Acrab’s $130M Gambit to Move AI Agents from the Cloud to the Edge

SINGAPORE – August 05, 2026 – In a move that signals a significant strategic bet on the future of artificial intelligence, Singapore-based Acrab today confirmed a US$130 million Series B financing round. The investment, which brings the two-year-old company’s cumulative funding to over US$350 million, is not merely another cash injection into the booming AI sector. Instead, it represents a well-capitalized challenge to the prevailing cloud-centric AI paradigm. Acrab is making a bold assertion: the next generation of AI—autonomous, task-oriented agents—will not live exclusively in distant data centers, but locally, within our homes and businesses.

Led by continued support from existing investors Vertex Ventures SEA & India and Vertex Growth, with new participation from European and Southeast Asian institutions, the funding is earmarked to commercialize Acrab’s full-stack computing platform. This is a direct play to create the foundational infrastructure for agentic AI, moving beyond today's generative models that answer questions to intelligent systems that can take action. With paths to industrial deployment and revenue expected this year, Acrab is rapidly transitioning from a stealthy deep-tech startup into a formidable force aiming to redefine the AI value chain.

The Strategic Shift to Agentic AI

For the past several years, the generative AI revolution has been tethered to the cloud. Power-hungry, large-scale models from providers like OpenAI and Google run on massive server farms, with users accessing their capabilities through APIs. While transformative, this model presents inherent limitations for the next logical step in AI's evolution: agentic AI. These are systems designed to understand context, remember preferences, and autonomously execute complex, multi-step tasks—from managing a smart factory floor to coordinating a user's digital life.

For such agents to be effective, they require near-instantaneous response times, unwavering operational consistency, and absolute data privacy. The latency involved in sending data to the cloud for processing and receiving a response can be a critical failure point in real-time applications. Furthermore, the sensitive personal or proprietary data these agents would handle makes routing it through third-party servers a significant security and privacy concern for both individuals and enterprises. Finally, the pay-per-use token model of cloud AI becomes economically challenging for always-on, perpetually active agentic systems.

This is the strategic opening Acrab is targeting. The company's core thesis is that the economics and performance requirements of agentic AI demand a new computing architecture—one centered at the edge. By building a platform capable of running powerful AI models locally, Acrab aims to sever the dependency on the cloud for high-stakes AI tasks, offering a solution that is faster, more secure, and potentially more cost-effective over the long term.

Under the Hood: A Full-Stack Hardware Play

Acrab’s strategy is not simply to create a piece of the puzzle; it is to deliver the entire board. The company is developing a full-stack AI computing platform that integrates purpose-built silicon, intelligent software, and a ready-to-deploy hardware system. At the heart of this stack is the GΞLIX 1, Acrab's first-generation System-on-Chip (SoC) built on a cutting-edge 5-nanometer process.

This chip is an engineering feat designed specifically for agentic workloads. It integrates a 20-core Arm CPU with a powerful multicore Neural Processing Unit (NPU) and 273 GB/s of unified memory bandwidth. This architecture allows it to run open-source AI models with up to 100 billion parameters locally—a scale that was, until recently, the exclusive domain of cloud data centers. In internal benchmarks, Acrab claims the GΞLIX 1 achieves a prefill rate over seven times faster than an Apple Mac Mini M4 Pro on specific model configurations, demonstrating its optimization for processing the large prompts and context windows essential for effective AI agents.

This custom silicon powers the company's flagship product, the Agent Box. This personal edge AI device is designed for persistent, on-site operation, enabling local inference and agent orchestration without constant cloud access. For businesses and consumers, this translates into tangible benefits: sensitive data remains on-premise, AI-powered tools function even with intermittent internet, and the recurring cloud token fees are replaced by a one-time hardware investment. Acrab's claims are aggressive, suggesting the Agent Box can deliver performance comparable to an Nvidia DGX Spark system—a workgroup appliance that costs thousands of dollars—at a fraction of the cost and power consumption.

Investor Confidence and a Crowded Battlefield

Accumulating over US$350 million in just two years reflects extraordinary investor confidence, particularly from seasoned players like Vertex. This enthusiasm is fueled by a confluence of massive market trends. The global agentic AI market is projected to skyrocket from single-digit billions in 2025 to well over $100 billion by the early 2030s, while the edge computing market is on a similar trajectory. Investors are betting that the company that provides the foundational hardware for this shift will capture immense value.

“While generative AI helps people find answers, agentic AI will be what empowers them to get things done,” a source familiar with the company's vision noted, highlighting the market's forward-looking perspective.

Acrab's leadership pedigree provides further assurance. CEO Dr. Ken Phua brings deep semiconductor experience from his time as a senior executive at Arm and co-CEO of Arm China, lending immense credibility to the company's ambitious silicon-centric strategy. However, the path forward is fraught with competition. Acrab is entering a battlefield populated by established giants and nimble startups. Nvidia's powerful Jetson modules dominate the robotics and autonomous systems edge market, while Intel, Qualcomm, and Apple are all investing heavily in their own on-device AI capabilities. At the same time, specialized startups like EdgeCortix and Hailo are tackling similar problems. Acrab's full-stack integration is its key differentiator, but it will have to prove its superiority in a market where deep software ecosystems and developer mindshare are critical moats.

From Lab to Market: The Path to Commercialization

Acrab is not content to remain a research project. The company has articulated a clear, two-pronged go-to-market strategy. The Agent Box serves as both a powerful proof-of-concept and a commercial product for early adopters in enterprise and prosumer segments. Simultaneously, Acrab is positioning its GΞLIX platform as a horizontal foundation for a broader ecosystem of device makers.

The company is actively working to embed its technology into AI-powered network storage systems, smart vehicles, and industrial and service robots. The explicit expectation of generating revenue within 2026 indicates that these partnerships are advancing rapidly. Having already validated its platform in demanding industrial environments, Acrab is now moving toward its first large-scale deployments, a crucial step in translating technological prowess into market traction. By providing a comprehensive developer toolchain and reference designs, Acrab aims to accelerate adoption and become the de facto standard for building the next generation of intelligent, autonomous systems at the edge.