Why Everspin Technologies Is Fading And Will Continue To Downtrend

Summary

• In October 2016, Everspin (MRAM) IPOed at $8 per share and traded there until a Barron's article in May caused a strong rally. It is now fading back to $8.
• Everspin's MRAM is too expensive to compete with DRAM, as DRAM will continue to get cheaper and better, and Moore's Law will continue for many years to come.
• Everspin has overly high analyst estimates, and management has hinted that it will miss guidance.
• Small companies like Everspin do not last in the memory chip business, economies of scale are essential for profitability, and system designers are reluctant to purchase from a small, and sole, vendor.
• At $5M per quarter cash burn, Everspin will need to do an equity raise sometime next year.

Even with the recent pullback, we recommend selling or shorting Everspin (NASDAQ:MRAM) as a long term short play. MRAM is a previous hot stock that is currently in the cooling down phase. We believe that many of MRAM’s shareholders don’t know the business very well and are holding because the company sells memory chips, which is a hot industry. We have spoken with several semiconductor experts to help us understand the industry and give a balanced analysis on the company. The reality is the memory chip product Everspin sells, MRAM, is too expensive compared to the competing DRAM chip, and this tiny company doesn’t sell in the bulk required to compete with behemoths like Micron (MU), Samsung (OTCPK:SSNLF), and Hynix (000660.KS) which all sell DRAM.

We believe MRAM will fall to below $10 within six months as the company begins missing earnings expectations and starts lowering guidance. The company's next earnings report is on November 13th. If the company doesn't submit unfavorable guidance for Q417, then we believe it's likely they will the next quarter. This will make many new investors that bought on the recent uptrend sell their shares, and the price will return to the $8-10 range it was before the uptrend started.

Long term, the stock could go to zero if MRAM never goes mainstream as the company is currently burning about $5M in cash per quarter. Analysts have overly high expectations for the company’s revenues and earnings in the upcoming quarters. Management has alluded to future events that will cause it to miss estimates. It said some customers won't purchase their next 256Mb memory chip coming out in early 2018 and will wait for the next one, the 1 Gb chip which will come out in late 2018/early 2019. We spoke with its investor relations representative who also suggested the company may miss estimates. The bearish thesis in a nutshell is: in the memory chip space, a lower price is king, not favorable architecture.

Everspin’s Recent Uptrend Was From Hype, It Is Now Trending Back Down

Everspin had its IPO in October 2016, and its stock stayed in a tight range between $6 and $10 up until late May 2017. The catalyst that made the stock break to the upside was a Barron’s article on Everspin and Impinj (PI). The following chart shows the stock movements since its IPO:

At the time of the red square in the chart above is when the Barron’s article was published at the end of May. The Barron’s article catapulted MRAM from $10 into the high teens. MRAM stock reached a high price of over $24 before it started a strong fade.

The Barron’s article had a very strong influence on MRAM’s share price, yet it doesn’t go into detail about the bullish case. It was just a couple of paragraphs that started the strong rally. What it does say is:

“Matthew Ramsay of Canaccord Genuity says the stock is currently worth 14 times projected earnings of 85 cents per share in 2018, or $12, above a recent $9.90. But it could have a multiple of four or five times sales if its newer markets take off, he says, which would put the stock price well above $12.”

However, from statements by Everspin’s management in earnings calls, its new 1 Gb product won’t be for sale until late 2018, and some of its customers are holding out for that product instead of buying its 256 Mb product coming out in the beginning of 2018. This is a warning by the company that its MRAM product that many customers are waiting for won’t be available until the end of the year. Revenue guidance will be affected by this. If MRAM is unprofitable in 2018, then the stock should be trading for well below $12 today based on Mr. Ramsay’s valuation method of 14 times projected 2018 earnings.

As stated in Everspin's Q2 2017 earnings call transcript:

“With the 1Gb chip sampling now, we believe, some customers may be reevaluating the product plans and may consider using the 1Gb device in their initial products instead of the 256Mb. This reevaluation of which density chip to use could result in our revenue ramp of our Spin Torque chips taking longer than we previously expected.”

From a Craig-Hallum report on August 10, 2017:

"One customer with which MRAM had a 256Mb design win has changed the design to use a 1Gb device, which will push out adoption by 3 quarters.”

Sellside firm Craig-Hallum is bullish on Everspin and regularly speaks with its management.

