Anthropic Could Spend $750 Billion on AI Infrastructure as Debt Concerns Hit Tech Stocks

Key Takeaways

• Anthropic’s existing cloud, chip and computing commitments exceed $600 billion, while another $150 billion of potential agreements could lift the total toward $750 billion.
• The Claude developer generated approximately $16.3 billion in first-half revenue, up from $1.1 billion a year earlier.
• Anthropic reported a net loss of $80.7 billion, although about $72 billion reportedly resulted from non-cash accounting charges linked to convertible securities.
• The company spent nearly $9 billion on computing and infrastructure during the first half, exceeding its infrastructure spending for all of 2025.
• A proposed $60 billion chip-financing package includes junior debt offering a yield above 10%, highlighting the rising cost of funding the AI buildout.
• The disclosures arrive as Anthropic prepares for a potential IPO that could value the company at approximately $2 trillion.

Anthropic’s long-term artificial-intelligence infrastructure obligations could approach $750 billion, intensifying Wall Street concerns that the AI boom is becoming increasingly dependent on debt, complex financing structures and uninterrupted demand for computing power.

A draft IPO prospectus reviewed by media outlets shows that Anthropic has already committed more than $600 billion to cloud capacity, chips and related infrastructure. The Claude developer is reportedly discussing additional agreements worth as much as $150 billion, potentially bringing its total future infrastructure exposure close to three-quarters of a trillion dollars.

The figure does not represent an immediate cash payment or a conventional one-year capital expenditure budget. It includes multi-year cloud-service contracts, chip leases, computing-capacity agreements and other infrastructure obligations extending across several years.

Nevertheless, its scale raises questions about whether Anthropic’s extraordinary revenue growth can keep pace with the cost of developing and operating increasingly powerful AI models.

How Anthropic’s Infrastructure Commitments Could Reach $750 Billion

Anthropic’s updated draft prospectus reportedly identifies more than $600 billion of existing future infrastructure expenses.

That total includes cloud and computing arrangements with Google, Amazon Web Services, Microsoft and other partners, as well as an obligation exceeding $160 billion connected to a Broadcom-supported lease of Google-designed AI processors.

The company is also negotiating further infrastructure agreements that could add as much as $150 billion.

The final amount could change because some of the additional agreements have not been completed. Payment timing, cancellation provisions and minimum-use conditions will also determine the actual financial burden.

Earlier versions of Anthropic’s IPO materials disclosed approximately $518 billion of cloud and computing commitments over the coming decade. Around 80% of that amount was reportedly non-cancellable or payable regardless of whether Anthropic used all the contracted capacity.

The increase from $518 billion to more than $600 billion indicates that Anthropic continued signing infrastructure agreements while preparing its public listing.

Revenue Climbs to $16.3 Billion in Six Months

Anthropic’s spending is being supported by exceptionally rapid revenue growth.

The company generated approximately $16.3 billion in revenue during the six months ended June 30, 2026, compared with $1.1 billion in the same period a year earlier. First-half revenue was also more than three times the approximately $4.6 billion generated during all of 2025.

Demand has been driven by Claude’s enterprise adoption, API usage and coding products such as Claude Code. Businesses are increasingly using Anthropic’s models for software development, customer support, data analysis and automated workflows.

However, revenue concentration remains a material risk. Approximately 20% of first-half sales reportedly came from only two customers. Many major customers also lack long-term commitments, creating a mismatch between relatively flexible customer contracts and Anthropic’s rigid infrastructure obligations.

If demand slows or customers switch to competing models, Anthropic could remain responsible for substantial cloud and chip payments without receiving the revenue originally anticipated.

The $80.7 Billion Loss Requires Context

Anthropic recorded a net loss of approximately $80.7 billion in the first half of 2026, but that figure does not reflect only cash consumed by normal operations.

Around $72 billion reportedly came from non-cash accounting charges related to changes in the value of convertible notes and other financing instruments. As Anthropic’s expected valuation increased, the accounting value of those obligations rose, producing a large reported expense.

The loss therefore overstates the company’s immediate cash outflow. Even after excluding those accounting adjustments, however, Anthropic remains an extremely capital-intensive business.

The company spent nearly $9 billion on computing capacity and infrastructure during the first six months of 2026. That was already more than the $7.3 billion it spent on infrastructure throughout 2025.

The distinction between accounting losses and cash operating costs will be crucial for investors evaluating Anthropic’s IPO. Non-cash charges may not directly drain liquidity, but infrastructure payments, interest costs and cloud commitments require real cash.

Anthropic Seeks a Record $60 Billion Chip-Financing Deal

Anthropic’s proposed financing arrangements illustrate how the AI infrastructure boom is moving beyond conventional corporate borrowing.

Bank of America, Citigroup and Morgan Stanley are reportedly syndicating a $60 billion financing package supporting Anthropic’s lease of Google-developed tensor processing units.

