Anthropic’s Cloud Partnerships Face Capacity and Cost Challenges

Anthropic is facing a growing infrastructure challenge as demand for its Claude AI models rises faster than it can secure computing capacity from cloud providers. The company depends on Amazon Web Services (AWS), Google Cloud and Microsoft-related infrastructure while competing with these technology giants in the market for AI models and enterprise software. A reported IPO filing indicates that Anthropic is exploring additional capacity through its own data-centre arrangements and third-party operators, with potential cloud commitments exceeding $750 billion, including agreements still under negotiation.

The situation highlights a central tension in the AI industry: companies need enormous amounts of computing power to develop and operate advanced models, but the suppliers of that infrastructure are also building competing AI products. Anthropic reported $16.3 billion in revenue for the first half of 2026, alongside an $80.7 billion net loss that was reportedly driven largely by accounting charges related to financing instruments. As it prepares for a potential November IPO, investors are likely to scrutinise its infrastructure commitments, its ability to obtain enough computing capacity and the long-term economics of running Claude at scale.

Key Takeaways

• Cloud capacity pressure: Anthropic reportedly expects difficulty obtaining sufficient computing capacity from major providers as demand for Claude grows.
• Potential commitments: Its projected cloud-related commitments could exceed $750 billion, including some deals still being negotiated.
• Competing partners: Amazon, Google and Microsoft have cloud businesses that also support their own AI ambitions.
• Third-party infrastructure: Anthropic is leasing data centres and pursuing additional capacity deals with specialist operators.
• Nscale negotiations: A potential second agreement worth about $45 billion has been reported, but it has not been finalised.
• Financial scrutiny: Anthropic reported $16.3 billion in first-half revenue and an $80.7 billion net loss, much of which was reportedly related to accounting charges.
• IPO implications: Investors will examine whether the company can secure reliable computing capacity without taking on unsustainable financial commitments.

Why Anthropic Is Running Into Cloud Partnership Problems

Anthropic needs access to large quantities of specialised computing hardware to train its models and serve customers using Claude. That includes processors, networking equipment, data-centre capacity, electricity and the systems needed to keep AI services running reliably.

Building this infrastructure independently is expensive and time-consuming. Cloud providers already operate large computing networks and can supply capacity without requiring Anthropic to construct every facility itself.

Anthropic has therefore developed relationships with major technology companies, including Amazon and Google. These relationships provide access to computing infrastructure, while investments and commercial agreements can help support the company’s expansion.

The difficulty is that the same companies are developing their own AI models, assistants and enterprise tools. Amazon has its own AI offerings, Google develops Gemini, and Microsoft has invested heavily in AI products and infrastructure.

This creates a complicated commercial relationship. Anthropic is a valuable customer because its models require substantial computing resources, but it is also a competitor seeking many of the same enterprise customers.

That does not mean these partnerships have ended. Rather, the reported capacity constraints suggest that commercial agreements alone may not guarantee Anthropic all the computing resources it wants, when it wants them.

More Than $750 Billion in Potential Cloud Commitments

The scale of Anthropic’s projected infrastructure spending is one of the most striking aspects of the reported IPO filing.

The company reportedly expects its cloud commitments could exceed $750 billion. This figure should not be interpreted as $750 billion already spent, nor as a fully executed set of contracts. It includes arrangements that remain under negotiation, making the eventual total dependent on future agreements and infrastructure plans.

Infrastructure issue	Why it matters
Long-term cloud contracts	Can secure capacity but create substantial future financial obligations
Computing shortages	Can limit model development and the ability to serve customers
Data-centre leasing	Offers another route to capacity without owning every facility
Third-party operators	Expand supply options but introduce execution and reliability risks
Power and permitting	Can delay projects even after commercial agreements are negotiated

The commitments reflect the unusual economics of frontier AI. Unlike a conventional software company that can serve additional customers mainly through existing software infrastructure, a model provider may need substantial additional computing resources as usage expands.

Even after a model has been trained, serving it to customers requires ongoing computation. Complex reasoning, coding agents and high-volume enterprise applications can consume significant capacity.

Long-term contracts may help Anthropic plan ahead, but they also create exposure to changes in demand, hardware efficiency and model economics. If computing becomes cheaper or a future model requires fewer resources, previously negotiated capacity may become less attractive. If demand grows faster than expected, the company may still need additional supply.

