Anthropic’s relationships with major cloud providers are becoming more complicated as the company races to secure the computing power needed to develop and operate its Claude AI models. The company relies on infrastructure partners including Amazon Web Services (AWS), Google Cloud and Microsoft, but these technology giants are also developing their own AI systems and services. A leaked IPO filing reported by The Wall Street Journal and covered by TechCrunch describes growing concerns about computing availability, partner competition and the enormous financial commitments required to sustain Anthropic’s expansion.
The financial scale is striking. Anthropic reported $16.3 billion in revenue during the first half of 2026, while its filing showed an $80.7 billion net loss, largely associated with accounting charges tied to financing instruments. It also spent $9 billion on computing during the first half of the year. The filing reportedly describes potential cloud-related spending exceeding $750 billion, including deals still being negotiated, while Reuters has separately reported talks for another $45 billion computing agreement with Nvidia-backed Nscale. These figures highlight Anthropic’s infrastructure ambitions, but they also raise questions about how the company will manage costs, capacity and financial risk as it prepares for a potential IPO.
Key Takeaways
- Cloud capacity is a constraint: Anthropic says access to sufficient computing power is essential to meet demand for Claude.
- Major partners are also competitors: Amazon, Google and Microsoft operate cloud businesses while developing their own AI products.
- Huge infrastructure commitments: Anthropic’s potential cloud-related spending could exceed $750 billion, including agreements still under negotiation.
- Another possible deal: Anthropic is reportedly discussing a second Nscale computing agreement worth approximately $45 billion.
- Heavy computing costs: The company spent $9 billion on computing in the first half of 2026.
- IPO scrutiny is increasing: Investors will need to assess infrastructure obligations, financing arrangements, customer concentration and profitability.
- Important distinction: Potential spending estimates, signed contracts and financing commitments are different financial categories and should not be treated as interchangeable.
Why Anthropic Is Struggling to Secure Enough Cloud Capacity
Advanced AI models require substantial computing resources for training, testing and serving users. As companies deploy AI assistants for coding, research, customer service and enterprise automation, demand for high-performance chips and data-centre capacity has increased.
Anthropic needs infrastructure for both model development and everyday Claude usage. Training requires large clusters of specialised processors, while serving customers requires reliable computing capacity that can respond to requests at scale.
The challenge is that access to suitable infrastructure cannot always be expanded immediately. Data centres require land, power connections, cooling systems, networking equipment and advanced chips. Even when a company has the money to pay for computing, suitable capacity may not be available where and when it is needed.
TechCrunch reported that Anthropic’s leaked IPO filing describes difficulty securing enough capacity from AWS, Google and Microsoft. The company is therefore looking beyond traditional hyperscale cloud providers to independent infrastructure developers and operators.
This creates a complicated planning problem. Anthropic must secure enough computing power to support future growth without committing to more capacity than it can economically use.
The scale of the spending problem
| Financial indicator | Reported figure |
|---|---|
| First-half 2026 revenue | $16.3 billion |
| First-half 2026 net loss | $80.7 billion |
| First-half computing expenditure | $9 billion |
| Potential cloud-related spending | More than $750 billion, including negotiations |
| Reported possible second Nscale deal | Approximately $45 billion |
Figures are based on reporting about Anthropic’s leaked IPO filing and related infrastructure negotiations. The potential spending estimate includes deals still under discussion and is not equivalent to fully signed obligations.
The $80.7 billion net loss also requires context. Reporting indicates that much of it was associated with accounting charges tied to financing instruments. It should not be interpreted as evidence that Anthropic spent $80.7 billion in cash on operations during the six-month period.
Nevertheless, the reported computing expenditure demonstrates how expensive it can be to operate an AI business at this scale. The company must continually invest in infrastructure while trying to turn rising usage into sustainable revenue.
Cloud Providers Are Becoming Anthropic’s Competitors
Anthropic’s dependence on major cloud platforms creates a strategic tension. Amazon, Google and Microsoft can provide the infrastructure needed to operate Claude, but each also has its own AI strategy and commercial incentives.
Amazon promotes its own AI services while supporting Anthropic through cloud infrastructure and investment. Google develops its Gemini models and operates Google Cloud. Microsoft offers its own AI products and cloud services while competing for enterprise customers.
These companies can benefit from supplying Anthropic with computing capacity. The relationship may generate infrastructure revenue, increase demand for specialised chips and encourage businesses to use their cloud platforms.
