Nvidia AI Infrastructure Push: Land, Power and Data Centres

Nvidia has made billions of dollars selling the chips that power the Artificial Intelligence boom. But according to CEO Jensen Huang, increasingly powerful AI models need more than just GPUs.

They need enormous amounts of land, electricity and data-centre Infrastructure to turn computing hardware into working AI systems.

That is why Nvidia is expanding its role beyond supplying chips. The company now plans to help AI companies secure the physical infrastructure required to build and operate large-scale AI computing facilities.

In a post on X titled “Securing the Infrastructure of Intelligence,” Huang explained that AI companies are increasingly facing a shortage of computing capacity rather than a lack of demand or algorithms.

“AI factories require a full stack of critical resources,” Huang wrote, referring to advanced chips, packaging, memory and networking, as well as land, power and “shell”.

Here, “shell” refers to the physical data-centre infrastructure required to house and operate AI computing systems.

Nvidia wants to solve AI’s infrastructure problem

The AI industry has been expanding rapidly, with companies such as OpenAI, Anthropic and Meta Investing heavily in computing capacity.

Large language models require huge amounts of computing power both during training and when serving users. As AI adoption grows, securing enough computing capacity is becoming an increasingly important challenge.

Huang says this is becoming a fundamental constraint on AI companies.

According to him, their growth is increasingly being limited not by algorithms or customer demand, but by the availability of compute.

That creates a new opportunity for Nvidia.

Rather than simply selling the GPUs that go into AI data centres, Nvidia wants to help ensure that the land, electricity and physical facilities needed to deploy those GPUs are also available.

The shift effectively puts Nvidia closer to the infrastructure layer of the AI industry.

Nvidia is leasing infrastructure to AI companies

Nvidia’s approach is not simply to give away land and electricity to AI companies.

Instead, the company is becoming involved in securing and supporting the infrastructure through leasing arrangements and partnerships.

One of the first major examples is its partnership with SB Energy at the PORTS-Pike Technology Campus in Portsmouth, Ohio.

The site is planned to host Nvidia AI factories, with OpenAI as the tenant under a 20-year lease.

Under the arrangement, OpenAI will build and operate the AI factory using Nvidia’s full-stack DSX AI factory platform.

That platform includes Nvidia’s:

• GPUs
• CPUs
• Networking technology
• Infrastructure software

The project is therefore intended to function as a large-scale AI computing facility rather than simply a conventional data centre filled with servers.

The massive 4.25GW AI factory plan

The scale of the planned project illustrates just how much infrastructure modern AI systems can require.

The initial plan calls for an AI factory at PORTS-Pike with 4.25 gigawatts of capacity.

That represents a huge amount of computing and electrical infrastructure dedicated to AI operations.

Nvidia says the site could potentially secure an additional 3.75GW, which would bring its potential capacity to 8GW.

The first phase is expected to come online between 2028 and 2030.

The facility will use Nvidia’s DSX platform and its broader combination of GPUs, CPUs, networking and infrastructure software.

This means Nvidia is not simply supplying individual components. It is helping establish a broader infrastructure ecosystem designed specifically around its AI computing technology.

OpenAI is the first major customer

OpenAI will be the first customer associated with the PORTS-Pike project.

However, Nvidia is not simply funding the entire facility and handing it over to OpenAI.

According to Huang, Nvidia will support defined portions of the lease and power payments, while also providing a specified residual-value commitment.

OpenAI will remain the tenant and will pay the lease.

The arrangement allows Nvidia to help secure the physical resources required for OpenAI’s growing computing needs without taking on the entire cost of the facility itself.

It also gives Nvidia a deeper role in the infrastructure supporting one of the world’s largest AI companies.

Nvidia is investing $1.5 billion in SB Energy

As part of the arrangement, Nvidia is also investing $1.5 billion in SB Energy.

The investment adds another layer to Nvidia’s strategy.

Rather than viewing AI infrastructure as something that customers will independently build around Nvidia’s chips, the company appears to be looking at the entire AI factory ecosystem as part of the long-term opportunity.

That includes securing suitable locations, ensuring sufficient power availability and building facilities capable of supporting successive generations of Nvidia hardware.

Why land and electricity have become critical for AI

The AI Boom is creating an unusual infrastructure challenge.

The performance of AI systems continues to improve as companies deploy more powerful chips and larger computing clusters. But those systems require enormous physical facilities and substantial amounts of electricity.

A company can therefore have access to advanced AI models and demand from customers but still struggle to expand if it cannot secure enough computing infrastructure.

This is the problem Huang says Nvidia is trying to address.

The bottleneck is no longer simply whether companies can buy enough chips. They also need somewhere to install those chips and enough electricity to run them.

That makes land, power and data-centre construction increasingly important components of the AI race.

Nvidia wants AI factories to last through multiple chip generations

Another important part of the strategy is the expected lifespan of the infrastructure.

AI hardware is advancing quickly, meaning today’s leading GPU systems can eventually be replaced by newer and more powerful generations.

Nvidia’s strategy is designed around securing sites that can continue hosting its AI systems as the hardware evolves.

In other words, the same physical facility could potentially be upgraded repeatedly rather than being built for only one generation of Nvidia chips.

That could make expensive infrastructure investments more useful over the long term.

This changes Nvidia’s role in the AI boom

For years, Nvidia’s position in the AI industry has largely been defined by its GPUs.

Companies building AI models need Nvidia’s computing hardware, while cloud providers and data-centre operators deploy those chips at enormous scale.

The new strategy expands that relationship.

Nvidia is now moving further into the infrastructure required to deploy its own hardware.

That does not mean Nvidia is becoming a traditional real-estate or electricity company. Instead, the company is attempting to make sure that critical infrastructure does not become a bottleneck for the demand for its chips.

The strategy could also strengthen Nvidia’s ecosystem by making it easier for major AI customers to deploy large quantities of Nvidia hardware.

The bigger AI infrastructure race

The PORTS-Pike project highlights how the next phase of the AI boom may depend as much on physical infrastructure as on advances in software.

Companies can develop better AI models, but those models still require computing facilities capable of running them.

That creates a chain of requirements: chips need servers, servers need data centres, data centres need land and power, and all of it needs to be connected through high-speed networking infrastructure.

Nvidia already controls a major part of the chip and networking layers. Its new infrastructure strategy takes the company closer to the physical foundation supporting the entire AI ecosystem.

Nvidia’s bet goes beyond selling GPUs

Nvidia’s move represents a significant evolution in its AI Business.

The company has benefited enormously from demand for GPUs as OpenAI, Meta, Anthropic and other AI companies race to build increasingly capable systems.

But Huang’s message is that simply producing more chips will not solve the industry’s next major bottleneck.

AI companies also need the physical capacity to deploy them.

By supporting leases and power payments, investing in infrastructure partners such as SB Energy and helping establish massive AI factories, Nvidia is attempting to address that constraint directly.

The 4.25GW PORTS-Pike project, with the potential to reach 8GW, shows the scale of the infrastructure now being considered for AI.

If the strategy works, Nvidia could increasingly become not just the company supplying the engines of the AI boom, but also one of the companies helping build the physical infrastructure in which those engines operate.