真乄科技業的頂尖投資團隊

Non-volatile memories are increasingly important for all the applications involving electronic content: from the replacement of hard disk drive by solid state memory to all the applications linked to wearable, date storage... With $78B last year, the stand alone memory market is one of the largest market segment in the semiconductor industry.The technologies in place (NAND flash memory for example) are well established and even more new technologies for non-volatile memories are now quite close to enter the market in order to provide adapted functionalities for enterprise storage applications, embedded memories, wearable applications and possibly mobile phone applications on the long run. RRAM, MRAM, STTRAM, PCRAM are competing technologies for a fraction of this memory business. These new technologies are mainly supported by start-ups that are new in the memory business. Such business model induces to set up a supply chain able to move in to high volume production. This is a unique opportunity for the foundries (TSMC but also TowerJazz...) to enter the memory business, for stand-alone devices. These new technologies are bringing new players both at memory design level and at manufacturing level.

The objective of Yole Développement's new report "Emerging Non Volatile Memory (NVM) Technology & Market Trends" is to provide an updated analysis of the applications trends, technology developments and industrial implementations of these new technologies.



We have decided to open our columns to Crossbar which is one of the most innovative company involved in these new technologies. Mr. Sylvain Dubois, VP Marketing & Business Development at Crossbar, answered our questions on his company and shared his analysis on the evolution of this industry.

Yole Développement: Can you briefly introduce Crossbar, its history and current activity?
Sylvain Dubois: Crossbar has been created in 2010 and is working on new generation of non-volatile memory solutions based on Resistive RAM technology. We went out of stealth-mode in 2013 with the successful demonstration of our first integrated Crossbar RRAM and CMOS test chip. We are now transferring our RRAM technology to manufacturing foundries and ready to start the commercialization phase of our RRAM technology.

YD: What are the competitive advantages of your offer?
SD: Crossbar RRAM technology is CMOS compatible, simple to be manufactured in the Back-End of Line of standard CMOS logic foundries using standard equipment tools and processes, even for our 3D RRAM. This will disrupt the existing memory ecosystem by enabling new players in the memory and storage business.
Crossbar RRAM is scalable below 10nm, this will enable very dense and cost-effective storage solutions and also innovative integrated memory + processing logic system architectures at the most advanced nodes.
Crossbar RRAM technology has superior performances vs electron-based technology (low-latency, byte-alterable, 1M cycles, 10Y retention, no block erase)

YD: What are the different applications targeted by Crossbar?
SD: We are targeting two set of applications: We will provide our RRAM technology as embedded memory IP blocks licensed and integrated in MCU/SoC and we will also design stand-alone products for low latency high capacity low power storage solutions.

YD: What are the market drivers for your solution?
SD: There is a huge demand for more data storage coming from the rise of the Internet of Things, both from the small and smart connected devices up to the mega storage data centers in the cloud. In the 2020 timeframe, there will be about 7.6 Billion people and 50 Billion connected devices uploading or streaming content to the Digital Universe that is expected to reach 44 ZettaBytes. That demand for storage capacity represents huge opportunities and businesses and organization who will be able to extract more value from this data will gain strategic competitive advantages. Big Data storage and analytics are key components of future competition and growth in business. Crossbar 3D RRAM is addressing these applications by enabling new generation of nonvolatile, ultra-dense, low-latency, low energy storage solutions.

YD: What is the timeframe for the development of such applications?
SD: We are currently transferring Crossbar RRAM technology to large manufacturing foundries. Samples will be available for strategic customers starting this year.

YD: At Yole Développement, we expect that enterprise storage will be the biggest market for emerging NVM until 2020 with Storage Class Memory applications. Are you involved there? How are data centers end users (like Google, Facebook...) driving the market?
SD: At Crossbar, we also believe that enterprise storage is a huge opportunity and we are actively working with specific strategic partners in this ecosystem.

YD: How do you expect to enter on this very concentrated market where 5 big memory makers are dominant?
SD: The top five memory manufacturers are currently working on similar Flash-based technology. Planar Flash is already facing scalability challenges below 20nm with deteriorating performances, endurance and retention. 3D Flash development requires massive investments to overcome the 3D manufacturing challenges specific to NAND. This will come with increasing manufacturing cost. 3D Flash will have to scale to small geometries to be cost-effective and they will face same challenges as planar Flash. These are the reason why the industry is realizing that Flash technology is running out of steam and that new emerging memory technologies will take over.

YD: In the long run, when do you expect RRAM to have a larger market share than NAND on the memory market ?
SD: It is hard to tell precisely when the storage industry will complete the shift towards RRAM. We might enter very soon in a transition phase where some applications will migrate to RRAM and other layers in the storage hierarchy will stay with HDD or 3D/TLC-NAND.

YD: There are many different kinds of RRAM on the market (CBRAM, OxRAM, CMox...). What is the positioning of your memory in this landscape and its distinctive features?
SD: We are targeting very dense crosspoint memory arrays. Last December we disclosed how Crossbar solves the sneak path current of cross-point memory arrays and we showed results that has never been published by the industry. At the Crossbar memory cell level, the choice of materials and simplicity of the memory cell structure allows a simple manufacturing process which is a major milestone in the go-to-market strategy of any emerging memory technology company.

YD: How easy or difficult is it to put a supply chain in place for such a technology?
SD: The transfer of our RRAM technology to manufacturing foundries is relatively simple. We are working on several business models to provide this technology to customers as an memory IP license or as a stand-alone memory product.

YD: From your point of view, how can a fabless company be successful in this memory business?
SD: Even the largest manufacturing foundries/fabs are not enough for the largest customers so we believe that a fabless model with strategic manufacturing partnerships is the right approach. This will also be a strategic advantage for big players who want to enter this business and develop their specific storage solutions answering their specific needs in the mobile or big data industry.

YD: Crossbar had recently presented a paper at IEDM conference on its array selector technology. What are the next important milestones in your technology developments?
SD: We are executing precisely on our corporate development plan announced at our out of stealth-mode phase in 2013. We are currently in the Fab transfer phase for embedded applications.

YD: Sneak path current is often mentioned as the main challenge for 3D RRAM. Do you confirm? What are the other technological challenges?
SD: The sneak path current is a major challenge of very dense crosspoint 3D arrays and this has been a great milestone achieved by the Crossbar team in 2014. Ramping up a new memory technology is a challenging adventure and we are working hard with our manufacturing partners, alpha customers, investors to take the right decisions at every steps towards mass-production. We built a strong team of experts from the memory industry with a clear focus to deliver and the agility required to face future technical challenges.

More information about the report Emerging Non Volatile Memory (NVM) Technology & Market Trends here.