LENZO
Category: AI Chips / Semiconductors
Japan-based semiconductor startup (NAIST spinout) developing power-efficient CGLA (Coarse-Grained Linear Array) AI inference chips, led by an ex-PlayStation chip engineer. LENZO was founded in 2024. The company is led by Kenshin Fujiwara (藤原健真). Based in Ikoma, Nara, Japan. Team size: 11-50. Total funding raised: $3.3M. Latest round: Seed. Key investors include Incubate Fund, Sony Innovation Fund (Sony Ventures), Mitsubishi UFJ Capital, Monozukuri Ventures, Oki Matsumoto (Monex Group founder), Takuya Hirose (ex-Remixpoint CFO).
- Founded
- 2024
- Headquarters
- Ikoma, Nara, Japan
- Team size
- 11-50
- Total funding
- $3.3M
Value proposition
Ultra-efficient AI inference hardware built on the proprietary CGLA architecture that delivers dramatically higher performance-per-watt than GPUs, targeting energy-constrained edge, robotics, and physical AI workloads.
Products and solutions
A-Series AI processors (65 TOPS INT8, 5-10W power envelope, up to 64GB LPDDR4, PCIe Gen3) built on TokenProcessor™ / CGLA architecture, supports PyTorch, TensorFlow, ONNX frameworks.
Unique value
CGLA (Coarse-Grained Linear Array) architecture combining programmability of CPUs with efficiency of specialized accelerators, embedding power-awareness at the hardware level for ~90% power reduction vs GPUs.
Target customer
Robotics, industrial automation, autonomous systems, edge infrastructure operators, blockchain/crypto mining operators, and enterprise customers seeking power-efficient compute.
Industries served
AI inference hardware, robotics, autonomous systems, industrial vision, edge infrastructure, blockchain/crypto mining.
Technology advantage
Proprietary CGLA architecture developed over years at NAIST (60+ patents from the NAIST Computing Architecture Lab); non-von-Neumann design that reduces memory-compute data movement bottleneck; high performance-per-watt in compact form factor.
How they differentiate
Focuses on power efficiency and adaptability rather than raw peak performance, targeting edge/physical AI and blockchain workloads where energy is the bottleneck; non-von-Neumann CGLA architecture vs GPU-centric designs.
Main competitors
NVIDIA (dominant GPU incumbent), Cerebras, Groq, SambaNova, Japanese peers Preferred Networks and EdgeCortix.
Key partnerships
NAIST (Nara Institute of Science and Technology) research origins, TSMC chip manufacturing, investors include Sony Innovation Fund, Incubate Fund, Mitsubishi UFJ Capital.
Notable customers
Engaging infrastructure operators, blockchain operators, and enterprise customers (early access program), no named public customers yet.
Major milestones
Founded Dec 2024, pre-seed round Oct 2025, 500M JPY seed round Mar 2026 (first silicon production), presented at CadenceLIVE Japan 2026, SEMICON Taiwan 2026, NIKKEI Tech Foresight LIVE.
Growth metrics
Seed round post-money valuation estimated at ~¥2.4B (~$16M); 371 LinkedIn followers.
Market positioning
Early-stage Japanese challenger to NVIDIA in power-efficient AI inference, entering via crypto mining market first to fund long-term AI semiconductor and software ecosystem development.
Geographic focus
Japan-originated, targeting global market; offices in Nara, Kyoto, and Yokohama.
Patents and IP
60+ patents from the NAIST Computing Architecture Lab (CGLA architecture research).
About Kenshin Fujiwara (藤原健真)
Ex-Sony Computer Entertainment (PlayStation 2 Emotion Engine & PlayStation 3 Cell Broadband Engine CPU/GPU developer); California State University BS Computer Science. Serial entrepreneur (4 prior startups, 3 exits); former VC; ex-CEO/founder of HACARUS.
Latest news about LENZO
More AI Chips / Semiconductors companies
Official website: https://lenzo.com/
