The rapid growth of artificial intelligence is creating unprecedented demand for computing power, pushing AI companies to secure increasingly large amounts of cloud and data centre capacity. Anthropic is the latest major AI developer to make a significant infrastructure commitment, reportedly signing a US$35 billion cloud computing agreement with Nvidia-backed cloud provider Lambda.

The agreement is designed to expand the computing capacity available for Anthropic's AI products, including its Claude family of models. According to reporting from The Wall Street Journal and Reuters, the deal will bring additional Nvidia-powered infrastructure online as Anthropic responds to growing demand for its AI services.

The scale of the agreement highlights how access to computing infrastructure has become one of the most important resources in the global AI industry. Developing increasingly capable AI models requires enormous quantities of processing power, both during the initial training process and later when models respond to requests from millions of users and businesses.

Behind Anthropic's latest agreement is a network of technology and infrastructure companies. Lambda will provide the cloud computing capacity, while Nvidia will supply the chips that power the infrastructure. The data centre supporting the arrangement is being developed by Hut 8 in Nueces County, Texas.

Nvidia also holds the lease on the Texas data centre, according to The Wall Street Journal. The chipmaker previously reached an agreement with Hut 8 to secure capacity at the facility before Lambda's cloud agreement with Anthropic emerged.

The arrangement illustrates how the AI infrastructure ecosystem is becoming increasingly interconnected. AI developers need access to powerful computing resources, cloud providers need large quantities of specialised hardware, and data centre operators need facilities capable of supporting the substantial power and cooling requirements associated with modern AI workloads.

Nvidia sits at the centre of much of this infrastructure expansion. While the company is best known for supplying the graphics processing units used to train and operate AI models, its role in the AI economy has expanded beyond simply manufacturing chips.

The company has invested in cloud infrastructure providers including Lambda while also using its financial resources and partnerships to increase access to AI computing capacity. That strategy can help smaller cloud providers compete for major AI workloads while simultaneously expanding the infrastructure capable of running Nvidia hardware.

For Anthropic, securing additional infrastructure has become particularly important as demand for Claude continues to grow. Reuters reported that the new Lambda agreement is intended specifically to bring additional Nvidia capacity online to meet increasing demand for Anthropic's AI products.

The Lambda agreement is also part of a much larger infrastructure expansion by Anthropic.

Only days earlier, Anthropic reportedly agreed to spend approximately US$45 billion over six years to rent computing capacity from Nscale at a data centre in West Virginia. That agreement is expected to provide access to around 460 megawatts of capacity using Nvidia's next-generation Vera Rubin processors.

Together, these agreements demonstrate how rapidly the economics of artificial intelligence are shifting from software development toward large-scale physical infrastructure.

Although AI services are experienced through websites, applications and conversational interfaces, the systems behind them depend on vast networks of servers, networking equipment, cooling systems and electricity infrastructure.

As AI models become more capable and usage increases, inference — the computing process used when an AI model generates an answer — is also becoming an increasingly significant infrastructure requirement. A successful AI service may need to process enormous numbers of requests continuously, meaning computing capacity must expand alongside user adoption.

This dynamic is contributing to the rise of specialised AI cloud companies sometimes referred to as "neoclouds." Instead of competing across the entire cloud computing market, these providers focus heavily on supplying GPU infrastructure optimised for artificial intelligence and high-performance computing.

Companies such as Lambda and Nscale are becoming increasingly important within this ecosystem because they can provide AI developers with access to large clusters of specialised processors without requiring those developers to build every data centre themselves.

The model can also provide greater flexibility. Building a large-scale AI data centre requires substantial capital, specialised engineering expertise, access to electricity and potentially years of planning and construction. Renting computing capacity allows AI companies to expand more quickly while infrastructure partners handle much of the physical development.

However, the enormous scale of these agreements also demonstrates how capital-intensive the AI industry is becoming. Access to advanced chips, electricity and suitable data centre locations could increasingly influence which companies are able to train and operate the most capable AI systems.

Energy availability is another growing consideration. Large AI data centres can require hundreds of megawatts of power, placing new pressure on electricity generation and transmission infrastructure in regions experiencing rapid data centre development.

Cooling requirements, water consumption, grid capacity and the environmental impact of new facilities are therefore becoming increasingly important parts of the wider discussion surrounding AI expansion.

At the same time, the concentration of AI workloads around a relatively small group of hardware manufacturers, cloud providers and infrastructure developers raises questions about resilience and supply chains. AI companies may increasingly seek relationships with multiple providers to reduce their dependence on any single source of computing capacity.

Anthropic's recent infrastructure agreements demonstrate this approach. By securing large amounts of computing power across different providers and locations, the company can increase the capacity available for its AI products while building a broader infrastructure base.

The US$35 billion Lambda agreement ultimately reflects a larger transformation occurring across artificial intelligence. Competition is no longer centred solely on which company can develop the most capable model. Increasingly, it also depends on who can secure the chips, data centres, energy and cloud infrastructure required to operate those models at enormous scale.

As demand for generative AI continues to grow, agreements of this magnitude could become an increasingly important part of the industry's development. The software may remain the part users interact with, but behind every AI response is a rapidly expanding physical infrastructure — and companies are spending tens of billions of dollars to ensure enough computing power is available when it is needed.