In a landmark move set to redefine the scale of artificial intelligence, NVIDIA and OpenAI have announced a strategic partnership to build an unprecedented amount of AI-specific computing infrastructure. The collaboration, detailed in a joint letter of intent, involves OpenAI deploying at least 10 gigawatts of data centers powered by NVIDIA systems, an undertaking that represents millions of GPUs. To finance the enormous capital expense of data centers and power generation, NVIDIA will invest up to $100 billion in OpenAI, with funds released progressively as each gigawatt of capacity comes online.
This multi-year initiative aims to create the computational foundation for OpenAI’s next generation of AI models, accelerating the pursuit of artificial general intelligence (AGI). The project dwarfs previous infrastructure build-outs in the tech industry and signals a new era of hyperscale investment focused exclusively on AI. The first phase of the massive deployment is scheduled to begin operations in the second half of 2026, leveraging NVIDIA’s next-generation platform to power what both companies are calling the future of the AI economy. “This investment and infrastructure partnership mark the next leap forward,” said Jensen Huang, founder and CEO of NVIDIA. “Deploying 10 gigawatts to power the next era of intelligence.”
An Unprecedented Scale of Power
The core of the agreement is the commitment to 10 gigawatts of power capacity dedicated solely to AI computation. To put this figure in perspective, 10 gigawatts is the approximate energy required to power 7.5 million homes simultaneously. This massive energy budget is a direct response to the escalating demands of training and running frontier AI models, which now require computational resources that are straining the limits of the global power grid. The new infrastructure will be built to handle both the training of increasingly complex models and the inferencing operations needed to serve them to hundreds of millions of users.
The collaboration is a natural extension of a decade-long relationship between the two AI pioneers. This partnership has evolved from NVIDIA’s first DGX supercomputer to the systems that powered the breakthrough of ChatGPT. OpenAI’s President, Greg Brockman, noted the history, stating, “We’ve been working closely with NVIDIA since the early days of OpenAI. We’re excited to deploy 10 gigawatts of compute with NVIDIA to push back the frontier of intelligence.” The project will be essential for OpenAI to serve its rapidly growing user base, which has expanded to over 700 million weekly active users.
The Next-Generation Technological Foundation
The Vera Rubin Platform
The first wave of these new AI data centers will be built upon NVIDIA’s yet-to-be-released “Vera Rubin” platform. Named after the pioneering astrophysicist, the Rubin architecture is the successor to NVIDIA’s Blackwell platform and represents a significant leap in specialized processor design. The platform will feature not only a next-generation GPU but also a new class of processor called a CPX, which is purpose-built for massive-context AI. This new hardware is designed to excel at tasks involving models that must process and reason across millions of tokens of information at once, a critical capability for applications like high-quality generative video and advanced coding assistants. The Rubin family of chips will be manufactured by TSMC on a 3-nanometer process and will utilize next-generation HBM4 memory for faster data access.
Phased Deployment and Roadmap
The deployment will be a multi-year effort, with the first gigawatt of capacity scheduled to come online in the latter half of 2026. This initial phase will be among the first large-scale deployments of the Vera Rubin platform. The partnership goes beyond a simple hardware purchase; the two companies have agreed to co-optimize their future roadmaps. This ensures that OpenAI’s future AI models and software are tightly integrated and fine-tuned for NVIDIA’s upcoming hardware and software, creating a highly efficient and powerful full-stack ecosystem. This collaboration complements OpenAI’s ongoing work with partners like Microsoft, Oracle, and the Stargate initiative, creating a diverse and robust infrastructure network.
Strategic Vision for the AI Economy
Both chief executives framed the partnership in transformative economic terms. “Everything starts with compute,” said Sam Altman, CEO of OpenAI. “Compute infrastructure will be the basis for the economy of the future, and we will utilize what we’re building with NVIDIA to both create new AI breakthroughs and empower people and businesses with them at scale.” The deal solidifies a critical supply chain for OpenAI, securing access to the most sought-after resource in the technology world: accelerated compute. By partnering directly with NVIDIA, OpenAI ensures it has the dedicated “AI factories” needed to maintain its leadership position and pursue its long-term mission of developing safe and beneficial AGI.
Confronting Gigawatt-Scale Challenges
The Power and Cooling Bottleneck
Building data centers on a gigawatt scale presents monumental engineering and logistical challenges that go far beyond securing computer chips. The primary bottleneck is power availability. Many national electrical grids were not designed for such concentrated, high-volume power draws and lack the infrastructure to support them. Connecting a gigawatt-scale campus to the grid often requires constructing new substations and high-voltage transmission lines, a process that can take years and face significant regulatory hurdles.
Furthermore, AI systems consume power at extreme densities. A traditional data center rack might draw 5–10 kilowatts, whereas a rack of modern AI servers can demand over 100 kW. This density generates an immense amount of heat that cannot be managed by traditional air-cooling methods. Consequently, these new facilities must be designed from the ground up with advanced liquid-cooling solutions, such as direct-to-chip or immersion cooling, which adds another layer of complexity and cost to their construction and operation.
Environmental and Logistical Hurdles
The enormous energy and water requirements for cooling these facilities raise significant environmental concerns. Sourcing gigawatts of clean, reliable energy is a critical challenge, as renewable sources like wind and solar are often intermittent. This will likely require a combination of power purchase agreements with renewable energy providers, the development of on-site power generation, and the use of large-scale battery storage systems to ensure uninterrupted operation. Beyond power, these massive construction projects must navigate a complex web of environmental permits, land acquisition issues, and potential community opposition, all of which can introduce delays.
