent need to duplicate their IT infrastructure in the cloud in 2018. As they witnessed a rise in memory prices (both DRAM and NAND), both data centers and enterprises rushed to pre-order. This only exacerbated the shortage and resulted in even higher prices in 2018!

Come 2019, enterprise cloud presence was largely completed, demand shrank and a supply glut followed. DRAM and NAND prices halved in the space of a year.

Source: extremetech.com

But, remember, this was 2018/2019. Establishing a presence in the cloud was the rage, and growth from enterprises and data centers dominated the memory industry. In Economics terms, overall demand did not change, but it was compressed in one year, leading to a fallow period in the next.

Here, again, investors naturally became wary of the volatility. Even when enterprise demand had normalised in 2020, fear remained. Here’s an excerpt from the article linked above; during the Q&A section of a Micron presentation in May 2020:

Analyst Question:

What about the volatility caused by data centers? The hyperscalers or the cloud service providers, these companies tend to be very lumpy in their purchases. They tend to have five quarters of huge strengths and five quarters of weakness. We saw 2019 was very, very weak and so far this year extremely strong…. the critical kind of thing that investors are trying to figure out how many more quarters have we got of this data center strength?

Micron CEO Response (my italics)

The COVID environment, it certainly does bring lower visibility and does bring some element of uncertainty in all industries. I would like to say that the kind of volatility that was experienced in 2018-2019 kind of timeframe....those cloud customers also would not want to have that kind of volatility, because it is better for them, as well as for us.

Cloud customers definitely have matured in terms of their buying patterns and understanding of the industry as well. And from our dialogue, we were closing with these customers, we are a major supplier to these customers, we worked closely to understand what their demand and product requirements are. And Micron is - they are valued partners to us, the cloud customers, and we are valued partners to them as well.

Micron’s Demand has diversified

Investors have yet to acknowledge that Micron’s demand has diversified materially. Now that the insane rush for the cloud has abated, Micron supplies a wider range of end markets.

This is confirmed by regarding the shift in segmental results.

One division, called, "Compute and Networking Business Unit ('CNBU'): Includes memory products sold into client, cloud server, enterprise, graphics, and networking markets.' This division includes cloud business, but it also encompasses home-computing (client) and gaming (graphics).

Nonetheless, the high dependence on 'CNBU' has reduced, whereas 'MBU' or the Mobile Business unit and Embedded Business Unit ('EBU') which includes automotive and industrial segments have increased.

	Revenue Mix by Segment

	Q2 F21	Q3 F18
CNBU	42.3%	51.4%
MBU	29.1%	22.5%
SBU	13.6%	14.6%
EBU	15.0%	11.5%

Source: SEC 10-Qs

Micron’s Future Earnings Will Beat Current Market Estimates

First here’s the guidance for next fiscal quarter (end May ’21)

FQ3-21	GAAP⁽¹⁾ Outlook	Non-GAAP⁽²⁾ Outlook

Revenue	$7.1 billion ± $200 million	$7.1 billion ± $200 million
Gross margin	40.5% ± 1%	41.5% ± 1%
Operating expenses	$930 million ± $25 million	$875 million ± $25 million
Interest (income) expense, net	$27 million	$25 million
Diluted earnings per share	$1.52 ± $0.07	$1.62 ± $0.07

The outstanding feature is the rise in profit margin to 41.5%, compared to the Q2 margin just released of 32.9% based on higher DRAM pricing. Also note the $1.62 for non-GAAP earnings is a material increase from the current Street consensus of $1.32, (see snapshot below date 2 April '21, although this will move higher shortly).

Regarding Micron’s EPS for Fiscal ’22 (Aug ’22), my earnings model delivers $11.2 per share, materially higher than current consensus of $9.33. Naturally, the critical input variable is DRAM ASPs over the fiscal ‘22. In my model, ASPs rise on average by 20% on average over the next 12 months. In simple terms, that equates to a straight-line 40% rise over the next year. See valuation section below for more detail on DRAM pricing assumed.

DRAM Demand, ASPs and Micron’s Future Earnings

One must note that Micron is not in the business of forecasting DRAM/NAND ASPs (Average Selling Prices) in the future, but only for one quarter as contracts already signed provide certainty.

As a result, an analyst is obliged to forecast future pricing to estimate Micron’s future earnings. In addition 90% of DRAM/NAND is transacted on a volume contract basis (in confidence between Micron and say Apple (AAPL) or Nvidia (NVDA)), whereas an analyst only has access to spot pricing as a proxy. Nonetheless, it’s a reliable proxy as affirmed by research firm DRAMeXchange here.