However, the same providers also have reasons to invest in their own models. If their AI products become more competitive, they may be less inclined to treat Anthropic simply as a customer whose growth they should support without qualification.
This does not mean the partnerships are ending. Cloud providers can serve customers that compete with them in other markets, and infrastructure sales can remain commercially attractive even when competition exists.
The difficulty is that Anthropic needs reliable capacity, favourable economics and predictable access. Its suppliers must balance those needs against their own infrastructure constraints, customer commitments and strategic priorities.
Anthropic Is Looking Beyond the Biggest Cloud Providers
Anthropic has been expanding its infrastructure relationships beyond the largest cloud companies. Independent data-centre developers can provide additional capacity, but they introduce different considerations around financing, operational experience and delivery schedules.
The company’s reported discussions with Nscale illustrate this strategy.
Reuters reported on October 19 that Anthropic was in talks for a second computing deal worth approximately $45 billion. The proposed facility would be built at Nscale’s Monarch data-centre campus in West Virginia and could provide 460 megawatts of capacity using Nvidia’s next-generation Vera Rubin AI chips.
The companies had already agreed in August on a separate deal reportedly worth $45 billion for computing capacity at the same campus. Anthropic also reportedly holds a right of first refusal on a potential third building.
The second deal remains under discussion, and the terms could change. It should not be described as a completed agreement.
| Infrastructure development | Reported status |
|---|---|
| First Nscale Monarch facility | Agreement announced in August |
| Second Monarch facility | Approximately $45 billion deal under discussion |
| Proposed capacity of second facility | 460 megawatts |
| Planned computing technology | Nvidia Vera Rubin AI chips |
| Possible third facility | Anthropic reportedly has a right of first refusal |
The potential expansion shows how much capacity Anthropic believes it may need. It also demonstrates the risks of depending on infrastructure projects that require substantial investment, power availability, equipment delivery and construction work before they can serve customers.
A contract for future computing capacity is not the same as an operating data centre. Delays in power connections, construction or hardware delivery can affect when capacity becomes available and when the customer begins receiving the expected benefit.
Why Anthropic’s Financing Arrangements Matter
The infrastructure race is not just about renting servers. It increasingly involves complex arrangements between AI companies, chip suppliers, cloud providers and financial partners.
Anthropic’s IPO disclosures have drawn attention to a financing arrangement involving Broadcom. Reuters reported on October 1 that Broadcom had agreed to lend Anthropic up to $42 billion to finance its lease of tensor processing unit (TPU) computing capacity.
The arrangement connects financing with access to computing hardware. Broadcom is involved in the design and supply chain for specialised AI chips, while Anthropic needs the computing capacity those chips enable.
Such arrangements can help AI companies secure infrastructure without paying every cost upfront. They may also allow suppliers to support demand for their hardware and related services.
However, financing introduces additional obligations. Investors need to understand interest costs, repayment terms, lease commitments, conditions that could accelerate payments and how the arrangements interact with other infrastructure contracts.
These details matter because infrastructure spending can become difficult to reverse. If customer demand grows more slowly than expected, a company may still need to pay for capacity it has committed to use or finance.
For Anthropic, the question is whether the revenue generated by Claude and related services will be sufficient to support the long-term cost of its infrastructure strategy.
The Difference Between Revenue Growth and Financial Sustainability
Anthropic’s first-half revenue of $16.3 billion demonstrates substantial commercial scale. Its reported third-quarter revenue of $16.8 billion, up 45% from the previous quarter, further indicates that the business continued expanding in the following period.
But revenue growth alone does not establish that the company’s infrastructure strategy is sustainable. Investors must consider operating expenses, financing costs, future capital requirements and the terms of long-term contracts.
The $80.7 billion first-half net loss is particularly important to interpret carefully. Much of the reported loss was attributed to accounting charges linked to financing instruments rather than ordinary operating expenditure. That distinction matters, but it does not make every underlying obligation irrelevant.
Investors will want to distinguish between cash costs already paid, contractual commitments due in future years, non-cash accounting adjustments and agreements that remain under negotiation.
They will also examine how much revenue depends on a relatively small number of customers. The leaked filing reportedly indicated that two customers accounted for roughly one-fifth of first-half sales. Customer concentration can make revenue more vulnerable if a major buyer reduces usage or changes suppliers.
Anthropic’s potential IPO will therefore put its infrastructure strategy under greater scrutiny. Investors will need a clear view of how the company expects to finance growth, manage long-term obligations and convert AI demand into cash flow.
