Nscale and Anthropic Discuss Second $45 Billion AI Computing Deal

Nscale, the Nvidia-backed AI cloud infrastructure company, is in talks with Anthropic for a second computing agreement potentially worth $45 billion, according to a report published by Moneycontrol on October 20, 2026. The proposed deal would cover another facility at Nscale’s Monarch data centre campus in West Virginia, providing 460 megawatts of computing capacity powered by Nvidia’s next-generation Vera Rubin AI chips. If completed, the agreement would expand the infrastructure available to Anthropic, the developer of Claude, as demand for AI computing continues to rise.

The negotiations come as both companies prepare for potential public listings and investors scrutinise the financial commitments behind the AI infrastructure boom. Nscale is targeting a valuation of approximately $30 billion for its planned IPO, while Anthropic continues to secure computing capacity to support its expanding AI services. However, the second agreement has not been finalised, and the reported $45 billion represents a potential contract value rather than money already paid or revenue already earned.

Key Takeaways

• Potential deal value: Nscale and Anthropic are discussing a second computing agreement worth approximately $45 billion.
• New facility: The proposed data centre would be built at the Monarch campus in West Virginia.
• Computing capacity: The facility would provide 460 megawatts of capacity.
• Nvidia technology: The proposed infrastructure would use Nvidia’s Vera Rubin AI chips.
• IPO implications: A second major customer agreement could strengthen Nscale’s investment case ahead of its planned listing.
• Existing relationship: Anthropic already has substantial infrastructure agreements with Nscale.
• Important caveat: Negotiations are ongoing, and the terms may change before any final agreement.

What Is the Proposed $45 Billion Deal?

The proposed agreement would give Anthropic access to additional computing resources through Nscale’s data centre infrastructure. The two companies are discussing a second major facility at the Monarch campus, which is being developed to accommodate large-scale AI workloads.

According to the reporting, the proposed building would deliver 460 megawatts of computing capacity and use Nvidia’s Vera Rubin chips. These processors are designed for demanding AI workloads, including the training and operation of advanced models.

The reported value of approximately $45 billion reflects the potential scale of the commercial arrangement. It should not be interpreted as an immediate cash payment, a completed infrastructure investment or guaranteed profit for Nscale.

The economics of a long-term computing contract depend on factors such as the duration of the agreement, the cost of building the facility, electricity supply, equipment procurement, financing and the terms under which computing services are delivered.

Nscale and Anthropic have not publicly confirmed a final agreement. The negotiations could conclude in the coming days, according to people familiar with the discussions cited in the report, but the terms remain subject to change.

How the proposed facility fits into the Monarch campus

The Monarch campus in West Virginia is intended to become a major computing hub for AI infrastructure. Nscale’s IPO filing describes the broader site as capable of scaling to more than eight gigawatts of gross power when fully developed.

That figure represents the potential scale of the entire campus, not the capacity of the proposed Anthropic building. The facility under discussion would provide 460 megawatts.

Developing data centres at this scale requires more than installing servers. Operators must secure electricity, construct buildings, establish cooling systems, connect networking equipment and ensure that the infrastructure can operate reliably under heavy workloads.

These requirements make data centre development capital-intensive and create execution risks. Delays in construction, power connections or equipment delivery can affect when a facility begins generating revenue.

Why Anthropic Needs More AI Computing Power

Anthropic develops the Claude family of AI models, which are used for tasks including software development, writing, research, business analysis and automated workflows.

As businesses incorporate AI into more products and internal processes, model developers need access to sufficient computing capacity to serve users reliably. Training advanced models also requires substantial computing resources, while inference—the process of generating responses to user requests—can consume large amounts of processing capacity as usage expands.

Securing infrastructure in advance can help an AI company plan its product launches and reduce uncertainty about future computing availability. Long-term agreements may also provide access to facilities and specialised chips that would be difficult to obtain quickly through short-term arrangements.

However, reserving more capacity also creates financial obligations. If demand grows more slowly than expected, a company may have to pay for computing resources that it cannot use as efficiently as planned, depending on the contract’s terms.

