As the field of artificial intelligence (AI) transitions from model development to production inference, the demand for computational power is accelerating rapidly. This surge in demand is steering us toward a future where AI factories operate continuously, generating tokens at an unprecedented scale. To accommodate this shift, businesses now require large-scale, multi-tenant accelerated computing solutions that can come online swiftly, maintain high utilization, and support the economic models of token-scale AI services.
Historically, emerging AI companies have faced significant challenges in accessing the capital-intensive infrastructure needed to fuel their innovations. Long-term commitments rarely provided sufficient incentives for financial backers to unlock funding for essential computing resources. This lack of accessible infrastructure has been a substantial barrier to entry for many startups and model builders.
In response to these challenges, NVIDIA is rolling out a transformative business model aimed at expanding compute access across the rapidly growing AI ecosystem, including startups, enterprises, research institutions, and regional players. This innovative approach allows AI clouds to procure NVIDIA’s cutting-edge infrastructure tailored for AI-native, enterprise, and independent software vendor (ISV) clients, via a revenue-sharing and credit-support structure designed for economic alignment.
NVIDIA’s New Business Model
Under this new model, AI clouds can sell NVIDIA-powered cloud services, fostering a dual revenue stream. NVIDIA will benefit from standard product sales while also receiving a portion of the cloud revenue generated from supported capacities. This strategic partnership is expected to expedite the adoption of NVIDIA technologies within high-growth AI sectors, providing the company with a recurring earnings stream linked to actual usage.
For model builders, inference providers, agent platforms, and enterprises aiming to scale their AI capabilities, this model translates to quicker access to a full-stack accelerated computing setup without the need to navigate the complexities related to site selection, power procurement, construction, and hardware initialization.
More Read
Building AI Factory Capacity
This initiative is already in motion, with various AI cloud companies working to construct DSX AI factories designed to support diverse customers and workloads spread across multiple regions. Notable examples include Sharon AI and Firmus Technologies, which are among the first companies collaborating with NVIDIA on this pioneering business model.
Sharon AI is in the process of deploying up to 40,000 NVIDIA Grace Blackwell GB300 GPUs. According to James Manning, cofounder and CEO of Sharon AI, “This strategic collaboration with NVIDIA marks a pivotal moment in Sharon AI’s mission to deliver sovereign, large-scale AI compute infrastructure.”
Meanwhile, Firmus is developing a DSX AI factory campus in Batam, Indonesia. This ambitious project is expected to scale up to 360 megawatts while integrating up to 170,000 NVIDIA GPUs into its infrastructure. Tim Rosenfield, co-CEO of Firmus Technologies, emphasized the importance of scalable and energy-efficient computing: “AI-native companies need access to scalable, energy- and cost-efficient compute infrastructure to compete globally.”
Market Demand for AI Compute
AI-native companies such as Baseten, Fireworks AI, and Together AI illustrate the direction in which compute demand is headed, emphasizing the need for immediate access to AI cloud capacity for a variety of tasks — from model training and post-training processes to fine-tuning and high-volume agentic inference aimed at developers, digital natives, and enterprises leveraging AI.
Their client base is in search of reliable and scalable access to large-scale NVIDIA-accelerated computing, especially as their usage continues to expand. Additionally, these companies require commercial flexibility as they transition their products from pilot phases to full production.
To secure the necessary compute capacity and advance AI model development, organizations are encouraged to connect with Sharon AI and Firmus.
For additional information about NVIDIA’s collaboration with AI clouds and the implications for future AI infrastructure, be sure to explore NVIDIA Cloud Partners and AI factories.
Inspired by: Source
