EDACafe Industry Predictions for 2025 – Everspin

By Dr. Sanjeev Aggarwal, President & CEO of Everspin

Sanjeev Aggarwal, Ph.D.

The memory market, along with its adjacent industries, will continue to propel advancements in memory technology, driving higher densities, faster access speeds and lower power consumption. As the sun sets on the ability for NOR Flash memory to scale to smaller nodes (< 40nm CMOS) or larger monolithic densities (> 256Mb), there is a demand for innovation of alternative memory technologies. Magnetic Random Access Memory (MRAM) is an obvious choice for meeting the scaling requirements that NOR Flash cannot meet. Leading foundries such as Global Foundries, TSMC and Samsung are offering embedded MRAM as an alternative to 2x and smaller CMOS nodes.

Everspin is taking the lead in providing stand-alone MRAM as an alternative to stand-alone NOR Flash, offering faster access and lower power, and enabling a simpler systems architecture that does not require data scrubbing and erase before writing. Everspin’s roadmap introduces Automotive Grade 256Mb – 2Gb density MRAM for this market in the 2025/26 time frame. Features include an Octal SPI interface, two orders of magnitude faster writes and two orders of magnitude greater write endurance with no data scrubbing for retention. These innovative memories support faster Over The Air (OTA) updates allowing 2 to 3 orders of magnitude faster configuration updates to the FPGAs in systems. This means a 1Gb UNISYST MRAM can be reprogrammed in about 2 seconds compared to several minutes with NOR Flash.

The global Low Earth Orbital (LEO) market is expected to grow at a CAGR of ~ 13%, significantly increasing this market from ~ $10 billion to ~ $23 billion by 2029. The demand is primarily driven by global broadband internet access through large satellite constellations and earth observation applications. This effort is being led by several companies, including Northrup Grumman, Thales, SpaceX, and Lockheed Martin, to list a few. Everspin’s innovative MRAM with Octal, Quad, or Dual serial interface at densities from 16Mb to 128Mb (PERSYST MRAM available today) and 1Gb (UNISYST MRAM available 2025/26) provides an ideal solution for these applications. MRAM is the preferred memory choice for harsh ambients, including extreme temperatures and radiation exposure. Unlike alternatives like NOR Flash, DRAM, or SRAM, MRAM is radiation tolerant and does not get corrupted in such ambients. There is a growing trend to use commercial (COTS) products in LEO applications, leading to expanding adoption of our PERSYST xSPI products. In space and deep space missions, MRAM technology has been radiation-hardened by Everspin partners, deployed on the Mars Perseverance Rover, and is actively collecting data. Similarly, this technology was deployed by NASA during its Lucy mission to Jupiter. We expect Everspin’s MRAM products to support and be a part of this market, which is expected to grow rapidly in the coming years.

Edge AI computing achieves more energy efficiency compared to cloud-based AI in part due to the energy consumption in transferring data from the edge to the cloud. For this reason, there is a push to bring more intelligent computing to the edge. Concurrently, architectures for edge AI chips are being simplified to support smaller models’ efficient computing and enable local inferencing and decision-making. This means that the memory in these solutions needs to store configuration information and perform compute operations. Everspin’s innovative UNISYST solution is designed to support these functions of store and execute in place. In addition, MRAM is persistent and, therefore, does not require a battery back-preserve configuration data. We expect this Edge AI adoption, where power consumption is critical, to drive the use of Everspin’s UNISYST MRAM solutions. In the years ahead, we anticipate AI solutions will require higher memory densities to support increasingly complex models while maintaining power efficiency.

About Author:

Dr. Sanjeev Aggarwal serves as a member of the Board of Directors and as President and Chief Executive Officer of Everspin Technologies, Inc. With over 25 years of expertise in the non-volatile memory and semiconductor industry, Sanjeev has been instrumental in shaping Everspin since its inception in 2008 through various leadership positions. Prior to his work on MRAM, Sanjeev was part of the team that developed and commercialized Ferroelectric memories at Texas Instruments. Sanjeev was honored as an Entrepreneur Of The Year 2024, Pacific Southwest finalist. He was also nominated for the Outstanding 50 Asian Americans in Business Award, 2024. Sanjeev is Senior Member, IEEE and his technical contributions include over 200 issued patents, more than 100 publications and numerous invited presentations. He graduated from Cornell University with a doctorate in Materials Science and Engineering and received his bachelors from Indian Institute of Technology, Varanasi in Ceramic Engineering.

Tags: AIComputing, AIModelEfficiency, AutomotiveMemory, COTSSolutions, DigitalTransformation, EdgeAI, EmbeddedMemory, Everspin, FPGAUpdates, LEOMarket, LowPowerMemory, MarsRoverTechnology, MemoryTechnology, MRAM, NORFlash, PERSYSTxSPI, RadiationTolerantMemory, SatelliteConstellations, SpaceMissions, UNISYSTMRAM

This entry was posted on Tuesday, January 7th, 2025 at 9:43 pm. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.