Right now, Everspin only sells its Toggle MRAM products, which are low density 16Mb and 32Mb chips. But since this business is only growing a little every year and is getting more and more outdated as competing chips improve, the company is advancing to higher density chips. It expects to be in production of its 256Mb Torque in early 2018. In March 2017, it was sampling its 256Mb products, and it usually takes 12-24 months after the sampling stage for a chip product to hit the market, from speaking to semiconductor industry insiders. The company's investor relations representative, Dave Allen, told us:

In March 2017, they had a production ready 256 Mb chip. That means they had a chip that meets the specs and has a yield target they need to have a successful product. The company announced then that its 256 product is ready for customers to sample and potentially use. It usually takes 9-12 months after that point where a customer can test it and build it into their road map.

After a customer finishes testing the chip, it then takes the company a little more time to finally mass produce it to sell. But if many customers want to wait for the 1 Gb chip Torque to come out and not buy the 256Mb chip, then that will further delay those revenues into late 2018 or early 2019. This will affect guidance and could cause the company to miss analyst expectations.

Everspin announced the sampling of its 1-Gigabit MRAM product on August 7, 2017, so that won’t be selling until late 2018. Also, Everspin’s new products might come out later than expected. The company has a history of delays. This article from August 2016 stated that Everspin will have its 256Mb Torque chips on the market in early 2017. So far, it hasn’t sold any Spin Torque chips, according to Mr. Allen.

Our bear case was strengthened from our conversation with Mr. Allen. In the interview, when we asked if some shareholders are going to be spooked if Everspin discontinues strong growth quarter over quarter, he replied:

That’s why we try to help educate the shareholders and for those who don’t pay attention, that’s unfortunate, but that’s all we can do.

And:

It’s not so much if MRAM will be adopted, but when.

We get from Allen’s above statements that he doesn’t know when MRAM will reach mass adoption, but he’s confident it will eventually. However, we don’t believe that all shareholders have long-term patience. Some are in it because of the hot memory chip sector, recent upward momentum in the stock, and the Barron’s article.

We also asked Allen how analysts got their high growth revenue and earnings estimates. He replied:

Let’s go back and revisit how analysts build their models. The company went public in October 2016. The company was talking to the analysts and bankers in the Spring/Summer of 2016. And their initial models have insight into what the company was thinking one and a half years ago. They have some insights that you wouldn’t get today as a public company because in the education process in the pre-IPO, trying to get a company public, they share forecasts.

If you read between the lines of what Allen said in this quote “trying to get the company public”, it suggests that Everspin may have presented overly optimistic, high expectations in order to have the bankers go along with the IPO.

If Everspin underperforms analyst expectations in the upcoming quarters, then the shareholders who don’t have the patience to wait for the company’s long-term vision will likely sell their shares. Everspin will have to raise money next year. It would take a shareholder with a strong belief in the company to want to hold through lowered guidance and a dilution before the company has proved that MRAM is a viable mainstream product.

Everspin’s History

Freescale Semiconductor invented the MRAM. Freescale didn’t want to compete in the commodity memory market and bleed money into it, so it spun off its MRAM department to create Everspin in 2008. Freescale was a spinoff of Motorola. That makes Everspin a spin-off of a spin-off.

The company raised $40M at $8 per share in its IPO in October 2016. The IPO was to raise money to bring its next generation MRAM Spin Torque product to market in a commercial way. But at $5M cash burn per quarter, the company has already burned through almost half of its cash from the IPO and hasn’t brought its Torque product to market yet. Everspin has also taken on an additional $4M in debt from its credit facility.

Moore’s Law And The Continued Improvement Of DRAM

We interviewed several memory experts to understand Everspin’s business. One informative interview was with Jim Handy of Objective Analysis, a semiconductor market research company.

Handy has over 30 years experience in the electronics industry, including 14 years as an industry analyst for Dataquest (now Gartner) and Semico Research. Investors can contact him from the website to get more information on the industry.

I sent Handy a slide from Everspin’s investor presentation and had some questions about it. Right away, Handy had good things to say about Everspin. He said:

Let me preface this by saying Everspin is a very credible company. They do a good job, they’re honest people, I’m familiar with them since their Motorola days.

When asked about Everspin’s MRAM product, he replied:

It’s a good technology. There are a lot of people placing bets on MRAM eventually replacing DRAM. The problem it has faced over the past several years, is people are waiting for DRAM to stop scaling, meaning stop getting its price to go down anymore. They are waiting for Moore’s law to end.”