The package is expected to include:

• $42 billion of senior secured financing
• $18 billion of junior debt
• A partial guarantee from Broadcom
• A $9 billion commitment from Blackstone-managed funds
• A yield above 10% on part of the junior tranche
• Potential convertible notes issued to Broadcom

Blackstone is expected to provide approximately half of the junior tranche and distribute the remainder to other investors. The senior portion benefits from stronger security and Broadcom’s investment-grade credit profile, while the junior debt carries substantially greater risk.

A double-digit yield is unusually high for financing linked to major technology companies. It suggests that private-credit investors require significant compensation for the risks associated with Anthropic’s rapid expansion, uncertain long-term cash flow and dependence on future AI demand.

Why AI Debt Is Worrying Wall Street

The earliest phase of the AI investment boom was largely financed using cash generated by highly profitable technology companies.

That structure is changing. Approximately one-third of current AI infrastructure spending is now estimated to be debt-financed, exposing banks, bondholders and private-credit funds to the economics of data centres and computing contracts.

The financing model generally assumes that demand for AI computing will continue increasing fast enough to keep expensive data centres and chips fully utilised. If model efficiency improves, demand slows or customers resist higher prices, infrastructure operators may struggle to generate the returns required to service their debt.

Unlike conventional software businesses, frontier-model developers must continuously purchase processors, electricity, networking equipment and cloud capacity. Each new model generation can require substantially more computing power, making it difficult to separate growth spending from the recurring cost of remaining competitive.

Anthropic’s obligations are particularly notable because many agreements extend further into the future than its customer contracts.

Energy Is Becoming Another Constraint

Financing is not the only challenge. AI data centres also require access to enormous amounts of electricity.

Global data-centre power demand is forecast to increase approximately 170% between 2025 and 2030, with more than 60% of the growth expected in the United States. Grid connections, power-generation capacity and local planning approvals are failing to expand at the same pace.

At least 86 infrastructure projects have reportedly been delayed or cancelled since the beginning of 2026, often because of local opposition, electricity shortages or construction constraints.

These bottlenecks create several risks for Anthropic:

• It may pay for chips before sufficient power is available.
• Data-centre construction could take longer than projected.
• Electricity and cooling costs could exceed initial estimates.
• Local opposition could delay new facilities.
• Unused capacity could lower returns while debt payments continue.

The result is that Anthropic’s success depends not only on building better AI models, but also on securing power, land, chips and financing at a scale normally associated with energy or industrial companies.

Tech Stocks React to AI Credit Concerns

Anthropic’s disclosures arrived during a period of heightened volatility in AI-related stocks.

Investors have become more selective after debt prices weakened at infrastructure companies and credit-default-swap costs increased across parts of the technology sector. Companies with significant borrowing requirements have generally faced greater scrutiny than businesses capable of financing AI expansion from operating cash flow.

CoreWeave and other specialised cloud providers have been particularly sensitive to concerns about leverage, contract concentration and future refinancing needs. Broader technology indexes have also struggled to maintain momentum despite strong demand for AI products.

Anthropic’s growth provides evidence that customer demand remains substantial. However, its infrastructure commitments reinforce the argument that revenue growth alone may not be enough to justify current valuations if capital and interest expenses rise equally quickly.

Markets.com offers selected stock CFDs available for 24/7 trading, giving you more flexibility to respond to price movements outside regular market hours. Explore stock CFD trading and discover up to $5,000 in combined rewards for eligible new clients.

Anthropic IPO Faces a Major Valuation Test

Anthropic is preparing for a potential IPO that could value the company at approximately $2 trillion.

The company previously raised $65 billion at a valuation of about $965 billion. A $2 trillion public valuation would more than double that level and immediately place Anthropic among the world’s most valuable technology companies.

Its listing, reportedly targeted for after the US midterm elections in November, will test whether public investors are willing to value frontier AI companies using the same aggressive assumptions previously accepted by private markets.

The bullish case is based on Anthropic’s rapid revenue expansion, strong enterprise adoption and exposure to a potentially enormous market for automated knowledge work.

The principal concerns include:

• More than $600 billion of existing infrastructure obligations
• Potential commitments approaching $750 billion
• High financing and interest costs
• Dependence on a limited number of customers
• Competition from OpenAI, Google and open-source models
• Uncertain long-term pricing power
• Risks associated with AI safety and model failures
• The possibility of future shareholder dilution

Can Anthropic Grow Into Its Infrastructure Commitments?

Anthropic’s first-half revenue demonstrates that demand for Claude is growing rapidly. The company has expanded from a specialised AI laboratory into a large enterprise technology provider in a remarkably short period.

However, the financial challenge is equally extraordinary. Anthropic must convert short-term customer demand into durable cash flow while meeting hundreds of billions of dollars in infrastructure obligations.

If AI adoption continues accelerating, the agreements could secure the computing capacity needed for Anthropic to compete with OpenAI and Google. If growth slows, the same agreements could become rigid liabilities that damage margins and require additional financing.

The company’s IPO documents will therefore be evaluated less like those of a traditional software company and more like those of a capital-intensive infrastructure operator.

Anthropic’s possible $750 billion AI buildout shows that the industry’s next phase will depend on much more than model quality. Access to chips, power and affordable financing may ultimately determine which AI developers survive the infrastructure race.