Malaysian businesses weighing up AI usually hit the same two walls: where the data goes, and how much the bill grows once everyone starts using it. A local company is now selling a single box aimed squarely at both. EMARQUE, a Malaysian technology firm, has launched the EMARQUE AI Eclipse, a desktop AI supercomputer built to run private AI on hardware a business owns rather than on someone else's cloud.
Contents
What the EMARQUE AI Eclipse is
The Eclipse is a compact desktop machine built around AMD's Ryzen AI Max+ 395 processor with 128GB of unified memory. That memory pool is the point: it lets the Eclipse load and run large AI models that would normally overflow the graphics memory on a conventional desktop GPU. It ships with either Windows 11 or Ubuntu Linux at the same price, and adds dual 10 Gigabit Ethernet for fast links to network storage and internal servers.
On the Windows version, EMARQUE bundles its own AI Utility, a front end for working with local models through chat, document question-and-answer, image generation, dictation and model management. For teams that want to go further, the Eclipse also works with established platforms such as Onyx and Hermes Agent, which businesses and developers can use to build private knowledge assistants, coding tools and agent workflows around models running on the machine itself.
Why a business would keep AI in-house
The case EMARQUE is making is about control and cost. Running a model locally means a company can work with its own documents, databases and business systems without sending that material to an external AI service. EMARQUE says it has already tested the approach in internal and customer pilot deployments, including private document retrieval, coding assistants and automation. A firm could, for example, let staff search years of internal files in plain language, or build an assistant around financial and operational data while keeping the underlying systems on site.
That framing is most relevant to finance, insurance, healthcare and legal services, where customer and operational records are sensitive. EMARQUE is careful to note that running AI locally does not remove the need for proper security, access controls or compliance. What it changes is the economics: selected workloads run on hardware the business already paid for, rather than adding a per-request charge every time the AI is used, which EMARQUE argues makes costs more predictable as usage grows.
It is the same argument now driving the national conversation around sovereign AI, where control over data, models and computing matters as much as raw capability. We have seen that debate play out at scale, from Malaysia's RM2 billion sovereign AI plans down to individual machines, and EMARQUE's pitch is that the same principles now apply to a single company.
Part of a wider move to on-device AI
The Eclipse lands as the same AMD Ryzen AI Max platform turns up in more local-AI tools. We have covered Perplexity running 27-billion-parameter models locally on Ryzen AI Max PCs, and a US studio that built its production assistant to run entirely on-device on the same silicon. EMARQUE says it has tested the Eclipse with models including Qwen3-Coder-30B-A3B, GPT-OSS 120B and Qwen3.5-122B-A10B, alongside smaller models for faster everyday tasks, with real performance depending on the model, quantisation and context length.
"Running a model is only the first step," said Kiri Kaniathran, founder and chief executive of EMARQUE. "The useful part starts when AI can work with your documents, applications and business systems while giving the organisation more control over its data."
EMARQUE AI Eclipse price and availability in Malaysia
The EMARQUE AI Eclipse is open for pre-order in Malaysia now, starting from RM17,999. Buyers can choose 1TB, 2TB or 4TB of storage, with Windows 11 or Ubuntu Linux offered at the same price. EMARQUE positions the Eclipse as an entry point into private AI rather than a replacement for larger systems, and says it also builds bigger AI workstations, GPU infrastructure and data-centre-scale setups for organisations that outgrow it. Full specifications are on the EMARQUE AI Eclipse product page.
For a Malaysian business stuck choosing between powerful cloud AI it cannot fully control and local tools that cannot run serious models, a RM17,999 box that keeps the data in the building is a new option worth weighing.
