Acrab secures over US$350 million in cumulative financing for agentic AI compute platform

Acrab has raised over US$350 million in cumulative financing as the Singapore-headquartered company comes out of stealth with its first-generation agentic compute chip and software solution.

The company is building hardware and software for agentic AI, a class of AI systems designed to understand context, coordinate tools and complete tasks on behalf of users. Acrab said the funding will be used to accelerate platform development, deepen research and development, expand collaborations with global technology partners and strengthen its presence in key international markets.

Vertex, the global venture platform backed by Temasek, was among Acrab’s earliest investors through Vertex Ventures SEA & India and Vertex Growth. Acrab said Vertex has increased its investment as the company reached key technical milestones. Other participants include global venture capital firms and strategic industry investors, although the company did not name them.

Building AI for local task execution

Acrab’s pitch is centred on agentic AI running closer to users, devices and physical environments. The company cited potential use cases across personal AI PCs, home hubs, in-vehicle intelligence, industrial operations and general robotics.

The company said agentic AI systems will increasingly need to understand user intent, coordinate actions and carry out tasks in real time. That creates a different infrastructure requirement from AI systems used mainly to generate answers, because the compute platform must support local language model processing, orchestration and task execution.

Acrab’s first-generation compute platform, GΞLIX, is designed to support local large language models for agentic AI workloads. The company said the platform has been validated in demanding real-world deployment environments and is moving towards first industry adoption and mass production. However, the company did not disclose customers, deployment sites, manufacturing partners or its production timelines.

Acrab’s platform combines silicon and software

Acrab describes its architecture as a full-stack compute platform covering AI silicon, local large language model processing, operating system technology, multimodal human-machine interfaces and agent orchestration.

That combination gives the company a broader technical scope than chip development alone. Acrab is trying to build the underlying platform for AI systems that can process inputs, coordinate software tools and act across different device environments. The company said it expects future AI experiences to become private-by-design, context-aware and capable of proactive assistance.

Dr. Ken Phua, CEO of Acrab, has held leadership roles in the silicon industry, including at Arm UK and Arm China. Acrab said his background in CPU architecture and heterogeneous computing informs the company’s approach to agentic AI infrastructure.

“We are living through a renaissance – the CPU is once again at the centre of the AI era,” said Dr. Ken Phua, CEO of Acrab. “CPUs are becoming increasingly important as AI systems evolve into heterogeneous computing environments, where execution depends not only on NPU performance, but on the seamless coordination between CPUs and NPUs. This is our belief and our core strength in architecture design. Delivering this new generation of agentic experiences calls for a fundamentally new compute foundation. Acrab is building it.”

Vertex backs the edge AI thesis

Vertex’s continued backing is tied to the view that agentic AI will require dedicated infrastructure at the edge. Kee Lock Chua, CEO of Vertex Holdings, said the firm backed Acrab early because it believed the next wave of AI would need a new compute foundation built for local execution.

Founded in 2024, Acrab has built a team with experience across semiconductor design, AI algorithms and operating systems. Its next test will be whether GΞLIX can move from validation into commercial adoption, and whether its full-stack approach can give device makers and industrial users a practical platform for agentic AI workloads.