Why Amazon, Google and Microsoft Are Both Partners and Competitors

Anthropic’s relationship with major cloud companies is not simply a supplier problem. It reflects the increasingly overlapping business models of the technology industry.

Amazon Web Services

Amazon has invested in Anthropic and provides computing resources through AWS. The companies have also worked together on infrastructure using Amazon’s AI chips.

The relationship gives Anthropic access to a large cloud platform and gives Amazon a prominent AI model partner for its customers. However, Amazon is also developing its own AI products and services, meaning it has commercial interests beyond Anthropic’s success.

Google Cloud

Google has invested in Anthropic and supplies cloud infrastructure, including access to its Tensor Processing Units (TPUs). These accelerators provide an alternative to relying exclusively on Nvidia-based systems.

Using different types of hardware can help an AI company diversify its supply and potentially improve the cost and performance of workloads. But the availability of computing capacity still depends on hardware production, data-centre construction, power and the needs of other customers.

Microsoft

Microsoft competes with Anthropic in enterprise AI through its own products and its relationship with OpenAI. Its cloud infrastructure business also serves a wide range of AI customers.

The reported difficulties in securing sufficient capacity from major providers highlight how quickly the market is changing. Cloud companies must allocate resources among their own AI initiatives, existing customers and outside model developers.

These relationships can remain commercially valuable even when interests overlap. The challenge for Anthropic is to negotiate enough capacity and maintain flexibility without depending too heavily on any one supplier.

Why Anthropic Is Turning to Third-Party Data Centres

Anthropic is exploring infrastructure options beyond the largest cloud platforms. One reported route involves specialist data-centre operators that build facilities and provide computing capacity to AI companies.

Nscale, a UK-based AI infrastructure company backed by Nvidia, is in discussions with Anthropic over a possible second major deal. The proposed agreement could be worth around $45 billion, similar to an earlier arrangement announced in August.

The additional project would reportedly support a second 460-megawatt data centre at Nscale’s Monarch campus in West Virginia. If completed, it could substantially expand the computing capacity available to Anthropic.

However, the proposed deal has not been finalised. Reports indicate that due diligence remains under way and that the project faces outstanding permitting and financing requirements. A contract of this size does not automatically mean the facility is fully funded, approved or ready to operate.

Third-party operators can help AI companies obtain capacity without building every data centre themselves. But they also introduce risks: a project may be delayed by permits, equipment shortages, financing constraints or power infrastructure problems.

Anthropic must therefore assess not just how much capacity a provider promises, but whether that capacity can be delivered on schedule and at a commercially viable cost.

The Financial Risks Behind AI Infrastructure Expansion

Anthropic’s infrastructure plans come at a time when investors are examining the cost of building AI businesses.

The company reported $16.3 billion in first-half 2026 revenue, a sharp increase from the previous year. However, the reported $80.7 billion net loss has attracted attention because of its scale.

The loss was reportedly driven largely by accounting charges associated with financing instruments. That distinction is important: a large accounting loss does not necessarily represent an equivalent cash outflow during the period. Investors need to examine the company’s detailed financial statements to understand operating expenses, non-cash charges, cash flow and future obligations separately.

Nevertheless, Anthropic’s computing costs are substantial. The reported IPO material indicated that the company spent approximately $9 billion on computing during the first half of 2026 alone.

This spending is tied to the resources required to develop advanced models and support their use by customers. It can be necessary for growth, but it creates pressure to generate enough revenue and profit to justify the investment.

The company’s financial challenge is to ensure that new computing capacity produces useful commercial output. If infrastructure costs grow faster than revenue, margins could come under pressure. If capacity is insufficient, Anthropic may struggle to meet demand, improve its models or deliver reliable services.

The balance between those two risks will be central to how investors evaluate the company.

Customer Concentration Adds Another Risk

The reported filing also raised questions about the concentration of Anthropic’s revenue. Two customers accounted for approximately one-fifth of sales, according to reporting on the disclosures.

Large enterprise customers can provide meaningful revenue and help validate an AI product for business use. But dependence on a relatively small number of buyers creates risks if one customer reduces usage, changes suppliers or negotiates lower prices.

