TetraMem
Category: AI Chips / Semiconductors
TetraMem develops analog in-memory computing (IMC) AI accelerators using memristor/ReRAM technology for ultra-low-power edge and data-center inference. TetraMem was founded in 2018. The company is led by Ning (Glenn) Ge. Based in San Jose, California, United States. Team size: 51-100. Latest round: Undisclosed. Key investors include SAIC Capital, Foothill Ventures, Brown Venture Associates, NYX Ventures (Los Altos), Galaxy Interactive, Photon Fund, Lam Capital, University of Massachusetts Amherst.
- Founded
- 2018
- Headquarters
- San Jose, California, United States
- Team size
- 51-100
Value proposition
Analog in-memory computing that fuses storage and computation in memristor crossbar arrays, eliminating the von Neumann bottleneck and achieving ~100x energy efficiency gain and ~10x throughput improvement over digital processors for AI inference
Products and solutions
MX100 (65nm test chip, published in Nature/Science), MX200 (22nm commercial AI inference accelerator for edge), MLX200 silicon validation completed, Memristor-based in-memory computing SoC with SK hynix achieving 21.3 TOPS/W at 65nm, 700°C-rated RRAM for extreme-environment AI
Unique value
Proprietary "computing memristor" technology that performs neural network multiply-accumulate operations physically inside memory cells, delivering 21.3 TOPS/W at 65nm — orders of magnitude more efficient than GPU-based inference
Target customer
Edge AI device makers (hearables, wearables, IoT sensors, smart cameras); smartphone/AR/automotive OEMs; data-center operators for energy-efficient AI inference
Industries served
Edge AI / IoT, Wearables & Hearables, Smartphones, Automotive (perception), AR/VR, Data Center AI Inference, Aerospace/Defense (extreme-temperature)
Technology advantage
Proprietary analog memristor/ReRAM crossbar array architecture; validated on 65nm (21.3 TOPS/W) with roadmap to 22nm and 12nm; 700°C-rated memristor published in Science; partnerships with SK hynix (world's #2 memory maker) and Andes Technology (RISC-V); CHIPS Act support via NY CREATES for 300mm wafer-scale manufacturing
How they differentiate
Unlike digital accelerators (GPUs, NPUs) that shuttle data between separate memory and compute units, TetraMem's analog crossbar arrays perform computation at the physical location of data storage using Ohm's Law, eliminating ~90% of energy wasted on data movement. The memristor technology is built on standard CMOS/ReRAM processes (no exotic materials) and has been validated on 300mm wafers through NY CREATES.
Main competitors
Mythic AI, Rain Neuromorphics, d-Matrix, Hailo, Lightelligence, Quadric IO
Key partnerships
SK hynix (joint research on memristor-based IMC SoC, published in Advanced Intelligent Systems), Andes Technology (RISC-V vector processor integration), NY CREATES / Albany NanoTech Complex (300mm wafer upscaling under CHIPS Act), UMass Amherst (Northeast Microelectronics Coalition Hub grant), Lam Capital (Lam Research venture competition)
Major milestones
2018: Company founded by four memristor pioneers, 2023: Partnership with Andes Technology for RISC-V AI accelerator, 2024: MX100 prototype shipped to partners, 2025: 700°C memristor published in Science, 2026: SK hynix collaboration published in Advanced Intelligent Systems (21.3 TOPS/W at 65nm), 2026: 300mm wafer-scale manufacturing validated at NY CREATES, 2026: MLX200 silicon validation completed
Market positioning
Early-stage semiconductor startup positioned at the intersection of analog in-memory computing and AI inference acceleration, targeting the post-HBM architectural shift for energy-efficient AI. Competing with both established GPU vendors (NVIDIA) and emerging analog AI chip startups.
Geographic focus
United States (Silicon Valley) with global partnerships (South Korea - SK hynix; Taiwan - Andes Technology)
Patents and IP
Co-founder J. Joshua Yang holds 121+ granted patents in resistive switching materials, memory, and neuromorphic computing. Co-founder Qiangfei Xia elected National Academy of Inventors Fellow.
About Ning (Glenn) Ge
Ex-HP Master Technologist (12 years); Ph.D. in Electrical Engineering, MBA
Latest news about TetraMem
More AI Chips / Semiconductors companies
Official website: https://tetramem.com