Why Cloud Capacity Could Shape the AI Industry
The competition for computing power is changing the economics of AI development. Access to advanced processors, electricity and data-centre capacity can determine how quickly companies release new models and how many customers they can serve.
For AI developers, insufficient capacity can limit growth even when customers are willing to pay. For infrastructure providers, the same shortage creates an opportunity to sell computing services and secure long-term contracts.
Yet the expansion also carries risks for the wider market. Data centres are expensive to build, and their commercial success depends on demand continuing long enough to justify the investment. High borrowing costs, construction delays, power constraints or weaker-than-expected AI spending can make projects harder to finance.
Anthropic’s growing network of infrastructure relationships reflects both sides of this market. The company needs capacity to compete, while its suppliers need customers willing to make large, long-term commitments.
The more tightly these relationships link chip suppliers, lenders, cloud providers and AI developers, the more important it becomes to understand who bears the financial risk if demand changes.
What This Means for Anthropic’s IPO
Anthropic is preparing for a potential November 2026 IPO, and its infrastructure disclosures give investors a more detailed view of the costs behind its growth.
Public-market investors typically examine revenue growth, margins, customer retention, capital requirements and the durability of a company’s competitive advantage. For an AI developer, infrastructure contracts and financing arrangements are likely to receive particular attention.
A large spending commitment is not necessarily negative. Securing computing capacity in advance may be essential to serving customers and building future products. Long-term contracts can also provide greater certainty about access to infrastructure.
The risk is that commitments become too large relative to future revenue or are tied to assumptions that do not hold. If AI prices fall, customers use models less than expected or more efficient systems reduce computing requirements, the economics of long-term contracts could change.
Conversely, if demand continues to grow rapidly and computing remains scarce, early access to capacity could give Anthropic a competitive advantage.
The IPO will provide an opportunity for investors to evaluate these trade-offs in greater detail, including the distinction between signed contracts and negotiations that have not yet been completed.
The Bigger Picture
Anthropic’s cloud partnerships reveal a central challenge facing the AI industry: building advanced models is only part of the business. Companies also need enormous amounts of computing power, reliable infrastructure, financing and long-term supplier relationships. The providers that supply these resources are increasingly competing in AI themselves, making the relationship between customer and supplier more complicated.
Anthropic’s strategy could help it secure the capacity needed to grow Claude, but the associated commitments create financial and operational risks. The ultimate test will be whether the company can translate its expanding revenue into sustainable cash flow while managing infrastructure costs, financing obligations and competition from its own cloud partners.
Looking Ahead
The next developments to watch include whether Anthropic finalises the proposed second Nscale agreement, how its other cloud contracts evolve and what additional details emerge through its IPO process. Investors will also be looking for clearer disclosures about which infrastructure commitments are binding, which remain under negotiation and how the company expects to finance them.
For the broader AI market, Anthropic’s experience highlights the growing importance of power, chips and data-centre availability. The companies that secure sufficient capacity at manageable cost may be better positioned to serve rising demand. But infrastructure expansion alone will not guarantee commercial success: durable customer demand, pricing power and disciplined financial management will remain essential.
Frequently Asked Questions
1. Why are Anthropic’s cloud partnerships becoming complicated?
Anthropic depends on major cloud providers for computing capacity, but Amazon, Google and Microsoft are also developing competing AI products. The company is additionally turning to independent infrastructure providers as it seeks more capacity.
2. How much could Anthropic spend on cloud infrastructure?
Reporting about its leaked IPO filing indicates that potential cloud-related spending could exceed $750 billion, including agreements still being negotiated. This is not the same as saying the entire amount is already contractually committed.
3. What is the reported Nscale deal worth?
Reuters reported that Anthropic is discussing a second Nscale computing deal worth approximately $45 billion. The proposed facility could provide 460 megawatts of capacity, but the discussions are ongoing and the terms could change.
4. Why does Anthropic need so much computing power?
It needs computing capacity to train AI models, develop new capabilities and serve users running Claude for tasks such as coding, research and enterprise automation.
5. Does Anthropic’s reported net loss mean it spent $80.7 billion in cash?
No. Reporting indicates that much of the first-half net loss was associated with accounting charges tied to financing instruments. Net loss and cash expenditure are different measures.
6. How could these commitments affect Anthropic’s IPO?
Investors will assess whether expected revenue can support infrastructure spending and financing obligations. Large commitments may secure future capacity, but they can also create financial risk if demand or pricing falls short of expectations.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
