NVIDIA commits up to $105 billion to support OpenAI’s Ohio data center, enhancing AI infrastructure for future developments.
NVIDIA Backs OpenAI’s Massive Ohio AI Data Center With Up to $105 Billion in Guarantees
AI News
NVIDIA is deepening its involvement in the infrastructure behind OpenAI’s next phase of artificial intelligence development, backing a massive data-center project in Ohio. The agreement places NVIDIA at the center of a major expansion of computing capacity intended to support OpenAI’s growing AI workloads.
The project is expected to reach approximately 8 gigawatts of capacity, making it one of the largest planned AI infrastructure developments. NVIDIA is also committing significant financial support to the project while supplying the computing technology that will power the facility. Together, these commitments show how closely advanced AI development is becoming connected to large-scale physical infrastructure.
The significance of the announcement extends beyond a single data center. Training and operating increasingly capable AI systems requires enormous amounts of computing power, electricity, networking, cooling, and physical space. As AI models and services continue to scale, access to this infrastructure is becoming a strategic factor in determining how quickly companies can develop and deploy new systems.
Why this matters: the AI race is increasingly becoming an infrastructure race. Building more capable AI systems requires not only better models and chips, but also the data centers, energy, networking, and capital needed to operate them at enormous scale.
The Deal
NVIDIA’s commitment to OpenAI’s Ohio data center is structured around several parts rather than a single direct payment. The chipmaker has agreed to provide up to $105 billion in guarantees connected to OpenAI’s 20-year lease of the facility, while also investing $1.5 billion in SB Energy, the SoftBank subsidiary developing the project.
The guarantee does not represent the full construction cost of the data center, nor does it mean NVIDIA is simply handing OpenAI $105 billion. Instead, the commitment is designed to support specific obligations, including portions of lease and power payments and the value of the underlying facility. OpenAI remains responsible for paying the rent under the agreement.
NVIDIA will also serve as the exclusive chip provider for the planned facility in Pike County, Ohio. The site is expected to scale to as much as 8 gigawatts of capacity, with the first 800 megawatts targeted to come online in 2028. That makes the arrangement strategically important for both sides: OpenAI gains access to a large future computing environment, while NVIDIA secures a major long-term deployment opportunity for its AI computing platforms.
The key distinction: NVIDIA’s $105 billion commitment is a financial guarantee supporting specific project obligations, not a $105 billion cash investment in the data center. The arrangement combines financing, infrastructure, power commitments, and NVIDIA computing hardware into one long-term AI infrastructure strategy.
The Compute Challenge
The scale of the Ohio project becomes easier to understand when AI is viewed as a computing-intensive technology. Modern AI systems depend on large amounts of accelerated computing not only during model training, but also when millions of users interact with those systems in real time. As models become more capable and AI applications become more widely used, the amount of infrastructure required to train and serve them grows with it.
01
Training a frontier model requires large clusters of specialized accelerators working together for extended periods. Increasing model complexity, larger datasets, and more sophisticated training methods can all increase the demand for compute.
02
Once a model is deployed, computing power is needed every time the system generates an answer, analyzes information, writes code, or performs another AI task. Serving users at large scale therefore creates a continuous demand for compute.
03
Newer AI systems increasingly handle longer reasoning processes, multimodal inputs, coding tasks, agents, and other workloads that can require substantially more computation than a simple text response.
OpenAI has previously described the relationship between available compute and its ability to develop and deploy advanced AI systems. In a February 2026 submission to the U.S. House Select Committee on Strategic Competition, the company said its available compute had increased from 0.2 gigawatts in 2023 to approximately 1.9 gigawatts in 2025. The company has also described large-scale data centers as part of the physical foundation required to train and deploy frontier AI systems.
