Intel And Non-Volatile Memory

Mar.17.14 | About: Intel Corporation (INTC)


Intel is up to their ears in NAND Memory Technology.

Disruptive technology and exotic materials coming.

Where is Micron in all of this?

About a year ago my nephew took an interest in investing and the stock market. So, I've helped him whenever I can, particularly in the technology area. Since then he has gone completely bonkers over investing; reading everything he can, connecting dots, matching the losers with the winners, etc.

It looks like the teacher (me) is becoming the student because he emailed me a copy of this article last night. Please read it…OK, done? on the first read, didn't you think that Intel (NASDAQ:INTC) was ready to dive headlong back into the memory business? I did. Nearly soiled myself. Then I read it again and it became obvious that something even more interesting than Intel simply making memory is going on here.

Let's disassemble this article:

A short preface. The article confirms that Intel is dead serious about the Solid State Drive business and that they think they have some disruptive NAND technology coming, they mention a high speed interface (NVMe), small size, close proximity to the processor chip, and they talk about 3D NAND; not Micron (NASDAQ:MU) and 3D NAND, Intel and 3D NAND. Exciting and scary at the same time. Exciting because Intel is home to the most advanced semiconductor technology on the planet. Scary because of what it MIGHT do to the new memory oligopoly, including Micron.

First of all, this is no rumor; the information comes directly from Intel during a press day at their Non-Volatile Memory Solutions Group (NSG) headquartered in Folsom, CA. Next, upon careful reading, there is no mention of actual memory chip fabrication directly by Intel.


Whereas the first few generations of SSDs were squeezed into traditional hard disk drive (HDD) form factors to fit into typical PCs, servers, and data center equipment configured for mechanical drives, Intel is now developing smaller and more customizable form factors for its SSD products to better suit new systems that can maximize the advantages of flash memory.

Going forward, Intel is also seeding the market for a new interconnect bus standard called NVM Express-or NVMe for short, the name obviously referencing the established PCIe standard for connecting PC components and peripherals to each other. The advantages of NVMe will be to position non-volatile memory closer to logic processors in clients, servers, and storage systems while shaking out bandwidth and latency bottlenecks to really accelerate the performance and total cost of ownership advantages SSDs bring to the table.

The two inch format for SSDs is what you would do in the beginning of a transition to SSDs; make the SSD physically compatible with the incumbent HDD (Hard Disk Drive) format so that a PC supplier could offer the SSD at a higher cost, higher performance option or as a replacement of an existing HDD. To me that two inch format has always seemed like opening a box of breakfast cereal and finding it 10% full. A 480GB SSD can be built on a PC board about the size of a stick of chewing gum. At that size the SSD can be located right next to the CPU chip in a PC (which improves performance and reduces power consumption) instead of several inches (of wire) away. That sounds like a nonsensically small deal, but in the world of 2 and 3 GHz clock rates a few inches of wire is a big deal, and driving those wires, which look like capacitors, at 2Ghz consumes a lot of power.

Leaving the two inch format behind is a watershed event that basically means leaving HDDs behind…forever. It is one thing making SSD that can replace HDD, but making PCs with SSDs that can NEVER accept a HDD is a horse of an entirely different hue. The key here is the controller technology that can provide the speed that a compact SSD, made of NAND, is capable of. An Intel built SSD controller chip made on the most advanced fabrication technology is the "secret sauce" that would be the differentiating factor in an Intel SSD.

Another excerpt:

"Transitioning to 3D

Interestingly, while the lithographic methods used to fabricate logic processors are poised to keep shrinking transistor scales in keeping with Moore's Law for at least several more generations, NAND and DRAM fabrication is starting to run up against some problems as it proceeds along half-pitch lithography steps, he said.

"We don't want NVM to become a bottleneck in Moore's Law," Crooke said. That's probably an understatement for somebody leading a major product group at the company co-founded by Gordon Moore, who laid out his prescient, remarkably durable formula for the simplified economics of the semiconductor industry back in 1967.

Intel's solution to this pressing problem? Moving to a three-dimensional layout for NAND flash memory circuits.

"With NAND, we back off on the lithography shrinks a little bit and go to 3D, stacking elements on top of each other," Crooke said.

Intel transitioning to 3D NAND? I thought that was Micron's job. Well, it is Micron's job, but obviously with a lot of help from Intel through the Intel/Micron joint venture.

The JV was updated back in February of 2012 to include emerging memory technologies (and to shore up the Micron balance sheet to acquire Elpida.)

There is a lot of discussion about 3D NAND lately. This is a decent discussion on the subject. The important thing about 3D NAND memory will be the ability and technology for deposition and etching high aspect ratio structures. Because of their leadership on TriGate transistors, Intel has learned a great deal about deposition and etching these types of structures.

The beginning 3D technology from Samsung (OTC:SSNLF) will be 24 layers and is now experiencing some slowdown. Industry pundits claim that 3D NAND will not be economically viable until at least 32 layers. Recent rumors suggest the Micron (Intel) 3D NAND will start at 32 layers and samples will be available by summer.

A very similar article about a year old contains much of what the latest article does, but is a presentation of the management of the Intel/Micron Joint venture.

Here is what Kayvan Esfarjani, Vice President of technology and manufacturing at Intel and Co-CEO of IMFT, has to say about the IMFT 2D 16nm NAND: "....leveraging its HKMG experience gained in logic circuits."

Since Micron doesn't make logic circuits, we have clear confirmation that the Micron (Intel) 16nm HKMG (High K Metal Gate) NAND chip is largely Intel technology, as would be expected from the leader in HKMG. So this sometimes-casual Intel/Micron JV called IMFT is actually an intense relationship driven by Intel technology and Micron manufacturing.

Another excerpt:

…Intel is putting its money where its mouth is when it comes to the SSDs it has on hand right now. The chip giant has moved its own client systems, the ones used by Intel employees in their cubicles all around the world, to "100 percent SSD".

This is the first action with a promotional slant to it, specifically for Solid State Drives. I think the market for SSDs will respond to information and promotional advertising. I would expect the next thing we hear is some major PC manufacturers announcing a series of PCs that only come with the tiny SSD format.

So the bottom line of all of this is that the IMFT (Intel Micron Flash Technologies) joint venture is more alive and well than we might have guessed it to be before this press event.

Intel HAS and IS providing HKMG technology help on the Micron 20 and 15nm NAND.

Intel IS providing technical help on Micron 3D NAND, as any close observer might have guessed.

Intel WILL be growing their Non-Volatile (NAND/SSD) memory business as I've suggested in the past.

A good bet is that IMFT will utilize fully depreciated excess fabs to expand the NAND business for both members of JV. In order to not upset the stability of the new memory oligopoly, the output of any addition IMFT fabs will probably be dedicated to the highest density NAND chips for SSDs with no component sales to the merchant market.

The market for SSDs is truly enormous and IMFT is really the only entity positioned to address it.

The question to be answered is who will follow an Intel-Micron partnership into the expanded NAND business if they have to spend $5 billion and wait 3 years for a fab of their own, while IMFT has immediate access to fabs that might have a total cash cost of $1 billion each?

This IMFT cooperation teamed with about six excess Intel fabs could add $25 billion in profitable SSD sales each of Intel and Micron. I think that IMFT sidesteps the general memory oligopoly and builds a SSD duopoly.

How does anyone compete with this dream team?

I'm very long Intel and Micron.

Disclosure: I am long INTC, MU. 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.