Anthropic is therefore balancing two competing requirements: securing enough infrastructure to support future growth while ensuring that the cost of that capacity remains sustainable.

The proposed Nscale agreement would add to Anthropic’s existing infrastructure commitments, reflecting the company’s continued effort to secure computing resources from multiple providers.

Nvidia’s Vera Rubin Chips Are Central to the Plan

The proposed facility is expected to use Nvidia’s Vera Rubin AI chips, the next-generation computing platform designed for demanding AI workloads.

Advanced AI accelerators are central to the economics of modern data centres because they determine how efficiently infrastructure can train and run AI models. Performance, memory capacity, networking and energy consumption can all affect the amount of useful computing work a facility delivers.

For Anthropic, access to newer hardware could support its ability to operate larger workloads and serve customers at scale. For Nscale, using advanced Nvidia technology could help position the campus as a destination for customers seeking high-performance AI computing.

The relationship also illustrates Nvidia’s role in the broader AI infrastructure market. The company supplies the chips and related technology that data centre operators use to build capacity, while cloud providers and AI developers make commercial commitments to use that capacity.

However, the presence of advanced chips does not guarantee that a facility will be profitable. Returns depend on utilisation, electricity costs, financing, operating expenses and the prices customers are willing to pay for computing services.

Nscale’s IPO Could Benefit From Another Major Contract

Nscale is preparing for a public listing and has been seeking a valuation of approximately $30 billion. A second major agreement with Anthropic could strengthen its pitch to prospective investors by demonstrating demand for its infrastructure.

The company has already attracted significant financing. It raised $3.36 billion in a pre-IPO financing round in September 2026, with participation from investors including Nvidia, Apollo Global Management, Citadel, Hudson Bay Capital and the Abu Dhabi Investment Council.

Nscale’s IPO filing also indicates that its contracted business has expanded rapidly. The company has reported contract value exceeding $103 billion, reflecting its large infrastructure commitments and customer agreements.

Yet contract value should not be confused with recognised revenue or cash collected. Long-term contracts may generate revenue over several years, while the infrastructure needed to fulfil them must often be financed and built before all the income arrives.

Investors evaluating Nscale will therefore need to examine more than the headline value of its contracts. They will want to understand the timing of revenue recognition, construction costs, financing requirements, customer concentration, power availability and the company’s ability to deliver promised capacity.

A large contract can provide visibility into future demand, but it can also increase the financial risk associated with execution.

Anthropic Already Has Major Infrastructure Commitments

The proposed second deal would add to an already substantial relationship between the two companies.

According to a regulatory filing for Nscale’s planned IPO, Anthropic has existing infrastructure agreements with Nscale worth as much as approximately $44.6 billion. The new negotiations could create a further commitment of similar headline size if an agreement is signed.

Anthropic has also secured computing infrastructure through other providers. Reported agreements include a $50 billion arrangement with Fluidstack, a $10 billion agreement with Volta Infrastructure Holdings and a $45 billion computing agreement with SpaceX.

These deals demonstrate how AI developers are attempting to secure capacity through multiple infrastructure partners rather than relying on a single provider.

Diversifying suppliers can help manage capacity constraints and give AI companies access to different facilities, technologies and locations. However, it also creates a complex network of long-term commercial commitments that must be coordinated with product demand and financial planning.

For investors, the challenge is determining whether the revenue generated by AI services will be sufficient to support the infrastructure spending taking place across the industry.

Why Investors Are Becoming More Cautious About AI Infrastructure

The proposed agreement arrives amid growing scrutiny of the financial risks associated with the AI data centre boom.

Building large facilities requires substantial upfront investment in land, buildings, power infrastructure, cooling equipment, servers and networking systems. Many operators finance part of that expansion through borrowing or external investment.

The resulting financial structure can be risky if construction is delayed, electricity connections take longer than expected or customers do not use the capacity at the anticipated rate.

Concerns have also emerged about the amount of debt used to finance AI infrastructure. Investors have been examining whether the industry’s rapid expansion will generate enough long-term cash flow to justify the scale of its commitments.

