The next stage of artificial intelligence will not be defined by chatbots alone. It will be shaped by enormous data centres capable of producing intelligence at an industrial scale—and NVIDIA and OpenAI want to build the infrastructure first.
When ChatGPT appeared, most people experienced artificial intelligence as software. They typed a question into a browser and received an answer within seconds. The physical machinery behind that response remained largely invisible.
That is now changing.
The expansion of advanced AI requires enormous quantities of processors, electricity, networking equipment and data-centre capacity. NVIDIA and OpenAI are working together to build this foundation, creating a relationship that extends far beyond the conventional purchase of computer chips.
OpenAI creates the models and services generating demand for computing. NVIDIA supplies the accelerated-computing platform that makes those systems possible. Together, they are helping introduce what the technology industry increasingly calls the “AI factory.”
What Is an AI Factory?
A traditional factory converts raw materials into physical products. An AI factory converts data and electricity into model training, predictions, generated content and automated decisions.
Its production can be measured in tokens—the pieces of information processed by AI models. Every answer, software instruction, image or completed task requires computing capacity.
Modern AI facilities contain far more than rows of individual GPUs. They combine processors, CPUs, high-speed networking, storage, cooling and specialised software into systems designed to operate continuously.
NVIDIA is positioning itself as the provider of this complete platform. OpenAI is one of its largest potential users.
The 10-Gigawatt Plan
In September 2025, NVIDIA and OpenAI announced a letter of intent to deploy at least 10 gigawatts of NVIDIA systems for OpenAI’s future infrastructure. The companies said the planned capacity could involve millions of GPUs.
NVIDIA also stated that it intended to invest up to $100 billion in OpenAI progressively as new gigawatts were deployed. The investment was presented as a phased commitment connected to infrastructure construction rather than an immediate payment.
The first gigawatt was targeted for deployment during the second half of 2026 using NVIDIA’s Vera Rubin platform.
Reaching the target requires suitable land, electricity generation, transmission systems, substations, cooling equipment, buildings and financing. The processors are only one part of the project.
This makes the partnership unusual. A software company and a processor company are effectively coordinating an infrastructure expansion comparable to a major industrial-development programme.
Why OpenAI Cannot Grow on Software Alone
Training an advanced AI model involves processing enormous datasets across large clusters of processors. Once the model is available, serving users creates a second requirement known as inference.
Inference happens whenever a model responds to a request. As AI becomes integrated into search, programming, research, customer service and business operations, this continuing workload can become larger than the original training process.
OpenAI therefore needs capacity for both experimentation and everyday use.
In February 2026, the company said it was expanding its NVIDIA collaboration through three gigawatts of dedicated inference infrastructure and two gigawatts of training capacity based on Vera Rubin systems.
This would supplement NVIDIA Hopper and Blackwell systems already operating through Microsoft, Oracle Cloud Infrastructure and CoreWeave.
More computing capacity can allow OpenAI to develop increasingly capable models, support additional users and make its services faster and more reliable. It may also provide room for AI agents performing longer, more complicated tasks instead of producing one short answer.
NVIDIA Is Selling a System, Not Just a Processor
NVIDIA’s competitive advantage comes from the combination of hardware and software surrounding its GPUs.
Its Vera Rubin platform includes specialised processors, CPUs, networking switches, data-processing units and storage technology. These components are designed to work together as one large computing system.
NVIDIA announced in March 2026 that the platform’s seven principal chips were in full production. The company says Vera Rubin supports multiple stages of AI development, including pretraining, post-training, reasoning and real-time agentic inference.
For OpenAI, close cooperation with NVIDIA provides an opportunity to prepare models and infrastructure software for future hardware.
For NVIDIA, access to OpenAI’s requirements offers valuable information about the workloads frontier AI systems will create. The companies have described this process as coordinating their respective technology roadmaps.