What Handy is referring to with Moore’s Law is an observation made by Intel co-founder Gordon Moore in 1965. Moore noticed that the number of transistors per square inch on integrated circuits had doubled every year since their invention. Moore’s Law predicts that this trend will continue into the foreseeable future, and it has since 1965, albeit sometimes a little slower than doubling every year.

On Moore’s Law, Mr. Handy said:

What people have been nervous about, since the 70s, is that Moore’s Law will come to a grinding halt. It never does. And DRAM is one of those technologies where people say: “Oh yes, the end is just around the corner.” And they’ve been saying that for years and years. But right now, MRAM is much more expensive than DRAM. So right now it only goes into niche applications.

What applications does MRAM go into? We asked Handy. He replied:

With DRAM, any data you put into it goes away when the power is turned off, unless you prop it up with a battery or a capacitor. With MRAM, any data stuck in it will be there when it wakes back up.

For the most part, MRAM parts go into industrial systems. Where if the factory floor has robot arms swinging around that could crash into each other, where if there’s a power failure, when the power comes back on again, the robot’s controllers aren’t going to be confused about where they are, and make these robots crash into each other.

MRAM is also used more with automatic teller machines, or gaming machines. The last thing you want in a casino is for a one-armed bandit to lose track of where it was in the payout, and do two jackpots in a row just because the power went out.

These guys (Everspin) don’t like to talk about how much more it costs. Let me give you an idea of where these guys are. Short term, they have a steady business, long term, it’s going to grow modestly, like 20% per year, something like that. And for the long term, if DRAM stops scaling, if the price stops going down like in Moore’s Law, then these guys have a huge opportunity because the DRAM market is almost $40 billion this year. If they can take over that then wow! But that’s unlikely to happen for another 10 years or so.”

I told Handy that it's our opinion that Everspin won’t be able to last 10 years without going mainstream. The company currently burns $5 million per quarter. With only 20% annual growth, it would take a lot of financing to keep the company afloat for another 10 years.

Analyzing Everspin’s Bullish Thesis

The bullish thesis of Everspin claims that MRAM is a better memory chip than DRAM. This might be true, but the higher price of MRAM doesn’t make up for its advantages over DRAM in most industries.

Everspin’s mainstream pitch is a customer can either buy MRAM by itself or buy DRAM + Capacitors + Batteries. The main advantage of MRAM is there is no need for capacitors or batteries, especially expensive batteries, whereas DRAM requires them to function.

The following is a slide from Everspin’s investor presentation. Click for the investor slides link here.

This is the only slide in the investor presentation that shows cost savings with MRAM. The reason why, according to Handy is:

These guys don’t like to talk about how much it costs. It is more expensive but I couldn’t tell you how much more expensive. I know that it’s a lot.

The slide above mentions there are no external batteries with MRAM but doesn’t show the price savings of not needing external batteries. We estimate what that cost is in a cost comparison below.

Mr. Handy said:

MRAM is used in places where they are willing to pay several times as much as DRAM and a battery to get rid of that awful, unreliable battery that they don’t want in there.

Also, the slide mentions needing less NAND when using MRAM. Those savings are very miniscule. The investor relations representative Mr. Allen said that those savings are just “gravy” and not the main source of savings with MRAM. Not needing a battery is the main source of savings.

A Direct Cost Comparison Between 10Gb Of MRAM Versus DRAM

Everspin’s 256Mb Spin Torque MRAM chip is expected to be on the market in early 2018. We found the price is about $15-20 per chip from a Craig-Hallum report distributed on May 30, 2017. That report says (emphasis ours):

Everspin helped investors understand the opportunity that is possible from larger customers for its 256Mb device. MRAM indicated that flash array customer may have as much as $1800 value per chassis (5-9 devices per SSD, 24 SSDs per chassis, $15-20 ASP per SSD), and that customers generally expect to ship at least 1000 chassis per quarter, or more depending on market.

The above quote says that the ASP of a 256Mb MRAM chip is expected to be $15-20. Everspin’s denser 1Gb chip, which is expected to be on the market in late 2018, will be even more expensive than its 256Mb chip. To be conservative, we’ll assume the price for a 1Gb chip will be the higher end of the range, at $20 per Gb.

From dramexchange.com, the DRAM spot price is $1.80 for a 2Gb chip as shown below:

We can estimate that a 1 Gb DRAM chip is about $1. By late 2018, DRAM’s 1 Gb chip will also likely be cheaper than it is now, but we’ll keep it $1 per Gb for this comparison.

