Firmus Technologies has entered into a multi-year agreement with Thinking Machines Lab, an AI research and product company, positioning Thinking Machines Lab as an anchor customer for the Firmus AI-Cloud Region in Asia. The collaboration will support the training and deployment of next-generation AI models on Firmus AI Factory infrastructure powered by NVIDIA GB300 systems. The announcement, made from San Francisco and Sydney, highlights increasing demand for specialized AI infrastructure that integrates compute, networking, and cooling.
Strategic Anchor Customer Agreement
Under the agreement, Thinking Machines Lab will become the anchor customer for the new Firmus AI-Cloud Region in Asia. The company will train and deploy its next-generation AI models on Firmus AI Factory systems featuring NVIDIA GB300 technology. This arrangement provides Thinking Machines Lab with dedicated access to infrastructure designed for frontier AI workloads.
Thinking Machines Lab Profile
Thinking Machines Lab was founded in 2025 with a mission to build AI that extends human will and judgment. The company develops strong models and tools that allow people to fine tune models with their unique knowledge and adapt them as that knowledge changes. It also focuses on interaction models that support live, multimodal engagement, enabling users to interrupt, correct, and change their minds aloud.
Purpose-Built Infrastructure
The Firmus AI-Cloud Region is built on the company's AI Factory architecture, which combines networking, compute, and software into a unified platform for frontier AI workloads. The Firmus HyperCube integrates liquid cooling, power, and mechanical systems to reduce cooling and energy demand. This design reflects a Model-to-Grid approach that recognizes networking, cooling, and power as essential to computational efficiency.
Executive Perspectives
Trent Viengsone, Vice President of North America at Firmus Technologies, said the company is delighted to welcome Thinking Machines Lab as an anchor customer for its AI-Cloud Region in Asia. He noted that the lab's work sits at the leading edge of AI research and development, and that the agreement reflects rising demand for purpose-built AI infrastructure. Viengsone added that Firmus looks forward to helping Thinking Machines Lab scale its ambitions on Firmus AI Factory infrastructure.
Model-to-Grid Efficiency
Dr. Daniel Kearney, Chief Technology Officer at Firmus Technologies, emphasized that the Model-to-Grid approach treats networking, cooling, and power as equally important as compute for efficiency. By designing these systems together, Firmus can provide Thinking Machines Lab with additional capacity to train, deploy, and scale advanced AI systems. This integrated philosophy supports the resilience and performance requirements of next-generation artificial intelligence.
Expanding AI Capacity in Asia
The selection of an anchor customer for the Asia region signals a strategic expansion of Firmus Technologies' AI-Cloud footprint. Asia has become a critical market for AI development, with growing demand for high-performance computing and energy-efficient data center infrastructure. The new region is intended to serve customers that require robust, scalable, and sustainable AI training environments.
Industry Context and Demand
The agreement highlights a broader industry trend toward purpose-built AI infrastructure that balances performance with environmental responsibility. As frontier models grow more complex, organizations are seeking integrated platforms that address power, cooling, and networking constraints. This deal positions Firmus Technologies and Thinking Machines Lab within a competitive landscape focused on efficient AI scaling.
Firmus Technologies and Thinking Machines Lab have established a multi-year collaboration centered on advanced AI infrastructure in Asia. The partnership combines Thinking Machines Lab's research mission with Firmus Technologies' AI Factory architecture and NVIDIA GB300 systems. It underscores how coordinated design across compute, networking, and cooling can support the next generation of AI models.