Nscale faces questions of its own as it prepares for an IPO. Investors have reportedly urged the company to reconsider its target valuation, while the business continues working through comments from the US Securities and Exchange Commission on its filing.

The company has also experienced a setback at the Monarch site: Microsoft, which had been expected to use part of the West Virginia campus, withdrew from its planned arrangement after previously signing a letter of intent. Anthropic subsequently emerged as a prospective tenant for the facility.

The development illustrates the importance of securing firm customers and dependable power connections before committing to large infrastructure projects. A data centre can have substantial theoretical capacity, but its commercial value depends on whether it can be built, powered and used on schedule.

What the Deal Could Mean for the AI Industry

If the agreement is completed, it would reinforce the trend toward long-term partnerships between AI developers and specialised infrastructure companies.

Model developers need dependable computing capacity to serve customers, while data centre operators need major customers to justify investments in buildings, power systems and advanced chips. Large contracts can align these interests by giving developers access to capacity and operators greater visibility into future demand.

However, the arrangement also links the financial performance of infrastructure companies to a relatively small number of large AI customers. If an important customer changes its plans, delays product development or reduces computing demand, the effects could spread through the infrastructure supply chain.

The proposed deal also shows how AI competition increasingly depends on access to physical resources, not just the quality of software models. Computing hardware, electricity, cooling and network connectivity are becoming essential components of the race to build and deploy advanced AI systems.

For companies such as Anthropic, the goal is to secure enough infrastructure to remain competitive. For providers such as Nscale, the challenge is to turn that demand into reliable revenue and sustainable returns on capital.

The Bigger Picture

The proposed $45 billion Nscale-Anthropic agreement reflects the extraordinary scale of investment required to support advanced AI systems. Anthropic is seeking additional computing capacity for Claude, while Nscale is developing infrastructure that could help it attract major customers and strengthen its IPO plans. Nvidia’s Vera Rubin technology would be central to the proposed facility, linking the deal to the broader market for AI accelerators.

The agreement also highlights the financial risks of the infrastructure boom. A contract’s headline value does not guarantee immediate revenue or profit, and data centre developers must navigate construction schedules, power constraints, financing costs and customer commitments. Whether these investments produce attractive returns will depend on sustained demand for AI computing and the ability to deliver capacity efficiently.

Looking Ahead

The immediate question is whether Nscale and Anthropic finalise the second agreement and disclose its commercial terms. Investors will also watch Nscale’s IPO preparations, including the company’s response to regulatory comments and how it explains the relationship between contracted business, future revenue and capital expenditure. Until the negotiations conclude, the proposed $45 billion deal should be treated as a potential agreement rather than a completed transaction.

Over the longer term, the success of the Monarch campus will depend on whether Nscale can bring new capacity online and secure enough utilisation to justify its investment. Anthropic, meanwhile, must ensure that the computing resources it commits to can support growing demand without undermining its financial performance. The outcome will provide another indication of how effectively the AI industry can convert enormous infrastructure spending into sustainable commercial returns.

Frequently Asked Questions

1. What is the proposed Nscale-Anthropic deal worth?

The companies are discussing a second computing agreement potentially worth approximately $45 billion. The deal has not been finalised, and the terms may change.

2. Where will the proposed data centre be built?

The proposed facility would be part of Nscale’s Monarch data centre campus in West Virginia, United States.

3. How much computing capacity would the facility provide?

The proposed building would deliver 460 megawatts of computing capacity, using Nvidia’s Vera Rubin AI chips.

4. Why does Anthropic need more computing power?

Anthropic needs computing resources to train and operate its Claude AI models and serve growing demand from businesses and other users.

5. How could the deal affect Nscale’s IPO?

A major additional customer agreement could strengthen Nscale’s case to investors by demonstrating demand for its infrastructure. However, investors will still need to assess construction costs, financing, revenue timing and execution risks.

6. Is the $45 billion deal confirmed?

No. The companies are still negotiating. The reported value is a potential contract amount, not confirmed revenue or a completed cash transaction.