Three layers of AI compute
| Workload | What it requires | Why it matters |
|---|---|---|
| Training | Large-scale accelerator clusters and sustained compute capacity. | Used to develop and improve increasingly capable models. |
| Inference | Computing capacity available whenever users request AI output. | Determines how reliably AI services can operate at scale. |
| Reasoning & Agents | Additional computation for multi-step and more complex tasks. | Enables AI systems to perform increasingly sophisticated work. |
The bigger picture: a frontier AI company does not need computing power only when it trains a new model. It needs infrastructure throughout the entire lifecycle of AI development—from experimentation and training to deployment, inference, evaluation, and increasingly complex agentic workloads.
AI Infrastructure
The scale of the planned Ohio facility is what makes the project particularly significant. OpenAI and its partners are developing an infrastructure campus designed to ultimately support approximately 8 gigawatts of capacity. For an AI company, that scale represents far more than additional server space: it means creating a physical foundation capable of supporting large volumes of accelerated computing for years to come.
The first phase is expected to bring approximately 800 megawatts online in 2028, while the broader campus is planned to expand over time. Reaching that level requires coordinated development across computing hardware, electrical systems, transmission infrastructure, cooling, networking, construction, and operations. In other words, the project is effectively an integrated AI infrastructure system rather than a conventional server facility.
Project at a glance
| Infrastructure element | Planned scale / role | Strategic importance |
|---|---|---|
| Overall capacity | Approximately 8 GW when fully developed | Provides a large long-term foundation for AI compute. |
| Initial capacity | Approximately 800 MW targeted for 2028 | Establishes the first major phase of the computing campus. |
| Computing hardware | NVIDIA computing systems | Supplies the accelerated compute required for large-scale AI workloads. |
| Energy infrastructure | Dedicated power and grid infrastructure | Enables the sustained electricity supply required by energy-intensive AI systems. |
Compute
Electricity alone does not create useful AI capacity. The facility must combine power with large clusters of accelerators, high-speed networking, storage, cooling, and software infrastructure. The result is a tightly integrated computing environment capable of handling demanding AI workloads.
Scale
The phased approach is also important. Rather than treating the project as a single deployment, the planned campus can expand its computing footprint over time. That gives OpenAI and its partners a physical platform for increasing capacity as AI workloads continue to grow.
This scale also explains why AI infrastructure projects increasingly involve organizations outside the traditional technology sector. Large AI campuses depend on utilities, energy developers, construction firms, equipment suppliers, financing partners, and local governments. Building the computing layer is therefore only one part of a much larger infrastructure challenge.
The strategic takeaway: the competitive advantage in advanced AI is increasingly influenced by how quickly companies can secure the physical infrastructure needed to turn computing hardware into reliable, scalable AI capacity.
Industry Shift
NVIDIA has built its position in the AI industry around accelerated computing, but its involvement in the Ohio project illustrates a broader evolution. The company is increasingly connected to the infrastructure surrounding AI compute, including the systems, financing arrangements, networking technologies, and data-center capacity required to deploy large-scale AI workloads.
This distinction matters because the value of an AI accelerator depends on the infrastructure around it. A powerful GPU cannot operate in isolation. It requires electricity, cooling, high-speed networking, storage, software, physical space, and a data-center environment designed to support dense computing workloads.
From component to platform
The modern AI data center combines multiple layers of technology. NVIDIA’s position allows it to participate across several of these layers rather than operating only as a conventional semiconductor supplier.
Strategic shift
Hardware → Infrastructure
Layer 01
GPUs and accelerated computing systems process demanding AI workloads.
Layer 02
High-speed interconnects allow large numbers of processors to work together efficiently.
Layer 03
Physical facilities provide the power, cooling, storage, and space required by AI systems.
Layer 04
Financial, energy, software, and infrastructure partners make large-scale deployment possible.
Traditional model
Hardware-centricEmerging model
Infrastructure-centricWhy the distinction matters
Compute
Processing capacity determines how much AI work can be performed.
Power
Reliable electricity is essential for operating dense AI clusters continuously.
Infrastructure
Networking, cooling, storage, and facilities turn hardware into a functioning AI platform.
