Does phase-change memory ride again with Sandisk?

Technology News | April 11, 2025

By Peter Clarke

Cette publication existe aussi en Français

There are indications that Sandisk Corp. is looking to create a non-volatile, phase-change memory under the direction of an executive who managed Intel’s 3D XPoint based Optane memory business.

When Sandisk, was spun off from data storage company Western Digital in February 2025, the company said it intended to deliver flash memory products while pursuing the development of emerging disruptive memory technologies.

Sandisk to pursue emerging memory technology after split

At the Sandisk Investor Day held on February 11, shortly before the spin off, Alper Ilkbahar, the incoming senior vice president of memory technology, described something he called 3D Matrix Memory as a future development for the company. Prior to joining Western Digital/Sandisk in February 2022 Ilkbahar had worked at Intel from September 2016; first as the general manager for datacenter memory and from October 2020 as manager of the Optane memory business.

In his presentation Ilkbahar said Sandisk had been working on 3D Matrix Memory technology since 2017 and had entered into a “productization phase” through a partnership with IMEC (Leuven, Belgium) that began in 2024 under the title Project Neo. This partnership continues and is being pursued to bridge the “lab-to-fab gap,” something Ilkbahar said was particularly necessary when working with “new material sets.”

In his presentation Ilkbahar did not disclose the material set or the operational basis of the 3D Matrix Memory. He did not even state that it was a non-volatile memory, but given his own history at Intel and his disclosure that the technology is being used as the basis of a US government development contract for a non-volatile strategic rad-hard memory, phase-change is clearly a contender technology.

In addition, two of the diagrams Ilkbahar referenced in the meeting show 3D Matrix Memory as having striking similarities to Intel’s 3D XPoint memory which was non-volatile and widely supposed to be based on chalcogenide phase-change technology.

Isometric projection and cross-sectional view of Sandisk 3D Matrix Memory: Source: Sandisk via Youtube.

The double-deck x-y-x crosspoint memory stack is one similarity. Another is the height-limited, or truncated dome-shaped shadow in the memory cell, a potential characteristic of a thermally-driven phase-change device, although the low resolution of the image makes this inconclusive. Magnetoresistive random access memory (MRAM) is also used for radiation-hardened memory but the cross-sectional view does not have the appearance of an MRAM device. The combination of factors points to 3D Matrix Memory being a continuation of the PCM and 3D XPoint development line.

Optane

3D XPoint technology was developed jointly by Intel and Micron starting in 2006 and was officially announced in 2015. It was the basis of Intel’s Optane memory business which was discontinued in July 2022 after years of financial losses and limited market adoption.

The Optane business made a loss for Intel of US$576 million on sales of US$392 million in 2020, according to the company’s 4Q21 financial filings. Across its product life from 2017 to 2022 – and not including the many years of R&D – the Optane business may have lost Intel more than a billion dollars. Optane was specifically excluded from a deal Intel made to sell its non-volatile memory and solid-state drive (SSD) businesses to SK Hynix for $9 billion, announced in October 2020.

In his Sandisk presentation Ilkbahar said: “We, as the inventors of 3D XPoint memory, are very proud to show you a new, innovative, scalable 3D architecture targeting to deliver DRAM-like performance, 4x the capacity at half the bit cost.”

Ilkbahar did not elaborate on his reference to 3D XPoint. He continued: “At the core of this technology we have a novel memory cell that allows us to scale into dense array architectures. At the product level we intend to attach the memory to industry-standard attach points like CXL buses.”

Ilkbahar showed the audience that the development vehicle with IMEC is a 4Gbit and 8Gbit memory platform and a plan to sample commercial customers with 32Gbit and 64Gbit memories. However, no time line was shown for these developments.

Project ANGSTRM

That 3D Matrix Memory is a form of phase-change memory is supported by the fact that in 2023 Western Digital/Sandisk was the winner of a US$35 million Air Force Research Laboratory award to develop an advanced next-generation strategic radiation-hardened memory under Project ANGSTRM. Project ANGSTRM has a 54-month term and the work is due to be completed by November 2028. Western Digital/Sandisk was the only winner out of a pool ten applicants.

Project ANGSTRM is chartered with developing a strategic radiation-hardened non-volatile memory with requirements of 4Gbit to 16Gbit monolithic memory capacity, endurance of 10^9 to 10^12 read/write cycles and data retention of 10 to 15 years.

It is generally accepted that PCM is one of the best of the emerging non-volatile memory technologies for demonstrating radiation hardness because its operation is based on the change of the material phase rather than on more-easily impacted charge storage. MRAM is also suited to radiation-hardened applications, particularly in space environments, due to its inherent resistance to ionizing radiation and other space-related hazards.

Questions

Intel gave up on selling Optane because of the losses incurred, which may have been related to fabrication costs and poor yield as well as improvements in rival technologies, such as 3D-NAND and stacked DRAM, in the form of high-bandwidth memory (HBM). 3D XPoint struggled to go beyond a double-decker configuration although a four-layer, second generation component was previewed in August 2020.

If the non-volatile memory cell of 3D Matrix Memory is based on a phase-change material there are a number of questions that need to be answered:

What is the material set?
Has Sandisk been licensed to develop the technology by Intel, Micron and/or others?
Does 3D Matrix Memory differ from 3D XPoint and, if so, how?
Has Sandisk identified the technology/market problems that Intel was unable to solve?
Can Sandisk demonstrate that it has solved them?

If 3D Matrix Memory is based on resistive RAM then different questions have to be asked:

What is the material set?
What is the principle of operation?
What are the failure modes, operational sensitivities and reliability issues?
How could a charge-storage non-volatile memory be made radiation-hard to a strategic level?

Up until now it was thought that there were no longer any chip companies pursuing phase-change memory for discrete, high-capacity components, but Sandisk may be the remaining high-capacity champion for a technology with a 55-year history.

Meanwhile STMicroelectronics has deployed PCM for embedded and rugged applications such as automotive microcontrollers (see ST samples MCU with embedded phase-change memory). It has plans to offer embedded PCM on an 18nm FDSOI manufacturing process supplied by Samsung Foundry.

Sandisk was approached by eeNews Europe to confirm or deny that 3D Matrix Memory is a phase-change memory, but the company did not respond by the time this article was posted.