Acrab raises over US$350M to build the compute foundation for agentic AI

As AI moves from generating answers to completing tasks, the compute layer has to move closer to where decisions are made. A Singapore company is building it.

The first wave of AI was about generation: models that could write, summarise, and answer. The next wave is about action. Agentic systems reason, plan, coordinate tools, and carry out tasks on behalf of the people and environments they serve, learning and adapting as they go.

That shift changes what the infrastructure underneath has to do. Most of today's AI stack was designed for cloud-based training and inference. Agentic workloads pull in a different direction: lower latency, local inference, multimodal interaction, and continuous decision-making, much of it running at the edge rather than in a distant data centre.

Building for action, at the edge

Acrab, a Singapore-headquartered frontier technology company, is building for that gap. Founded in 2024, it develops a full-stack compute architecture for agentic AI, spanning AI silicon, local LLM inference, an operating system, multimodal human-machine interfaces, and agent orchestration. Its first-generation platform, GΞLIX, runs local LLMs to support agentic workloads, and has been validated in demanding real-world environments as it moves towards first industry adoption and mass production.

Acrab's bet is on the processor at the centre of all this. As inference moves on-device, the company argues, performance will depend less on raw NPU throughput and more on how tightly CPUs and NPUs work together in heterogeneous systems.

"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," said Dr Ken Phua, CEO of Acrab. "Delivering this new generation of agentic experiences calls for a fundamentally new compute foundation. Acrab is building it."

It is a view shaped by experience. Acrab is led by Dr Phua, whose career in silicon includes heading Asia Applications Engineering at Arm and later serving as co-CEO of Arm China.

A market forming at the edge

The opportunity Acrab is chasing sits close to the user. Agentic systems are starting to appear in personal AI PCs and home hubs, in vehicles, across industrial operations, and in robotics. In each of these settings, AI that is private-by-design, context-aware, and able to act on its own depends on compute that can run locally and respond in real time. The company's wager is that this returns the CPU to the centre of the picture, coordinating the mix of processors that on-device intelligence now requires.

Vertex's perspective

Vertex Ventures SEA & India and Vertex Growth were among Acrab's earliest backers, and have increased their commitment as the company reached successive technical milestones. Emerging from stealth, Acrab has now raised over US$350 million in cumulative financing since its founding, with the rounds also drawing a group of global venture capital firms and strategic industry investors.

"We backed Acrab in its earliest days, because we shared a conviction that the next wave of AI won't run in the cloud, it will run at the edge, and it will need a new compute foundation built for it," said Kee Lock Chua, CEO of Vertex Holdings and Managing Partner of Vertex Ventures SEA & India and Vertex Growth. "Acrab is building exactly that. Our confidence has only deepened as the team has translated that thesis into a validated platform, and we've increased our commitment at every step."

With the new capital, Acrab plans to accelerate platform development, deepen research into next-generation agentic compute, expand collaborations with global technology partners, and build its presence in key international markets. If the agentic era unfolds the way the company and its backers expect, much of that intelligence will run quietly at the edge, on a compute layer built for the job.