Acrab Raises $480M to Push Edge AI Chips as Agent Era Arrives
Acrab
Google/DeepMind
NVIDIA
Qualcomm
Microsoft
Meta
Amazon/AWS
Singapore-based AI startup Acrab has completed a $130 million Series B funding round, bringing total funding to over $480 million. The company, founded in 2024, has launched its first edge AI chip GΞLIX 1 and the Agent Box device, targeting the growing demand for on-device AI processing. Acrab focuses on high-end edge AI servers and aims to provide a new computing foundation for agentic AI workloads.
Acrab, a Singapore-based AI computing infrastructure company founded in 2024, has completed a $130 million Series B funding round led by Vertex Growth and other investors, bringing cumulative funding to over $480 million. The new capital will be used to expand production capacity, grow the technology ecosystem, and accelerate development of the next-generation AI computing platform. Just a month before the funding announcement, Acrab unveiled its first-generation edge AI chip GΞLIX 1 and the Agent Box, a personal AI center designed for on-device processing of billion-parameter models. The company has already begun customer onboarding and is preparing for mass production, a rapid pace for a chip startup. Acrab's strategy shifts focus from cloud AI to edge computing, arguing that while cloud remains dominant, the rise of agentic AI—which requires low latency, privacy, and continuous task execution—will drive demand for powerful on-device chips. The GΞLIX 1 chip boasts 700 TOPS peak performance, unified memory, CPU-NPU collaboration, and support for long-context tasks, with internal tests showing prefill speeds 7 times faster than leading edge platforms. Acrab also highlights the architectural challenges of agentic AI, referencing Microsoft's research on CPU-GPU boundaries, and positions its software stack to enable full agent workflows locally. The company, led by former Arm executive Ken Phua, aims to license its platform to device makers in AI PC, automotive, robotics, and industrial sectors. Future success hinges on which devices adopt GΞLIX 1 and whether initial customers scale production.
- Abbreviations
- GPU = Graphics Processing Unit — графический процессор
- NPU = Neural Processing Unit — нейропроцессор
- CPU = Central Processing Unit — центральный процессор
- TOPS = Trillions of Operations Per Second — триллионы операций в секунду
- PC = Personal Computer — персональный компьютер
- NAS = Network-Attached Storage — сетевое хранилище
Source: QbitAI 量子位 —
original
