Narbeh Derhacobian, CEO of fabless nonvolatile memory startup Adesto Technologies Corp. (San Jose, Calif.), is looking expand the remit of his company's conductive bridging RAM (CBRAM). That means higher capacities, smaller geometries, most likely a deal with an Asian foundry – but most of all it means application-specific memories tuned to what the market wants for wearable equipment and the Internet of Things.
Things are looking good for Adesto. The company is one of the first of many contenders to bring a new-generation of resistive RAMs to market. But back in 2012 the company appeared to take a sideways step when it announced the acquisition of Atmel's Serial Flash product families based on EEPROM non-volatile memory technology.
Adesto had been founded in 2006 by Derhacobian with the mission to develop a non-volatile memory called based on the movement of copper ions in a programmable metallization cell technology licensed from Axon Technologies Corp., a spinoff of Arizona State University. So turning to sell an old line of serial EEPROMs looked like it might be distraction.
Not so according to Derhacobian. What the acquisition really brought to Adesto was relationships with 300 customers and dozens of distributors and real-world knowledge of market requirements. "In the first phase you are seeking design wins which often means competing on price. In the second phase you can catch companies in the design cycle and start to sell the features of the technology," Derhacobian said. The EEPROM products provided the introduction that got customers talking to Adesto and prepared to consider CBRAM.
At the time the acquisition was said to "complement Adesto’s line of Conductive Bridging RAM (CBRAM) devices for embedded applications, which had began shipping earlier in the year."
The fact was that Adesto's first generation of CBRAM products had a problem with thermal sensitivity, meaning the devices could not withstand reflow solder temperatures and dwell times. "The team did a fantastic job. We switched it around in a year. We now have six months of baking at 200 degrees C with no failures. It is rock solid," Derhacobian told Electronics 360.
The second generation of CBRAM was intended to go on sale in 2013 and Adesto quietly delivered on that promise and has now shipped more than one million units. (see Adesto Ships 1 Million CBRAM Memories). The RM24x family of CBRAM devices have memory capacities of 32kbit to 1Mbit and functional and electrical compatibility with serial EEPROMs and a choice of SPI and I2C interfaces. "A million units may not seem a big milestone but going from 0 to million is a big step for us," said Derhacobian.
It was a soft launch for the RM24x family, which is manufactured for Adesto by Altis Semiconductor using a 130nm CMOS process. Altis also offers a foundry service with CBRAM as embedded non-volatile memory although there appear to have been few, if any, takers so far.
"CBRAM needed to gain maturity in the market place so it can find its niche," said Derhacobian. "Distributors are stocking CBRAM and using them to replace EEPROM, particularly in Bluetooth LE applications. CBRAM is very cost-effective at low density. A 1Mbit EEPROM die size is 5 or 6 square millimeters while the 1Mbit CBRAM is 2 square millimeters," he observed. It also consumes less power.
Having built up performance and reliability data as an EEPROM replacement Derhacobian now expects CBRAM to go on and start winning design slots based on more direct uses of its inherent non-volatile capabilities. "Over the next three or four quarters you will see a lot more announcements relating to CBRAM features," he said.
But to really blossom surely CBRAM needs to be manufactured much closer to the leading edge and have a second source, even if that is a second wafer fab within the same company?
"We are in advanced discussions to put CBRAM into an Asian foundry at a sub-55nm node," Derhacobian told Electronics 360. That will allow CBRAM to go higher in capacity and increase the access to CBRAM as an embedded non-volatile memory but from Adesto's point of view the CBRAM strategy is not about the highest capacity.
Derhacobian made the point that Adesto is fabless but at the highest capacities – multigigabit – the only memory chip companies that succeed are the ones that own the means of production.. "We are building memories for the Internet of Things era; application-specific memories. CBRAM is scalable in geometry and capable of multi-level cells, but the question is how to compete. We're not going to focus on commodity NOR flash replacement, unless we see a strategic reason for it. We are looking to create added value application-specific memories."
The hiatus while Adesto re-engineered its first-generation CBRAMs and got busy selling EEPROMs means it has been some time since Adesto raised venture capital. The company is backed by Applied Materials, through its Applied Ventures arm and some undisclosed semiconductor companies, said Derhacobian. "We don't plan to seek money from private companies. We are profitable and we are on plan. The plan is to build a strong company that can bring innovative technologies to market."
That's slightly surprising. The company needs to engineer its third-generation of CBRAM products on a more advanced node than Altis can supply, which might take a year or more in terms of time to revenue. A venture capital injection would surely help that go as smoothly and quickly as possible. But if Adesto can achieve its engineering goals quickly using earnings – and some funds quietly slipped to it by strategic investors – then the one step sideways to go two forwards will have been justified.
Related links and articles: