TokenVisor aims to optimise monetisation of AMD GPU LLM neoclouds
Embedded LLM has launched TokenVisor, a platform designed to assist AMD GPU neocloud providers in monetising and administrating Large Language Model (LLM) workloads more effectively.
TokenVisor has been developed as a control plane specifically for AMD GPU-powered neoclouds, providing users with a new approach to deploy, manage, and commercialise LLM resources.
The platform enables GPU owners to set custom pricing, monitor GPU usage, automate resource allocation, and implement rate-limiting policies. These features are intended to help neocloud service providers quickly commercialise their offerings while providing businesses with enhanced governance and cost control over AI workloads.
Embedded LLM collaborated with AMD in developing TokenVisor, responding to a growing demand in the neocloud sector for clarity around return on investment (ROI), billing, and streamlined workload management. TokenVisor is described as meeting these needs through its focus on control and oversight.
The system was engineered in consultation with members of the AMD GPU neocloud community and features a user interface designed to facilitate model deployment and capacity management. Early feedback from users in the sector has highlighted benefits such as quicker time-to-revenue and rapid onboarding following installation of GPU hardware and TokenVisor.
Commenting on the announcement, Ooi Ghee Leng, Chief Executive Officer of Embedded LLM, stated:
TokenVisor is the hypervisor for the AI Token era – unlocking decentralised GPU computing's potential requires tools as powerful and flexible as the hardware. Co-launched at Advancing AI 2025, an event that celebrates AI innovation and open-source collaboration, marks an important milestone for the AMD GPU neocloud community.
Mahesh Balasubramanian, Senior Director of Product Marketing, Data Center GPU Business at AMD, outlined the intended value to the neocloud sector:
TokenVisor brings powerful new capabilities to the AMD GPU neocloud ecosystem, helping providers efficiently manage and monetise LLM workloads.
TokenVisor features comprehensive support for both popular LLM and multi-modal models, and is supported by responsive technical assistance. These aspects have been identified as differentiators by early adopters seeking a faster return on investment from their AI infrastructure.
Embedded LLM, headquartered in Singapore, positions TokenVisor as part of the country's efforts to become a regional hub for AI and cloud infrastructure. The company also refers to TokenVisor as reflecting Singapore's strategy for AI sovereignty and regional technology leadership.
The platform's open architecture is intended to support both enterprise and cloud customers who own and operate AMD GPU resources, with Embedded LLM stating its commitment to industry standards for integration and optimisation of AMD-based AI stacks.
As interest grows in the commercial application of LLM and generative AI technologies, solutions like TokenVisor may play a role in enabling organisations to manage growing AI workloads while keeping costs and operational oversight in focus.