NVIDIA’s involvement in the OpenAI project therefore highlights a broader change in the AI industry. As computing demand grows, companies that control or influence critical infrastructure can gain strategic importance alongside the companies developing the models themselves.
The strategic takeaway: the future of AI infrastructure is unlikely to be defined by chips alone. The companies capable of connecting compute, networking, energy, data centers, software, and capital will increasingly influence how quickly advanced AI can scale.
Economics of AI
The scale of the Ohio project also highlights a fundamental economic reality of frontier artificial intelligence: developing and operating increasingly capable AI systems requires enormous amounts of capital. The cost is no longer limited to research teams and model training. Companies must also finance computing hardware, data centers, electricity, networking, cooling systems, maintenance, and the infrastructure required to operate these systems over many years.
This creates a different economic model from traditional software development. A conventional software product can often reach millions of users without requiring a proportional expansion of physical infrastructure. AI services are different because every model request ultimately consumes computing resources. As usage increases, companies must continuously expand the capacity available to serve those requests.
The cost stack
The economics of advanced AI can be understood as a connected infrastructure stack. A large investment in one layer creates requirements in several others.
01 · Hardware
GPUs and specialized computing systems form the processing layer required for training and serving advanced AI models.
02 · Facilities
Computing hardware requires specialized buildings, networking, storage, cooling, security, and physical infrastructure.
03 · Energy
Large AI clusters require substantial and reliable electricity, making energy infrastructure an increasingly important part of AI economics.
04 · Operations
Running AI infrastructure also involves energy, maintenance, networking, staffing, hardware replacement, and other recurring expenses.
Capital intensity
As AI companies increase their available computing capacity, they must also invest in the physical systems that allow that compute to operate. This makes scaling frontier AI a capital-intensive process rather than a purely software-driven expansion.
Long-term economics
Large infrastructure projects are typically designed around expected long-term demand. For AI companies, that creates pressure to increase model usage, enterprise adoption, and AI-powered products sufficiently to make massive infrastructure investments economically sustainable.
From model to infrastructure
01
Develop increasingly capable models and AI systems.
02
Acquire the computing capacity required for training and inference.
03
Deliver AI capabilities through products and services at scale.
04
Generate sufficient economic value to support continued investment.
The NVIDIA–OpenAI arrangement therefore provides a useful example of how the economics of AI are changing. The development of frontier models is increasingly intertwined with long-term infrastructure commitments, and the ability to secure that infrastructure can become just as important as the ability to develop the underlying model technology.
The economic takeaway: frontier AI is becoming an infrastructure-heavy industry. The companies that can efficiently combine capital, compute, energy, and AI products will be better positioned to sustain the enormous costs associated with scaling advanced artificial intelligence.
References & Further Reading
The discussion in this section draws on publicly available information about artificial intelligence infrastructure, computing capacity, semiconductor investment, data-center development, and the economics of scaling frontier AI systems. The following sources provide additional context for readers who want to explore the subject in greater depth.
Industry information covering accelerated computing, AI infrastructure, data-center systems, and the technologies required to operate large-scale artificial intelligence workloads.
Research and technical material providing context on the increasing computational requirements associated with developing and deploying advanced AI systems.
Analysis of electricity consumption, data-center expansion, and the growing relationship between digital infrastructure and global energy systems.
Additional analysis of capital expenditure, computing capacity, infrastructure investment, and the long-term economics of large-scale AI deployment.
Editorial note
Infrastructure costs, investment figures, and AI industry developments can change rapidly. Readers should consult the original sources for the latest figures, announcements, and technical details.
Continue exploring the economics of AI — from computing hardware and energy infrastructure to the systems that turn frontier models into real-world products.
Editor in Chief
Software engineer and full-stack developer building modern digital experiences, products, and ideas.
codewithtabish.comReady to do everything better? Get daily tips, tricks, and tech guides from our expert team.
By clicking Sign Up, you confirm you are 16+ and agree to our Terms of Service and Privacy Policy.
Have fun. Be respectful. Feel free to criticize ideas, but not people.