At $20 per Gb of MRAM, that’s a whopping 20x the price of DRAM. If not in volume, the price would be more expensive, but to be conservative, we’ll use the $20 per 1Gb MRAM price.

Everspin doesn’t quote the price of batteries, they just say that the lack of a need for batteries is the biggest cost savings for MRAM users. We couldn’t find a consistent price of batteries online for the specific use with DRAM. But one analyst we spoke with said the combinations of batteries and capacitors can cost up to $60 to support 10x1Gb. We are using that as a ballpark number for this comparison.

Using the above numbers, the following is a price comparison of 10 Gb of DRAM to 10 Gb of MRAM:

As shown above, MRAM doesn’t need capacitors or batteries, so the cost of those is zero for Everspin customers. However, DRAM is so much cheaper than MRAM, 5% of the cost or even cheaper, that it’s still much more cost efficient for companies to use DRAM despite MRAM’s advantages. Our comparison shows a memory chip customer will pay roughly $70 total for 10Gb of DRAM + Capacitors + Batteries versus roughly $200 for 10Gb of the MRAM solution.

Furthermore, DRAM, capacitors, batteries, and NAND are all getting cheaper over time as the technology improves. Everspin will continually fight an uphill battle against this. By the time its 1Gb MRAM chip will be on the market in late 2018 or early 2019, then 2Gb DRAM might only cost $1 apiece, and 1Gb will be cheaper than $1. In two years, 4Gb DRAM may only cost $1. The speed of technology advancement usually makes chips better, denser, and cheaper every year. This is stated by Moore’s Law.

The following are some differences between DRAM, MRAM, and SRAM.

Source: MRAM vs SRAM vs DRAM-Difference between MRAM, SRAM and DRAM

Also, from RF Wireless World, the following points are comparisons between MRAM and DRAM:

• MRAM performs faster read/write compared to DRAM which requires capacitor charge/discharge.
• Both MRAM and DRAM feature similar densities.
• MRAM is non-volatile unlike DRAM
• MRAM has lower cell leakage.
• MRAM has lower voltage requirements compared to DRAM, which refreshes its data often.

Everspin’s MRAM pitch also suggests that it has better reliability and increases space over the DRAM solution. From the Q117 earnings call:

“Our Spin Torque MRAM is attractive in these systems because it can increase usable space by eliminating Supercaps.”

And

“With our non-volatile MRAM attributes, this customer now has a battery free solution which improves reliability, system uptime and total cost of ownership.“

However, price is much more important to most customers than these relatively minor architectural advantages.

History Will Repeat Itself: A Capstone Turbine And Everspin Comparison

Right before White Diamond Research began, about 3.5 years ago, its founder Adam Gefvert, CFA, wrote two bearish articles in 2014, here and here, on a hot energy stock at the time, Capstone Turbine (CPST). Since then, the stock has collapsed by 98%, as shown in its chart below.

Source: finviz.com

The dynamics that caused Capstone Turbine’s rise, followed by its epic decline, are very similar to what is currently happening with Everspin today. In a couple years, MRAM will likely have a similar chart as CPST above. Capstone Turbine continually loses money quarter after quarter, and Everspin will also likely never be profitable for similar reasons.

The bullish argument for Capstone Turbine was its microturbines have advantages of lower emissions and lower maintenance costs over competing generators. However, they are more expensive than natural gas/propane generators and diesel gas generators by 5-10x. As evidenced by CPST’s share price fall, the price of the generators is more important to customers than its advantages. The big, successful generator manufacturers like Honeywell (HON), General Electric (GE), or Schlumberger (SLB) don’t sell microturbines. They sell the cheaper natural gas generators.

Similar to the uneconomical dynamics of Capstone Turbine, Everspin’s upcoming new 256Mb and 1 Gb MRAM chip will be a whopping 20x+ more expensive than the competing DRAM chip. Big chip makers like Micron, Samsung, or Hynix don’t sell MRAM. They sell DRAM. The only company that sells MRAM is Everspin.

Chip Manufacturers Must Sell In Bulk To Be Profitable

DRAM is sold in bulk by technology behemoths such as Samsung and Micron. They don’t sell MRAM and aren’t interested in doing so. DRAM has gotten so efficient to produce and is so widely used, that it has become a commodity like oil or copper. The DRAM market index dramexchange.com shows the current spot prices of DRAM, NAND and SSD.