The graph below shows spot prices for a memory kit commonly used today in a desktop or gaming console.

rpisaw

Source: Author’s tracker camelcamel.com

As seen, DRAM spot prices have only just begun to rise from a trough that prevailed in the latter half of 2020; since December last year, prices have picked up from a bottom, and they have risen steadily since. It is this strength that fed into Micron’s raised guidance for the next quarter.

Although no analyst can provide certainty on future DRAM ASP’s (nor can Micron), it is my contention that the supply/demand imbalance in DRAM (71% of Micron’s revenue mix) is likely only to intensify. PC demand has been boosted by work-from-home trends; the advent of 5G phones will require up to 5X as much memory (DRAM and NAND) as their 4G predecessor; future self-driving cars (Level 3 – Level 5) will require new generations of memory with significantly increased bandwidths; even in data centers the rise of AI and ML that use graphic servers need up to 6X as much DRAM as a standard server.

end users

Source: Micron Q2 F21 Deck

Coupling the above demand drivers, the muted supply expansion, and the fact that DRAM ASP’s have only just risen from a deep trough, my base case is DRAM pricing will rise by 40% in the next 12 months. I envisage prices will then stabilise in a plateau for two successive years, a stark contrast to the deep cyclicality of the past. During that plateau, DRAM ASPs will equate to or even exceed the sharp 2018 peak.

Valuation

I have previously compared Micron’s valuation rating to Nvidia in an article The Absurdity Of Micron's Valuation Versus Nvidia (date August ’20). Since publication, Micron has indeed outperformed Nvidia, but that is not relevant here.

Although Nvidia makes GPUs (Graphic Processor Chips) and Micron makes DRAM/NAND, the products are complementary in that a GPU always needs memory. As seen in the revenue growth rates (YOY) below, Nvidia and Micron are exposed to similar forces over the long term. Note also, Nvidia registered higher revenue growth rates in the near term, as GPUs made inroads into Machine Learning (ML) and AI (Artificial Intelligence) applications in data centers whereas Micron faced the 2019 memory slump.

There is no doubt that Nvidia posted higher operating margins and hence merited a higher valuation ratio (EV/EBITDA being my favourite measure).

I propose Micron’s operating margin will rise over the next three years, closing (but not eliminating) the gap to Nvidia’s margin (see below). I also propose Micron’s revenue growth will be in the 20-25% over the next three years.

Valuation Table

TIME PERIOD	EBITDA (TTM)	EV/EBITDA RATIO	EV TARGET	NET DEBT	TARGET PRICE
	$MN		$MN	$MN	PER SHARE
NOW	9,787	15	146,805	- 1,086	127.38
IN 12 MONTHS	13,701	20	274,020	- 1,086	227.45

In light of the above improvements, Micron deserves a substantial rerating on its current EV/EBITDA ratio of 10.6 vs Nvidia’s of 59.4. Using a target EV/EBITDA ratio of 15 on current EBITDA (TTM) would align the ratio with Micron’s transformation out of losses in future troughs. This leads to a current price of $127 per Micron share.

In summary, for the next 12 months, I expect revenue growth of 35% as higher DRAM prices (+20% on average for the next 12 months, +15% volume growth =35%) filter through the profit statement to deliver earnings growth (EBITDA) of 40%. This offers an EBITDA in one year of $13,7 Bn. By then, I expect the market to begin to grant Micron a higher multiple as earnings will continue to surprise; the EV/EBITDA ratio hence rises to 20. (Remember NVDA today is 59.4). Using a share count that increases by 5% over the next year, my price target in one year is $227. Despite the huge run in the last year, Micron's re-rating still lies ahead. Strong Buy.

Investment Risk

It’s important to note that this article does NOT claim Micron will be immune to cycles, but earnings will be less cyclical in an upward trajectory, and the company will remain profitable in future troughs. The memory industry will always incur the ebb and flow of supply and demand. For example, after the bulk of 5G phones in the developed world have been bought, the new wave of phones in the developing world might not equate to the same memory despite higher unit volumes…and there might be a ‘gap’ before the next wave, say mass adoption of Full Self Driving (FSD) cars and navigation software. Yet the wave function of memory pricing will point upwards, in our current era where 'Data is the new Oil' as stated by the oil conglomerate Shell (RDS.A).

A risk to my recommendation is that the ‘gap’ in successive memory waves described above will become too wide and will lead to a valuation derating. Needless to say, forecasting that risk for that unforeseeable future is not possible.

This article was written by