Everspin’s MRAM IPO

by Robin Harris on Monday, 10 October, 2016

Everspin has filed for their IPO. They’re looking to raise $40 million from the public market. They’ve been shipping product for over 10 years, so this is a real company, not a dream and a slide deck.

Why MRAM?
Everspin’s Magnetic RAM has a number of advantages over flash and DRAM:

  • DDR RAM write latency – much faster than 3D XPoint and much, much faster than flash.
  • Endurance that is much higher than flash and 3D XPoint.
  • Byte addressable, like DRAM and unlike flash.
  • Can replace DRAM on DIMMs – no need for complex controllers for wear leveling and garbage collection.
  • The latest MRAM gen is built on a simple 3-layer process, which means that it could, at high volumes, be cheaper than DRAM.

The logical question then is: why hasn’t MRAM killed DRAM? A couple of key reasons:

  • Cost. MRAM could be as cheap as DRAM, but only after climbing down the learning curve, where DRAM has a 50 year head start.
  • Density. Everspin will be sampling a 1 Gbit chip later this year, far behind the 8 Gbit DRAM chips commonly available today.

But even with a cost-per-bit 10x that of DRAM, MRAM has a defensible niche as it ups production and lowers cost. It replaces DRAM-based NVDIMMs because it doesn’t need the batteries and assorted packaging cruft those require. It also works nicely as a buffer, where speed, endurance and non-volatility are vital, and its cost is buried in a larger product.

The filing
Everspin’s S1 shows that the company is doing about $25 million a year, and spending about $10m a year on R&D. Despite gross margins in the 50% range, their investments in growing the business have them on track to lose about $18m this year.

The StorageMojo take
Given Wall Street’s warm welcome for Nutanix I expect that Everspin will have a successful IPO. It may finally be sinking in with the investment community that storage is the most valuable part of our IT infrastructure.

The larger picture is that the IPO demonstrates that NVRAM technologies are real, that 3D XPoint isn’t the only contestant, and that the pace of change in the storage market is still accelerating. These are all Good Things.

StorageMojo wishes Everspin the best of luck on their IPO.

Courteous comments welcome, of course.

{ 2 comments… read them below or add one }

John Other October 14, 2016 at 2:28 pm

Hope this brief tangent is okay:

It is high time we saw more IPOs where money goes to developing tech like this.

By like this, I mean a company that needs capital for a business that primarily develops a technology. Not a company that is a consumer of technology that claims its programs are technology per se.

Maybe I am just still reeling from a colleague talking about “Uberifying” as a adjective to some of the intended mechanism of the code he is writing. I mean I was shaken – I know he was reaching for a explanation of what he is trying to achieve (which I didn’t get already we got distracted by other needs to discuss today), but he honestly intended or hoped that I would infer from that use a specific meaning of process approximate to a level of technological invention. Shocked that my senior by some years, alum U.Ill. into the bargain, would be so sloppy, I knee jerked with a attack on Chicago style economics supporting Uber obviously being a way for auto manufacturers to stealthily stalk legislation changes to enable fast track driverless cars by another route, being to make people so mad in general at the whole politics of drivers that drivers for hire cars are outlawed, just so we don’t have to read about their fights or the billions spent on lobbying, any more…

Thanks for your blogs, always appreciated, Robin! The more so, the less time I can keep within any sight of your subjects.

I keep grinding my teeth that one of the things we have to do to get this economy moving (I’m a Brit and in London, but my links are far better connected to the US, also economically*) is to do whatever it takes to generate a real IPO market. London has forever been loosening rules. How come we put Russian mining giants with almost no free float stock, straight onto not only the LSE main list, but even the main FTSE index? And our AIM market may be organised, but it comes across like the pink sheets if they were run by well connected (socially rather than business) old school brits with plum accents. We do at least have a 3yrs of profits rule, for new main list entrants to the main market, but you can get off the shelf balance sheet stuffing bolt-ons and make that hurdle, if you have any kind of decent “accountant”.**

I know SOX is the most cited deterrent for listing in the US.

So why is Companies Act 2006 in the UK, not cited equally? It too makes hundreds of far more modest offences criminal and subject to imprisonment on summary trial. Moreover summary trial means not a jury trial, not even necessarily a senior judge. (And there are wide and longstanding complaints form within the judiciary about lack of specialist abilities and capacity to try cases.)

I think that that juxtaposition almost demands a comparative study, along that axis.

The UK I think suffers a much bigger headache, which is just never going to be dealt with because first it would have to be acknowledged: few tech companies on our markets have much in the way of technology. We get one ARM and think we are all every bit as worthy. Admittedly ARM is a brilliant story which many economies would greatly like to enjoy. But to take one company which is at least recognisable across the pond, Autonomy, well the story pans out about the way I think all the tech IPO stories in the UK pan out, when I start looking. It begins with great interest, and ends with me wanting to proverbially sue the board for misrepresentation. (If HPE were truly meaning business, they would begin a private prosecution here in Companies Court, or ask that a civil Judge in the High Court try the facts to a criminal burden of proof, after which the Crown Prosecution Service ought to feel obliged to act… (notably they did not, when a civil case tried and finding guilty to criminal standard, counsellors who were rigging the postal vote in our neighborhood) .. I think the smaller markets, AIM and there are two others I keep forgetting exist (tells you everything doesn’t it, that we have two more markets one I am certain I recall exists just for tech IPOs!) actually are culturally impeded. After the “Big Bang” of deregulation, when US bulge bracket banks burst in and each took a chunk of the old line firms in the City of London, hereditary interests and the Gentleman’s Club set of financiers were move along, and found themselves doing much the same as before, which was operating without any scale or capital you’d notice in today’s economy.

