Fusemachines (NASDAQ: FUSE) has expanded the computing infrastructure available for its AI Twin development program through an arrangement with IBM Platinum Business Partner modulAIre. The company said modulAIre will provide access to its AI-optimized IBM Fusion infrastructure at nominal cost under the companies’ existing strategic partnership.
The added capacity is intended to let Fusemachines run more model experiments, test architectures and conduct evaluations as it develops AI Twin technology. The company introduced its AI Twin concept on Sept. 17 as an intelligent digital representation that could understand an individual’s knowledge, responsibilities, preferences and organizational context, then participate in authorized work on that person’s behalf.
For the engineering teams building those systems, the immediate benefit is additional infrastructure for iterative development. Fusemachines said its researchers will use the environment to experiment with and improve models and systems underlying AI Twin. That work can involve large language models, agentic reasoning, voice and conversational AI, memory, knowledge retrieval, enterprise context and real-time human-AI interaction, according to the company.
The infrastructure is supplied by modulAIre, which has customized IBM Fusion hardware for AI optimization, scalability and storage-to-inference operations. The companies said the platform integrates with technologies including Red Hat OpenShift, providing an environment for developing, deploying and managing AI applications and models. IBM (NYSE: IBM) is the maker of the Fusion infrastructure referenced in the arrangement.
This is an infrastructure-access expansion, not a general commercial launch of a new IBM or Fusemachines product. The announcement does not specify the amount of compute capacity, the number of GPUs, service-level commitments or a timetable for AI Twin availability. It also does not establish how the proposed AI Twin systems will be governed, which enterprise applications they will connect to or what controls will apply to authorized work performed on a user’s behalf.
Even so, the arrangement shows how AI product development is becoming tied to the operational characteristics of the platforms used to train, evaluate and deploy models. Access to a configurable environment can help an engineering team test different architectures and increase the pace of iteration without first building or procuring every component of the underlying stack. It can also allow experiments and evaluations to be repeated against a stable development environment, although the company did not describe its evaluation methodology. That benefit is contingent on the infrastructure and tools actually matching the workloads an organization needs to run.
modulAIre said its platform is designed to scale compute, storage and GPU resources as requirements change. For platform teams, that architecture is relevant because model-development work can place changing demands on accelerated computing, storage and deployment tools. The release does not describe scheduling, automation or observability features in detail, so those operating characteristics should be evaluated separately by prospective users.
Fusemachines said the expanded access will complement its existing research and development resources and support its broader enterprise AI products and agentic AI offerings. In IT-operations terms, the story is about the infrastructure layer behind those products: a partner-provided environment intended to simplify AI workload deployment and management while giving developers more capacity to evaluate and refine models.