Acrab raises $130M to scale Agent Box on-device AI inference platform
The AMW Read
Incremental funding update for a known edge-AI silicon player, with segment-level impact on on-device inference hardware but no structural shift.
Acrab raises $130M to scale Agent Box on-device AI inference platform
Acrab announced a US$130 million raise to advance its Agent Box platform, which enables on-device memory, multimodal interactions, and large-model inference. The round signals Southeast Asia's deepening push into AI hardware for enterprise and personal use. It brings Acrab's total disclosed funding to $350M, per the AI Market Watch index, which tracks roughly 5,000 companies (coverage, not a census). The raise follows the company's July launch of the GELIX 1 edge AI chip and Agent Box local inference system, entering the competitive edge-AI silicon market.
The investment underscores a structural shift toward on-device inference, moving workloads off centralized cloud GPUs and into edge silicon. Acrab's focus on local memory and multimodal interactions targets enterprise and personal devices where latency, privacy, and bandwidth constraints favor local processing. This positions Acrab within a crowded edge-AI silicon market that includes established players like Qualcomm and startups such as Hailo and Axelera. The funding also highlights Southeast Asia's ambitions to capture value in the AI hardware stack, a niche historically dominated by US and Chinese companies.
For builders, Acrab's capital and platform roadmap suggest growing viability of running large models entirely on-device, which could reduce inference costs and simplify compliance for enterprise deployments. Investors should watch whether Acrab can convert this funding into design wins across laptops, robotics, or IoT form factors, and how it differentiates on power efficiency and software ecosystem against well-funded rivals.