Specifications
Family
Three workload-native developer platforms: Acies, Actus and Aegis
Acies — Inference-Native
AI inference developer platform for modern edge workloads with an open, standards-based NPU, an open software stack and an easy-to-adopt form factor
Actus — Event-Native
Real-time developer platform for event-driven embedded applications such as motion control, sensor fusion and predictive maintenance
Aegis — Safety-Native
Developer platform for mission-critical physical AI systems that need low-latency and agentic processing
Design Principle
'Workload-native' — each platform is tailored to one class of physical AI device rather than treating physical AI as a single undifferentiated processor problem
Target Systems
Industrial machines, transportation platforms and embedded systems operating in the physical world
Throughput Gain
Multi-threaded architecture improves data throughput by 40% versus a conventional single-thread approach
Software Efficiency
Improves AI stack utilisation by more than 30%
ISA Flexibility
Multi-threaded cores with open RISC-V architecture options for customisation
Tooling
MIPS Atlas Explorer for hardware/software co-development
Safety Path
Streamlined safety validation with a defined certification pathway for the Aegis tier
Automotive Focus
Targets high power consumption, data movement inefficiency and vendor lock-in in vehicle compute
Vendor
MIPS, by GlobalFoundries (GF) — processor IP, software-first tools and GF silicon technologies
Positioning
Brings Physical AI capability to industrial machines, transportation platforms and embedded systems in the physical world
Announced
September 2, 2026
Status
Development platforms; the stated next milestones are measured inference efficiency and supported model lists for Acies, deterministic timing data for Actus, and a published safety architecture for Aegis
Overview
MIPS launched Acies, Actus and Aegis as three separate workload-native platforms rather than one general-purpose AI processor. The premise is that edge AI inference, real-time event processing and safety-critical control have fundamentally different requirements — memory behaviour, determinism, certification — and that a single processor tuned for all three ends up mediocre at each.
Acies is the inference tier: an AI inference developer platform for edge workloads with an open, standards-based NPU, an open software stack and an easy-to-adopt form factor. Actus is the event tier: a real-time platform for event-driven applications such as motion control, sensor fusion and predictive maintenance. Aegis is the safety tier: a platform for mission-critical physical AI systems that need both low latency and agentic processing, with a defined certification pathway.
MIPS, now part of GlobalFoundries, quotes a 40% improvement in data throughput from its multi-threaded architecture and more than 30% better AI stack utilisation, backed by open RISC-V architecture options and the MIPS Atlas Explorer co-development tooling. The platforms target industrial machines, transportation platforms and embedded systems. Buyers should note these are development platforms: the practical milestones are measured inference efficiency for Acies, deterministic timing data for Actus and a published safety architecture for Aegis. QS Compute supplies embedded AI platforms, development kits and industrial edge compute — request a quote.
Key Benefits
Right-sized silicon: separate inference, real-time and safety tiers instead of one compromise part. Measurable efficiency: 40% better data throughput and 30%+ better AI stack utilisation. Open stack: standards-based NPU, open software and RISC-V ISA options. Certification path: a defined safety route for mission-critical deployments.
Applications
Industrial machine control and automation, transportation and vehicle compute, robotics motion control and sensor fusion, predictive maintenance, and safety-critical embedded AI systems.
Request a Quote — MIPS ACIES, ACTUS & AEGIS — WORKLOAD-NATIVE PHYSICAL AI PLATFORMS
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →