Anthropic Expands AI Infrastructure Push With $45B Nscale Deal
Anthropic is continuing its aggressive expansion of artificial intelligence infrastructure, reportedly agreeing to spend around $45 billion on computing capacity supplied by British AI infrastructure company Nscale.
The agreement represents one of Anthropic’s biggest infrastructure commitments to date and adds to a rapidly growing collection of deals designed to secure the enormous computing resources required to train and operate advanced AI models.
According to reports, the arrangement will give Anthropic access to computing infrastructure powered by Nvidia’s latest Vera Rubin technology, with capacity expected to begin supporting the AI company’s services in late 2027.
Nscale Emerges as a Major AI Infrastructure Provider
Nscale is a relatively young participant in the rapidly expanding AI infrastructure sector. The company launched from stealth in May 2024, but has quickly secured major customers and expanded its network of large-scale data centre projects across Europe and North America.
Microsoft is already among the company’s most significant partners. In April 2026, Nscale announced an expanded agreement with Microsoft involving more than 30,000 NVIDIA Rubin GPUs at its Narvik campus in Norway.
Anthropic’s new agreement is expected to run for six years, with computing resources reportedly being delivered from Nscale’s flagship data centre campus in West Virginia.
The West Virginia site is itself becoming a major part of Nscale’s international strategy. Earlier in 2026, the company announced plans involving up to 1.35GW of AI computing capacity at the site using Nvidia Vera Rubin technology.
Nvidia’s Vera Rubin Platform Will Power the Expansion
A central element of the Anthropic-Nscale agreement is Nvidia’s Vera Rubin architecture.
Introduced as the next generation of Nvidia’s AI computing platform, Rubin was designed around multiple tightly integrated components working together as a single AI supercomputing system.
The original Vera Rubin platform combines six different chips, including the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU and Spectrum-6 Ethernet Switch. Nvidia says the architecture is designed to improve both the speed and economics of large-scale AI training and inference.
For Anthropic, gaining access to this new generation of hardware could prove increasingly important as demand for its Claude AI models and associated enterprise services continues to grow.
Anthropic Has Been Building Compute Capacity at Speed
The Nscale agreement is far from an isolated investment.
Over the past eight months, Anthropic has pursued a series of major infrastructure partnerships as it attempts to strengthen its position against competitors such as OpenAI.
Earlier in August, the company reportedly secured a $10 billion computing agreement with AI infrastructure startup Volta.
Volta was founded in January, making the scale of the agreement particularly notable given the company’s age. The arrangement is expected to provide Anthropic with cloud computing resources over a six-year period through a data centre in Norway equipped with Nvidia Vera Rubin hardware.
The Volta arrangement demonstrates how quickly new infrastructure providers are emerging to meet demand from AI developers whose computing requirements increasingly extend beyond the capacity of traditional cloud arrangements.
AMD Becomes Another Major Infrastructure Partner
Anthropic also expanded its hardware strategy in July through an agreement with AMD.
The arrangement included a commitment by AMD to make a strategic equity investment of up to $5 billion in Anthropic, while the AI company plans to deploy as much as 2 gigawatts of AMD Instinct MI450 Series GPU capacity.
The first gigawatt is expected to begin deployment during the first half of 2027.
The AMD agreement also illustrates Anthropic’s increasingly diversified approach to computing infrastructure.
Rather than relying on one chip manufacturer or cloud provider, the company is building capacity across several hardware ecosystems, including Nvidia GPUs, Google TPUs, Amazon Trainium processors and AMD’s Instinct platform.
SpaceX Adds Another Source of Computing Power
In May, Anthropic revealed another significant computing partnership, this time with SpaceX.
Anthropic said the agreement gave the company access to more than 300 megawatts of new capacity and over 220,000 NVIDIA GPUs at SpaceX’s Colossus infrastructure.
Separate reporting indicated that the wider arrangement draws computing capacity from two different SpaceX data centers and is providing Anthropic with approximately $1.25 billion worth of capacity each month.
The partnership creates an unusual business connection between Anthropic and Elon Musk’s SpaceX, given Musk’s long-running rivalry with OpenAI CEO Sam Altman.
Yet the agreement also demonstrates the scale of today’s AI infrastructure market. As demand for computing resources intensifies, access to capacity can increasingly outweigh traditional competitive relationships between technology companies.
Amazon, Google and Broadcom Add Gigawatts of Capacity
April brought another round of major infrastructure expansion.
Anthropic strengthened its long-standing relationship with Amazon through an agreement that provides access to an additional 5 gigawatts of computing capacity for training and deploying Claude.
Anthropic said its expanded Amazon relationship represents a commitment of more than $100 billion over the next ten years to AWS technologies, highlighting the enormous capital requirements involved in maintaining frontier AI systems.
During the same month, Anthropic expanded its partnership with Google and Broadcom for additional next-generation TPU capacity.
The companies said the new infrastructure is expected to begin coming online in 2027, adding several more gigawatts to Anthropic’s future computing resources.
The AI Industry’s Compute Race Continues
Anthropic’s infrastructure spending reflects a much broader transformation underway across the technology industry.
Developing increasingly capable AI systems requires enormous amounts of processing power, electricity, specialised chips and data centre capacity. As a result, access to computing infrastructure has become one of the most important strategic advantages available to leading AI companies.
Anthropic is therefore not alone in securing capacity years in advance.
Google, OpenAI and Meta are among the major technology companies pursuing similar strategies as competition shifts beyond AI models themselves and increasingly towards the infrastructure required to develop and operate them.
The reported $45 billion Nscale agreement underlines just how valuable that infrastructure has become.
For Anthropic, the growing collection of deals with Nscale, Volta, AMD, SpaceX, Amazon, Google and Broadcom represents more than simply additional computing power. It is an attempt to establish a diversified, large-scale technology foundation capable of supporting Claude as demand for generative AI continues to expand worldwide.
0 Comments