This additional capital will be used to help accelerate the evolution and rollout of the company’s MRAM memory products and its patented SmartMem SOC subsystem, addressing the growing demand from AI/Generative AI and IoT markets, and initiate the introduction of a new line of SmartMem SOC Chiplets/Chips.
With the AI/Smart Systems realm expanding exponentially, the availability of memory solutions that merge power, performance, and size is becoming increasingly important. According to Numem, it is looking to redefine industry standards with its AI centric solutions.
The company’s low power NuRAM delivers 20x-290x lower leakage power and 10-30x lower standby power, seen as a game-changer for AI processing from intelligent edge IoT to data centres.
NuRAM is 2-3x smaller than traditional SRAM reducing memory size and making it possible to pack more memory density into the same area, reducing power hungry external DRAM memory access (~60x higher power).
In terms of performance NuRAM operates 2-5x faster with significantly lower dynamic power than most MRAM/RRAM on the Reads and up to 500x faster on the Writes. NuRAM also boasts an endurance that is up to 100x better.
NuRAM integrates a complete Memory Management & Control which makes using the memory as simple as an SRAM by taking care of all the necessary steps to operate the memory.
The SmartMem SOC Subsystem extends some of these advantages to other persistent memories including other MRAM, RRAM/ReRAM, PCRAM and even embedded/external Flash. SmartMem SOC functions can be integrated in NuRAM as SOC Compute in Memory along with some DSP/AI Processing Accelerator compute functions.
"The rise of AI/Generative AI, Autonomous Driving and Smart IoT/embedded applications demand a significant increase in compute and memory capabilities," said Bill Leszinske, Operating Partner at Cambium Capital and Numem Board Member, "Numem's IP building blocks and chiplets are enabling advanced memories like MRAM/RRAM to enable new levels of performance and ultra-low power for these new workloads."