Ohio Becomes Part of the AI Map
The PORTS-Pike Technology Campus in Ohio provides a physical example of this expansion.
In August 2026, OpenAI announced that it had entered an agreement to secure approximately eight gigawatts of IT capacity at the campus.
NVIDIA was named as the project’s exclusive provider of AI-compute infrastructure, while SB Energy and the United States Department of Energy are also involved.
The project demonstrates how AI investment can reshape regions far from traditional technology centres. A large campus can create demand for construction workers, engineers, electricians, suppliers and permanent technical employees.
It can also create pressure on local resources. Residents and officials will expect clear answers about electricity costs, water consumption, noise and environmental impact.
OpenAI said it would pay project-specific energy and infrastructure expenses and added $40 million to SB Energy’s existing $40 million community-benefits fund.
Large AI projects will increasingly be judged not only by their computing performance, but also by their relationship with the communities hosting them.
The Business Logic for NVIDIA
The expansion of OpenAI supports demand for NVIDIA’s processors, networking and software. Its effect is already visible in the company’s financial results.
NVIDIA reported quarterly revenue of $96.2 billion for its second quarter of fiscal 2027, representing growth of 106% from the previous year. Data Center revenue reached $89 billion, increasing by 117% year over year.
The company is now participating more actively in the creation of the market it serves. Instead of simply waiting for customers to finance and construct data centres, NVIDIA is investing in developers and supporting infrastructure projects.
This can create a powerful cycle. More capacity allows customers to deploy additional AI services, which can increase computing demand and generate further orders for NVIDIA systems.
It also creates financial risk. The cycle remains sustainable only if end users continue paying enough for AI products to justify the infrastructure behind them.
OpenAI Is Avoiding Complete Dependence
Despite the size of the NVIDIA relationship, OpenAI continues working with multiple infrastructure partners.
Its wider portfolio includes Microsoft, AWS, Oracle, AMD, Broadcom, Cerebras, CoreWeave, SoftBank and SB Energy. These companies contribute processors, cloud services, networking, capital, energy or construction capacity.
This approach is partly practical. OpenAI expects to need more computing power than one provider may be able to deliver quickly.
Multiple partnerships can also reduce supply risk and create competition involving price and performance.
NVIDIA remains central, but it cannot assume that every future OpenAI workload will automatically use its technology. Alternative processors and more efficient model designs will continue challenging its position.
The Questions Nobody Can Answer Yet
The infrastructure plan depends on several assumptions that have not been fully tested.
Will demand for paid AI services grow quickly enough? Can data centres secure sufficient electricity without increasing local costs? Will processors become substantially more efficient before current projects are completed? Can companies earn sustainable revenue from AI agents and professional applications?
There are also regulatory questions involving market concentration, energy consumption and increasingly connected investment relationships.
NVIDIA may simultaneously be a supplier, infrastructure partner and investor in companies purchasing its technology.
That structure can accelerate construction, but it may also make it more difficult to separate independent customer demand from growth supported by supplier financing.
The Industrial Phase of Artificial Intelligence
The partnership between NVIDIA and OpenAI demonstrates that artificial intelligence has entered a new stage.
The first stage focused on research. The second brought generative AI to consumers. The emerging third stage is about building enough physical infrastructure to make AI a permanent part of the global economy.
OpenAI needs computing capacity to expand its models and products. NVIDIA needs successful AI applications to justify continued investment in increasingly powerful systems.
Their ambitions are closely connected, but success is not guaranteed. The alliance must transform electricity, processors and billions of dollars into services that produce lasting value.
If it succeeds, the AI factory may become as important to the twenty-first century as the semiconductor plant, power station and cloud data centre were to earlier technological eras.
NVIDIA and OpenAI are betting that demand will arrive—and that the infrastructure they are constructing will become the foundation on which the next generation of digital products operates.
This article reflects publicly available information as of September 3, 2026. It is for informational purposes only and does not constitute financial or investment advice.