Small memory companies like Everspin that are profitable don’t exist. It’s a commoditized industry with razor thin margins. They need economies of scale to be profitable. From speaking with experts in the industry, the way selling memory chips like MRAM works is the company pitches procurement specialists. These procurement specialists look for the lowest priced memory chip. They usually aren’t looking for architectural advantages.

Microchips is a capital intensive business. They aren’t cheap to make, and it’s necessary to keep upgrading them to keep up with fast-improving technology. If Everspin stops its R&D, it will fall behind its competition even more.

It’s also a disadvantage that Everspin is the only vendor that sells MRAM. Most data center and server companies aren’t going to buy a chip that comes from a sole vendor. Server designers need more than one vendor because once they design a certain chip in their system, the vendor needs to be around for the next 2-3 years. The engineering team isn’t going to risk it to buy from a small business. That small business can also have a production malfunction, which would mess up the server or cause delays. Everspin only has a $160M market cap and is burning cash, so there’s a small risk that it will go bankrupt in two years. Whereas an engineering team knows that a company like Micron is here to stay and is reliable. It wouldn’t be prudent for a system designer to replace Micron with Everspin.

Why Everspin Will Likely Miss Expectations In Its Next Two Quarters

Everspin had its IPO on October 7, 2016, and was priced at $8 per share. The following table shows its reported quarterly revenues and earnings and the expectations of each since the IPO.

Source: Bloomberg

In our opinion, these estimates are high because Everspin had to promise analysts big numbers in order for them to successfully market its IPO. Now that it has been a year since the IPO, these estimates will likely need to be adjusted downwards. The company has always guided for the following quarter in its earnings calls.

On August 10, Everspin reported Q2 2017 EPS of (40c), barely missing the consensus estimate of (38c) on $8.9M of revenue, compared to analyst expectations of $8.75M. This earnings result was in line with expectations. This isn’t a good result as the stock had closed the previous day at $17.98, about 80% higher than the price on the Q1 earnings report. A large beat should have been expected.

Mr. Ramsay of Canaccord dramatically raised his price target to $18 from $12 on August 14^th after the reported Q217 results. The analyst cited that “the results are better than expected revenues and gross margins despite what the company called a transitional quarter.”

It doesn’t make any sense for Ramsay to raise his price target on the earnings result because they were already guided by the company. As stated in its Q117 earnings call, Jeffrey Winzeler, Everspin’s CFO stated:

“Looking ahead to the second quarter of 2017, we expect revenue to range between $8.6 million and $8.9 million. We expect the resulting GAAP loss per share will range between a loss of $0.39 and $0.37 per share based on average weighted shares of 12,378,000.”

For the stock to go up 80% (from $10 to $18) and not beat its guidance should result in a selloff of the stock. The company isn’t experiencing the growth that its share price increase suggests it should have.

Investors and analysts haven’t downgraded Everspin on this past earnings. However, the next earnings report is much more important. If Everspin isn’t in line to beating expectations or lowers guidance, then a sell-off will likely commence. Everspin will likely make earnings expectations, because it has already guided it, but its Q417 guidance will likely be less than analysts’ estimates.

Sell-side analysts have forecast Everspin doubling its revenues next year. It’s certainly not on track to do that, and we don’t believe it will make it. The company hasn’t shown any evidence that suggests sales will be that spectacular.

Conclusion

Everspin is in the hot chip space sector and caught a strong rally that was kickstarted by a Barron’s article in late May. This has created a solid long-term shorting opportunity as we believe analysts and shareholders don’t fully understand the business. MRAM cannot compete with DRAM because of its lower density and higher price. Chip density and price are the two factors that are important to memory chip customers. MRAM’s system advantages don’t make up for its much higher price of around 20x.

Small chip makers like Everspin don’t survive because economies of scale is very important. Furthermore, a system engineer is hesitant to buy from a small, lone chip vendor because of the risk that something could happen to it down the road, and there's no other vendor as a backup.

As reality sets in, and as Everspin misses estimates and/or guidance, the stock will take a big fall, and we believe will return to single digits within the next six months. In the long term, Everspin stock will likely fall 90%+ as its chip technology continually falls behind the big chip players like Samsung and Micron, and Everspin continues burning millions of dollars every quarter for years to come.

Editor's Note: This article covers one or more microcap stocks. Please be aware of the risks associated with these stocks.

Disclosure: I am/we are short MRAM. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.