ARCEXA Vision
Ideas for the Next Generation of Computing
Explore ARCEXA's broader portfolio and research hardware roadmap. Dedicated semiconductor and R&D pages explain published development status; the homepage highlights our current work.
ARCEXA Portfolio
The ARCEXA Portfolio
Four connected areas spanning silicon, AI systems, engineering software, and advanced research. Programs at different stages are grouped here for exploration.
Program status: The homepage highlights SkyEdge, Yazhi-Vision, and Atlas as current work. Other development and research statuses appear below. ARCEXA products are not yet commercially available.
Semiconductor
SKY Family
- SkyEdgeSkyEdge-R1 · Chip design underway
- SkyRPU (SkyGPU)SkyGPU-G1
- SkyServerCPUSkyServerCPU-C1
- YazhiYazhi-A1
- SkavySkavy-E1
- RudraXRudraX-B / KAVACH
- AI AcceleratorDedicated AI accelerator IP
Edge AI SoC
SkyEdge
SkyEdge is a low-power edge AI vision SoC concept designed to process camera, audio, and sensor data locally. Its reference architecture traces images from camera input through an image signal processor and DSP cluster to an NPU for inference, with a CPU coordinating the system.
For power efficiency, the concept uses an always-on sensing island to wake the main compute only when needed, alongside banked on-chip memory, a two-tier interconnect, and switchable power domains.
These architectural features and performance targets come from an illustrative learning reference; they are design directions, not verified SkyEdge-R1 specifications.
SkyEdge-R1 · Integrated RTL candidate. Its 128 owned tests pass; coverage, stress testing, formal verification, CDC/RDC checks, FPGA bindings, and synthesis remain. SkyEdge-E1 is a paused 28 nm successor study.
Rapid Processing Unit
SkyRPU (SkyGPU)
A heterogeneous compute processor aimed at fast, concurrent workloads. The listed program is SkyGPU-G1.
- Event-driven computing
- AI acceleration
- Parallel and HPC workloads
- Massively concurrent processing
- Low-latency workloads
Neuromorphic and event-based workloads are a potential direction for exploration.
Server CPU
SkyServerCPU
A server-class CPU program focused on scalable compute infrastructure.
SkyServerCPU-C1 is the named program. Further architecture and development details have not been published.
Application-class RISC-V CPU
Yazhi
An independently developed, high-performance RISC-V CPU family. Yazhi-A1 is the first application and edge AI core baseline.
Development baseline; not ready for sign-off. All 28 regression tests pass. CSR, PTW, FP retirement, coverage, formal verification, ISA comparison, and full integration remain open.
Embedded RISC-V CPU IP
Skavy
A clean-room, scalable RISC-V CPU IP family. Skavy-E1 is a five-stage, in-order core for embedded computing.
Skavy-E1 · Active development. Sixteen of 59 planned RTL leaves are complete. The first batch passes all 16 simulation tests with 100% recorded line and branch coverage for that batch; 43 leaves remain.
Power semiconductors
RudraX
A family covering battery management, PMICs, thermal monitoring, telemetry, gate drivers, VRMs, and EV and industrial power controllers.
Planned. RudraX-B / KAVACH is a proposed 48 V BMS controller. RTL and analog implementation have not begun.
Dedicated AI compute
AI Accelerator
Dedicated AI, NPU, and ML accelerator architectures and IP for specialized compute workloads.
IPX
- Semiconductor IP PortfolioAnalog, mixed-signal, and interface IP
Reusable semiconductor IP
Semiconductor IP Portfolio
Reusable analog, mixed-signal, and interface IP blocks intended for integration across semiconductor programs.
DevX Core
- Event-Driven Execution CoreSparse, asynchronous, and sensor-centric compute
Event-Driven Execution Core
DevX Core
A specialized computing core designed for event-driven, sparse, asynchronous, and sensor-centric workloads.
“Process only when an event occurs.”
Target applications include event-driven sensors, sparse computation, and neuromorphic workloads.
AI Systems
- Yazhi-VisionEdge AI vision software in development
- ARC AI
EDA & Engineering Software
- AtlasMVP stage
- Helios
- Vulcan
- VSPIn active development
Advanced Research
ComputeM
- Compute Architecture
- Neuromorphic ComputingActive research
- AI/ML Hardware
- Advanced Accelerators
- Future Computing
- NeuronLab
- QuantumLab
ARCEXA Research
Research & Development
Research Hardware Roadmap
Planned hardware programs and exploratory research initiatives. ARCEXA hardware products are not yet commercially available.
Planned Hardware Programs
Explore planned hardware programs, from edge devices to research silicon. Contact ARCEXA for the latest program status and available technical details.
Phase 1 · Near-Term Edge & AI Devices
Phase 2 · Medium-Term Semiconductor IP
Phase 3 · Long-Term Custom Silicon
Phase 4 · Research Silicon Experimental Processors
R&D Programs
Exploratory research initiatives
Semiconductor Research
- Advanced Computing Architectures
- Processor Architecture Research
- Synchronization Models
- Network-on-Chip Research
- EDA Algorithms
Neuromorphic Computing
Active research
- NeuronLab Architecture
- Thalamus Architecture
- Cortex Architecture
- Alpha Flop Research
- Event-Driven Computing
AI Research
- Multi-Agent Systems
- AI Copilots
- AI for Semiconductor Design
- Autonomous Engineering Systems
Quantum Research
- Quantum Computing Software
- Quantum Toolchains
- Quantum Simulation
- Quantum and AI Systems
Intelligent Systems Research
- Digital Twins
- Autonomous Systems
- Intelligent Robotics
- Adaptive Infrastructure