I have another take, that may be affecting the entire economy of startups:

– just the sheer scale of individual valuations is having widespread effects on perceptions.

– and this includes expectations as to exit valuation.

I think that lawmakers ought in any event to look hard at the SEC rules at which point one has to file public information. The present system of limiting by number of shareholders I believe ceased to work a long time ago. About whenever the first moment attorneys felt comfortable enough with boiler – plate derivatives contracts, whereby equity ownership can be moved without the share registration being moved, or the financial benefits split in the same way, unseen and privately.

I also think that the immense scale of some investments might be seen to have wider economic effect. The operations of Uber for one, are vast, for any company. I think the overriding purpose of all legislation, is to ensure that any entity capable of affecting public interest or utility, is at the very least open to public scrutiny. Both financially an in terms of conduct or the board and so on.

But above all, picking a regulation requirement at a certain scale of value, would incline the capital cycle to include the public market, and dampen the greedier of instincts that sadly do wash down from loft heights, onto the heads of nearby shanty – town startup hard-scrabblers… albeit I have seen absolutely no early stage startup founders looking like they were ever in need of anything in their lives, lately, or if ever, in London. Startups here merely locate to try to look rough and hewn from hunger to succeed, by parking themselves near crowds of youngsters milling between $5MM townhouses on streets I remember as derelict and squatted, the house fronts now stores selling the passers-by clothing that seems lately to be trending towards the Zoolander interpretation of “Derlelict”.coming

But let’s not worry, we can always rely on our American cousins coming up with something new and fantastic. If the pound isn’t ten to the dollar by then!

– – –

My apologies, I nearly forgot the reason I started to open a comment:

Cannot some enterprising motherboard manufacturer come up with a board which can take the Intel compatible equivalent to the IBM Power 8 Centaur chips, that deliver L4 cache and extra DDR channels?

This is absolutely what I want to see, in general.

I know the market for the biggest scale-up servers is so lucrative that 64GB and even (I try not even to think about these) 128GB DIMMS are ridiculous prices, but I cannot think it is so lucrative as so server makers are not interested in the possibility.

A little while back, just before IBM passed things to Lenovo, they had a X series server with a optional memory expansion 1U rack. It wasn’t cheap, we quoted them. But we only reluctantly passed, as the there was a kink in the licensing breaks that worked better for fewer sockets.

To this end, we recently discussed building 4 way from Tyan. The board we looked at, had I think 16 riser cards, for a total of 128 DIMM slots.

Now that still holds our interest. Just we really are going to miss blunt CPU speed that’s getting relatively worse until the next CPU release cycle, from a base that is making us wince to start with. Intel badly needs to decide on its QPI roadmap, because this surely, given a interface you can plug into, will generate products for RAM expansion as well as GPU.

Another point just made to me over my shoulder now, is that in some instances, having a 8U server (I though they were 12, but 8 does sound possible) loaded to the hilt with RAM, actually saves physical space in the rack. I dunno, this is London where cages are half full because there’s insufficient grid electric capacity…

Of course, if anyone knows of any other Intel motherboard with >32 DIMM slots, please let us know here! 🙂

John Other October 14, 2016 at 3:24 pm

My bad, it was a Supermicro board with 96 DIMM slots, and there on the page I just checked, has appeared a interesting new server which seems to accept up to CPUs as cards, giving 192 DIMMS as a max. The upper bound of E5 Xeon memory support seems to be 48 DIMM slots. I merely am among the ranks that surely are not small in number, wishing to be able to add memory to lesser chips as the E7 line. Or just wishing that Intel would offer even one E7 part with low core and high clock.

Probably RWT is the place to ask this, but does anyone know off – hand why ARM vendors have not attacked the high DIMM count market?

The upcoming open supplier Power 9 systems also seem to have taken a turn, only one of those takes 1TB RAM as 32GB DIMMs.

Storage, too, could easily be needing to think of new ways to connect.

I am thinking that NVMe over OmniPath will come soon. If not this year, then early next, we want to look properly at Mangstor’s networking of NVMe. Something concerns me that Intel is unsure of itself here, and “Any Solution Today” will de facto edge out even the most mouthwatering roadmaps. This is tricky, because it is already actively on the tips of our tongues, not only prompted by this article, “When will networking approach a possible solution for addressing memory in a non message passing system?” i.e. at what moment might we take “a tray of cheap DIMMS” (and ideally something like Everspin’s tech) and just plug it into the switch, and perform just sweetly for what the expectations can reasonably be? For some applications, a means to expand memory this way today might be attractive. Now my memory is telling me I read up on just such a product /tech only the other week.

Anyhow, storage tiering will get to be pretty interesting whenever you can stuff trays of DIMMs next to trays of SSDs.

I’m separately expecting someone to come up with JBO-M.2 arrays soon enough, but M.2 cooling and inherent poor 4K perf in the chips chosen for such product cause that to be inelegant.

Yup, just gimme the Everspin stuff and DIMM expansion on 2 sockets please!

Anyhow else thinking this will be like London busses, and three arrive at once? Or all the things arrive about the same time, that fix this space?

Certainly DIMM expansion would start to drive the demand for and pricing on MRAM. I’m sure there’s quite a domino run to set off.

Leave a Comment

Previous post:

Next post: