Numem

What Numem does

Numem is a fabless semiconductor company founded in 2016 and headquartered in Sunnyvale, California, developing MRAM-based memory products for AI workloads spanning edge devices to data centers. The company builds on foundry-standard STT-MRAM (Spin-Transfer Torque MRAM) cells and differentiates via three patented technology platforms: the NuRAM MRAM cell architecture (delivering up to 2.4-2.5x smaller area than SRAM and up to 100x lower standby power), the AI Memory Engine (AIME) subsystem IP, and the SmartMem SoC subsystem (including the PEP Engine, SuperBIST, and SmartMMC). Core products include discrete MRAM chips and chiplets (up to 1GB per chip, multi-GB via die stacking, with UCIe-compatible chiplet form factor targeting up to 4TB/s per 8-die stack), Numem MRAM IP for embedding into xPU or ASIC designs, and AIME IP that also works with third-party MRAM, RRAM, PCRAM, and Flash.

The company operates four revenue streams: hardware sales of chips and chiplets (sampling targeted late Q4'25), IP licensing of MRAM and AIME subsystems, and design services spanning architecture definition through silicon validation. Go-to-market is sales-led and enterprise-focused, targeting AI chip manufacturers, hyperscalers, automotive OEMs, and edge AI device makers, supported by a network of offices in Sunnyvale, Mesa (AZ), London, Ahmedabad, and Tokyo. Customer segments span Edge AI, Data Center, AI Servers, Automotive (ADAS/IVI), Wearables, and Physical AI (humanoid robots, autonomous machines). The leadership team is concentrated with former Intel executives — including CEO Max Simmons (ex-Intel NAND/Optane commercialization), board member Rob Crooke (founding CEO of Solidigm, ex-Intel NVM Solutions GM), and senior technical advisor Ashu Bakhle (ex-Intel VP Engineering) — reflecting a deliberate strategy to leverage deep NVM commercialization experience to transition the company from technology development to broader commercial scale.

Name

Numem

Legal name

Numem Inc.

Website

https://numem.com

Company type

Private

Founded year

2016

Operating status

Operating

Headcount range

11–50 employees

Short description

Numem is a fabless semiconductor company developing MRAM-based memory chips, chiplets, and IP for AI workloads. Built on foundry-standard STT-MRAM with proprietary AIME subsystem, it serves AI chip manufacturers, data centers, automotive OEMs, and edge AI device makers via hardware sales, IP licensing, and design services.

Ownership category

akta.pro rank

Product category

Semiconductor Memory (MRAM)

NAICS

Computer Storage Device Manufacturing (334112), Computer and Peripheral Equipment Manufacturing (33411), Computer Systems Design and Related Services (54151)

SIC

Semiconductors & Related Devices (3674), Computer Storage Devices (3572), Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

Persistent/Non-Volatile Memory (MRAM/ReRAM/PCM/FeRAM) (HDAHAKAF)

akta.pro secondary industries

AI Memory & Storage for Training (HBM/DDR/SSD/Object Storage) (HDAAAAAF), AI Hardware IP & Chiplet/Packaging for Accelerators (EDA/IP, advanced packaging) (HDAAAAAL), Memory Modules & Subsystems (DIMM/SO-DIMM/HBM stacks) (HDAHAKAH)

Keywords

• MRAM memory products
• AI memory subsystems
• Non-volatile memory IP
• Memory chiplet solutions
• Embedded memory design

Headquarters

HQ city

Sunnyvale

HQ country

United States

HQ region

North America

Offices5 records

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales, Supply Chain

Revenue model

1. Discrete MRAM Chips and Chiplets: High-efficiency MRAM-based memory delivered through discrete BGA packages and chiplets for SoC integration. Advanced stacking enables scalable, high-density memory in compact form factors up to 1GB per chip and scalable to multi-GB capacity through die stacking.
2. Numem MRAM IP Licensing: Licensing of proprietary MRAM IP for integration into xPU or ASIC designs. Optimized for power-sensitive AI and edge applications as an ideal upgrade for traditional on-chip SRAM or embedded Flash.
3. AIME IP Licensing: Licensing of the AI Memory Engine IP subsystem that works with Numem MRAM, third-party MRAM, RRAM, PCRAM, and Flash technologies for system-level optimization.
4. Design Services: Specialized design services for integrating ultra-low-power non-volatile memory and intelligent memory subsystems into advanced SoCs, ASICs, chiplets, and AI platforms. Supports customers from architecture definition through silicon validation and production.

Pricing tiers

Go-to-market motion1 record

Distribution channels4 records

Marketing channels7 records

Core offering

Numem designs and sells MRAM-based memory products—including discrete BGA chips, chiplets, and synthesizable IP cores—augmented by its patented AI Memory Engine (AIME) subsystem architecture. Offerings target AI workloads across edge devices, AI servers, data centers, automotive, and wearables, delivering SRAM-class performance with significantly lower power consumption and higher density than conventional SRAM/DRAM. The company also provides design services to integrate its memory and subsystem IP into customer SoCs, ASICs, and chiplets.

Product overview

Numem offers a portfolio of MRAM-based memory solutions comprising discrete chips/chiplets, IP cores, and design services. The core offering includes Numem MRAM Chip/Chiplets (discrete BGA packages and chiplets up to 1GB capacity), Numem MRAM IP (embedded memory IP with 2.4X smaller area than SRAM), and AI Memory Engine (AIME) IP (intelligent memory subsystem supporting multiple memory types). These are complemented by SmartMem SOC subsystem technology (PEP Engine, SuperBIST, SmartMMC) and specialized design services for SoC integration. The portfolio addresses AI workloads from edge devices to data centers.

Differentiator

Problem solved

Functional benefit

Brands

• NuRAM: Numem's MRAM-based memory technology delivering SRAM-class performance with 2.5x smaller area and up to 8000x lower leakage power than SRAM.
• SmartMem
• AI Memory Engine (AIME)

Products and services

• Numem MRAM Chip / Chiplets High-efficiency MRAM-based memory delivered through discrete BGA packages and chiplets for SoC integration. Supports up to 1GB per chip with scalable multi-GB capacity through die stacking, standard SRAM and JEDEC-compliant LPDDR interfaces, with UCIe support for AI and heterogeneous computing integration.
• Numem MRAM IP Next-generation memory IP based on proven MRAM technology delivering fast access times, ultra-low leakage power, and up to 2.4x smaller cell area than SRAM; designed as an ideal upgrade for xPU or ASIC designs optimized for power-sensitive AI and edge applications.
• AI Memory Engine (AIME) IP Fully synthesizable and configurable memory subsystem IP that enhances power efficiency, performance, and endurance for Numem MRAM, third-party MRAM, RRAM, and Flash. Enables SRAM-like speed with up to 100x lower standby power and integrates into Edge AI, Physical AI, and Data Center AI platforms.
• Design Services Specialized design services for integrating ultra-low-power non-volatile memory and intelligent memory subsystems into advanced SoCs, ASICs, chiplets, and AI platforms. Supports customers from architecture definition through silicon validation and production, including custom memory solutions across diverse process technologies.
• MRAM Chiplet Solution Nonvolatile, high-speed, ultra-low power chiplet solution delivering up to 4TB/s per 8-die memory stack and 4GB per stack package, with multi-state flex power functions (active/standby/deep sleep) and UCIe interface compatibility for AI and heterogeneous computing integration.

Quantifiable outcome

• Up to 2.4-2.5x higher memory density in the same embedded footprint vs SRAM
• +6 more outcomes

Segments6 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Feature6 records

Partnerships

One partnership is on record.

• IC'AlpscoreStrategic or Co-development Partner · 25 June 2024IC'Alps, a leader in ASIC/SOC design and supply chain management, collaborated with Numem to develop an ultra-low-power SOC featuring RISC-V processors, 2 MBytes of NuRAM, and a DSP/AI Custom Datapath Accelerator. The custom SoC was developed in an advanced technology node. IC'Alps provided physical implementation in a secure environment. The partnership aims to extend collaboration to serve new customer SOC projects.

Scale indicators5 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• Everspin Technologies: Everspin is the leading commercial supplier of MRAM and STT-MRAM products for industrial, automotive, and enterprise applications, making it the most direct competitor to Numem's MRAM chip and IP portfolio.
• Avalanche Technology: Avalanche Technology develops MRAM and pMTJ STT-MRAM products targeting aerospace, industrial, and AI applications, competing directly with Numem's discrete MRAM and embedded IP offerings.
• Crossbar Inc. Crossbar develops ReRAM (Resistive RAM) for AI edge, IoT, and embedded applications; since Numem's AIME IP supports ReRAM integration alongside MRAM, Crossbar is a competing NVM technology with overlapping target customers.

Broad incumbents

• Samsung Foundry: Samsung is both a foundry partner for STT-MRAM (on which Numem's technology is built) and a developer of its own embedded MRAM and memory products, making it both an enabler and a competitor in the AI memory space.
• GlobalFoundries: GlobalFoundries offers foundry-standard embedded MRAM (eMRAM) on its 22FDX and 12LP+ platforms, directly competing with Numem's MRAM IP for embedded memory designs in AI, automotive, and IoT SoCs.
• TSMC: TSMC provides embedded MRAM as part of its foundry IP portfolio for SoC integration, overlapping with Numem's IP licensing business and serving as a potential foundry partner for Numem's discrete MRAM chip offerings.
• Intel (Non-Volatile Memory Solutions Group): Intel developed 3D XPoint/Optane and embedded MRAM technologies before divesting its NAND/SSD business to SK hynix (forming Solidigm); Numem has recruited former Intel NVM leaders, positioning it as an indirect competitor to Intel's legacy memory roadmap.
• Rambus Inc. Rambus provides memory interface IP (HBM, GDDR, DDR) and is expanding into chiplet and AI memory subsystem IP, overlapping with Numem's AIME memory subsystem IP and design services for AI accelerators.

Emerging players

• Weebit Nano: Weebit Nano commercializes ReRAM/PCRAM IP for embedded non-volatile memory targeting AI, automotive, and IoT SoCs, competing with Numem's MRAM IP for design wins in power-sensitive applications.
• Renesas Electronics: Renesas has integrated embedded MRAM into its microcontrollers and automotive SoCs, making it both a potential customer and indirect competitor to Numem's automotive-grade MRAM IP offerings.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks5 records

Customer concentration

• LinkedInlinkedin.com/company/numem

Revenue estimate

Valuation estimate

Number of profiles

Profiles7 records

Funding overview

Funding rounds1 record

Investors2 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Numem