Samsung-Grandis spotlights MRAM potential, and NAND’s chokehold

August 3, 2011 – Korean semiconductor giant Samsung Electronics has acquired Grandis, a maker of spin-transfer torque random access memory (STT-RAM), a flavor of magnetic random-access memory (MRAM). Details were not disclosed about the deal, other than that it closed in July and covered "the full scope" of Grandis from technology and assets to HR, which will be folded into Samsung’s memory chip R&D operations.

Grandis was launched in 2002, backed by Applied Ventures (AMAT’s venture arm), Sevin Rosen Funds, Matrix Partners, Incubic, and Concept Ventures; industry reports suggest their backing to date has been around $15M. Top exec Farhad Tabrizi was Hynix’s VP of worldwide marketing; other top Grandis execs bring backgrounds from Hitachi Global Storage, IBM, Western Digital, and Micron. The company has received two phases of DARPA funding totaling $15M since 2008, and in late 2010 signed on with Notre Dame and others for another $10M DARPA project.

Still the next memory-in-waiting?

MRAM’s promise is for its nonvolatility, power efficiency, and operation at ultrahigh speeds, for applications requiring high-density memory, or lower power consumption (e.g. smart phones). It’s also touted for its scalability, beyond 32nm or whenever current memory technologies finally lose steam. (Renesas was an early Grandis partner, eyeing MRAM’s potential for 65nm devices back in 2005; Hynix did the same in 2008 looking at "beyond 40nm" process nodes.)

Unfortunately for them, current memory technologies have continued to scale well enough to keep next-gen memory technologies at bay. Intel famously said back in 2003 that NAND flash wouldn’t be able to scale past the 60nm node, and now it’s at 20nm and counting, Jim Handy from Objective Analysis told SST. Meanwhile, Toshiba/Sandisk reportedly has a 19nm device dubbed "1X," and another called "1Y" in the works that suggests there’s at least another node in the hopper, he said.

Ferroelectric memory was the heir-apparent next-gen memory many years ago. "Then it was MRAM, then PCM [phase-change memory], then MRAM again," Handy said. And ultimately the industry could devise something entirely new and different.

MRAM as an IP play?

Note that this deal comes just weeks after Toshiba and Hynix announced their own MRAM partnership, aiming to eventually create a joint venture for production and extend patent cross-licensing and product supplies — joining forces to minimize risk and accelerate MRAM’s pace toward commercialization. "MRAM is our next growth platform," stated Hynix CEO Oh Chul Kwon. Kiyoshi Kobayashi, president/CEO of Toshiba’s semiconductor and storage products company, added that they "will strongly promote initiatives in integration of storage solutions including MRAM, NAND, and HDD."

And maybe that’s all there is to this Samsung-Grandis deal: IP positioning. "We see this acquisition simply as a preemptive move by Samsung to secure potential IP and technology in the MRAM arena, and not necessary representing a significant move forward in bringing the technology to mass production," Michael Yang, principal analyst for memory and storage at IHS iSuppli, told SST.

Keep in mind that Samsung already has put its bet down on another new memory technology-for-whenever-NAND-gives-up: PCM (which it calls PRAM), and claims to be shipping actual devices. (IBM and industry/academia consortia recently touted improvements in PCM, and Numonyx/Micron, Samsung, and KAIST provided updates earlier this spring.) Buying Grandis suggests at least Samsung is covering its IP bases in next-gen memory tech, or perhaps even an outright change of strategic direction, Handy says.

Ultimately, NAND will dictate

But in the end, it all comes back to NAND flash’s scalability, as memory makers continue to push NAND flash further and further — and in more volumes at lower costs. "Nobody’s predicting when it’ll end," Handy said, and "none of [the next-gen memory technologies] stand a chance unless flash reaches scaling limit." Until the end is seen for NAND flash, next-gen memory won’t get the push it needs to prove manufacturing cost-competitiveness and achieve commercialization.


Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.