Customer concentration becomes especially important when a company has substantial long-term infrastructure commitments. If computing contracts are fixed or difficult to change, a decline in demand from a major customer may leave the company paying for capacity it no longer needs.

Anthropic can reduce this risk by expanding its customer base, increasing adoption across industries and developing products that encourage recurring use. But customer diversification takes time, and the economics of large AI workloads may remain sensitive to a relatively small number of high-volume users.

For investors, the key question is whether Anthropic can turn strong demand into a broad and durable revenue base that supports its infrastructure spending.

Why the IPO Makes These Partnerships More Important

Anthropic is reportedly targeting a November 2026 IPO. A public listing would expose its financial performance, risks and future commitments to broader investor scrutiny.

Investors will want to understand how much of the company’s projected cloud spending is already contractually committed, how much remains conditional and what happens if infrastructure projects are delayed.

They will also examine whether Anthropic has enough flexibility to switch between providers, negotiate pricing and adjust its capacity as model efficiency improves. Diversification can reduce reliance on individual suppliers, but managing multiple hardware types, cloud platforms and data-centre operators adds operational complexity.

The company’s ability to forecast demand will also matter. Securing too little capacity can restrict growth, while committing to too much can leave the business with expensive infrastructure obligations.

A successful AI company needs more than popular models. It needs a sustainable system for financing, securing and operating the infrastructure behind those models.

The Bigger Picture

Anthropic’s cloud challenges illustrate a defining feature of the AI industry: the companies building the most advanced models are also among the largest consumers of the infrastructure required to run them. Their cloud providers can be strategic partners, investors and direct competitors at the same time. As demand rises, that overlap makes access to computing capacity a question of commercial leverage, hardware supply and long-term capital planning.

Anthropic’s move toward third-party data centres and additional capacity agreements may reduce its dependence on individual cloud providers, but it will not eliminate infrastructure risk. Projects still need financing, power, permits, hardware and reliable operations. The company’s ability to convert computing capacity into sustainable revenue will be as important as its ability to develop capable AI models.

Looking Ahead

The immediate focus will be on whether Anthropic finalises its potential second Nscale deal and secures enough capacity from its cloud partners to support its growth plans. Investors will also want clarity on which infrastructure commitments are signed, which remain under negotiation and how the company expects to fund them. The reported $750 billion figure is a projection that includes proposed arrangements, not a statement of money already spent.

As Anthropic approaches a potential IPO, its infrastructure strategy will become a central part of the investment case. The company needs sufficient computing resources to remain competitive, but it must also demonstrate that its commitments are financially manageable. The outcome could help reveal whether the industry’s enormous spending on AI infrastructure can translate into durable business returns.

Frequently Asked Questions

1. Why are Anthropic’s cloud partnerships getting complicated?

Anthropic depends on major cloud providers for computing capacity, but Amazon, Google and Microsoft also compete in AI. Reported capacity shortages and the need to negotiate additional infrastructure have made these relationships more complex.

2. Is Anthropic committed to spending $750 billion on cloud computing?

The reported figure describes potential cloud commitments that could exceed $750 billion, including some arrangements still under negotiation. It should not be treated as money already spent or as a fully signed set of contracts.

3. What is Anthropic’s potential deal with Nscale?

Anthropic is reportedly discussing a second deal worth around $45 billion for a proposed 460-megawatt data centre at Nscale’s Monarch campus in West Virginia. The agreement has not been finalised and the project faces outstanding development requirements.

4. Why does Anthropic need so much computing power?

Training advanced AI models and serving them to customers require specialised chips, networking, electricity and data-centre capacity. Greater usage and more complex AI tasks can increase these requirements.

5. Does Anthropic’s $80.7 billion net loss mean it spent that much cash?

Not necessarily. Reports indicate that much of the first-half loss was related to accounting charges associated with financing instruments. Net income, cash flow and operating expenditure are different measures and must be evaluated separately.

6. How could these infrastructure challenges affect Anthropic’s IPO?

Investors may scrutinise the company’s long-term spending commitments, its access to reliable computing capacity, customer concentration and ability to turn revenue growth into sustainable profits. These factors could influence how investors assess the business and its valuation.