Nearly half of Anthropic’s 2025 revenue flowed through Amazon and Google cloud marketplaces, highlighting the AI company’s rapid growth but also its exposure to a small group of powerful technology partners

IPO filing exposes the economics behind Anthropic’s rapid expansion

Anthropic’s confidential IPO filing has provided investors with a rare look into the financial structure of one of the world’s fastest-growing artificial-intelligence companies — and it highlights a significant dependence on the same Big Tech companies that simultaneously invest in, supply infrastructure to and compete with the AI developer.

According to documents reviewed by Reuters, around 47 per cent of Anthropic’s 2025 revenue, or approximately $2.16 billion, came through cloud marketplaces operated by Amazon and Alphabet’s Google. The two companies are not only major distribution partners but also investors in Anthropic and critical suppliers of computing capacity. At the same time, Amazon and Google operate their own AI businesses and compete for enterprise AI spending.

The structure gives Anthropic access to enormous enterprise sales networks, but it also creates financial and strategic dependencies that potential public-market investors will have to evaluate as the company moves toward what could become one of the largest technology IPOs ever.

Revenue has exploded, but losses remain substantial

Anthropic’s growth has been extraordinary.

Revenue surged roughly 12-fold in 2025 to nearly $4.6 billion, according to Reuters. However, the company's operating losses more than doubled and exceeded $8 billion during the same period.

The company is seeking a valuation of around $2 trillion, according to Reuters, putting enormous expectations on future growth and profitability.

Anthropic's business is primarily driven by usage of its Claude AI models. Around $3.8 billion of its 2025 revenue came from consumption-based usage, while subscription revenue was approximately $789 million.

The company expects consumption-based revenue to remain the dominant component of its business for the foreseeable future.

This model gives Anthropic a potentially powerful link between customer usage and revenue, but it also means sales can fluctuate if enterprises reduce AI workloads, optimise usage or switch to competing models.

Amazon and Google have become crucial distribution channels

The concentration of revenue flowing through cloud marketplaces has increased rapidly.

Sales through Amazon and Google represented around 11 per cent of Anthropic's revenue in 2023. That proportion increased to approximately 32 per cent in 2024 and reached about 47 per cent in 2025.

Cloud marketplaces effectively allow enterprises to purchase Claude through infrastructure and procurement relationships they already maintain with major cloud providers.

For Anthropic, this can significantly reduce the difficulty of building a global direct sales operation.

Instead of persuading every large enterprise to establish a separate commercial relationship with an AI company, Anthropic can reach customers through existing cloud contracts.

The advantage is particularly important in enterprise technology, where procurement, security, compliance and infrastructure requirements can make direct adoption of a new AI platform more complicated.

The price of distribution is significant

The same arrangement also comes at a cost.

Reuters' analysis of the filing estimates that Anthropic paid approximately $351 million in distribution fees to cloud platforms in 2025. That implies that cloud partners collected roughly 16 cents for every dollar of sales routed through their marketplaces.

Anthropic records these channel-partner costs within its sales, marketing and partnerships expenses.

The economics therefore involve a trade-off: Anthropic gets access to a much larger customer network, but gives up part of its revenue to the companies operating those channels.

For an AI business already spending heavily on computing infrastructure, distribution costs add another layer to the challenge of eventually reaching sustainable profitability.

Big Tech is simultaneously partner, investor and competitor

The most unusual aspect of Anthropic's structure is the overlap between its commercial relationships and competitive landscape.

Amazon and Google have invested billions of dollars in Anthropic. At the same time, both companies provide infrastructure needed to train and operate Claude and distribute the model to enterprise customers.

They also operate competing AI platforms.

This creates a complicated set of incentives. Anthropic benefits from the scale and reach of its partners, while those same partners have their own AI products, cloud businesses and strategic priorities.

Anthropic acknowledged in its filing that dependence on a limited number of partners and suppliers can create conflicts of interest and potentially affect its access to computing resources.

Compute commitments are enormous

Anthropic's growth strategy requires extraordinary amounts of computing infrastructure.

At the end of 2025, the company had approximately $54.6 billion in non-cancellable hosting and computing commitments, according to Reuters. By early 2026, its total long-term commitments had exceeded $417 billion, covering around 3.5 gigawatts of dedicated computing capacity.

Separate company disclosures show just how rapidly its infrastructure requirements have expanded.

In April, Anthropic announced an agreement with Amazon involving up to five gigawatts of additional computing capacity and said it was committing more than $100 billion over 10 years to AWS technologies. Amazon also announced a $5 billion investment in Anthropic, with the possibility of another $20 billion.

Anthropic has also expanded its relationship with Google and Broadcom for multiple gigawatts of next-generation TPU capacity expected to come online from 2027.

The scale of these commitments demonstrates both the extraordinary demand for AI compute and the capital intensity of operating frontier AI models.

Microsoft adds another major cloud relationship

Anthropic has also broadened its cloud distribution network beyond Amazon and Google.

Microsoft joined the company's cloud-computing ecosystem through a deal announced in November, according to the IPO filing cited by Reuters.

Anthropic has said Claude is available through Amazon Web Services, Google Cloud and Microsoft Azure, giving enterprise customers multiple routes to access the models.

The diversification could reduce dependence on any single infrastructure provider over time.

However, it does not eliminate the broader issue: Anthropic remains dependent on external cloud infrastructure and enormous amounts of computing capacity to deliver its products.

Customer concentration is another risk

The IPO filing also highlights concentration on the demand side.

Two unnamed customers each accounted for approximately 12 per cent of Anthropic's revenue in 2025. The company also warned that many of its largest customers do not have long-term contractual commitments and could reduce or stop spending.

That makes customer retention a crucial part of Anthropic's growth equation.

The company's usage-based model can generate rapid revenue expansion when customers increase AI workloads. But the same model could make revenue more sensitive to changes in usage patterns, technology budgets and competitive offerings.

Cloud partners also handle a large portion of collections

Anthropic's dependence on its cloud partners extends beyond distribution.

The companies were responsible for collecting around 60 per cent of Anthropic's $909 million in outstanding customer bills at the end of 2025, compared with 42 per cent a year earlier, according to Reuters.

This means delays, disputes or operational issues within those payment channels could potentially affect Anthropic's cash flow, even where the underlying customer relationship is directly with Anthropic.

For investors assessing a high-growth company preparing to enter public markets, working-capital dynamics can be as important as headline revenue growth.

Revenue accounting becomes another IPO talking point

Anthropic's accounting treatment for cloud marketplace sales has also attracted attention.

The company records the full value of marketplace contracts as revenue because it considers itself the principal providing the AI service. The share retained by cloud platforms is recorded as a sales and marketing expense.

That approach results in a higher reported revenue figure than if Anthropic recorded only the net amount after cloud-platform fees.

Reuters reported that OpenAI has criticised this methodology, arguing that it makes Anthropic's reported revenue appear billions of dollars larger than an alternative presentation would.

Anthropic has defended its approach, saying it follows established accounting principles and is the principal in these transactions.

The distinction will matter to investors because revenue comparisons between fast-growing AI companies can become difficult when their distribution structures and accounting treatments differ.

The bigger AI infrastructure race

Anthropic's filing also illustrates a broader feature of the current AI industry: leading model developers are becoming deeply intertwined with cloud providers, chip companies and infrastructure operators.

Anthropic needs enormous computing resources to train and serve increasingly capable models. Cloud companies, meanwhile, benefit from the infrastructure demand generated by AI developers.

This creates a mutually reinforcing cycle in which investment flows from technology companies into AI developers, while AI developers commit enormous sums back to the same technology ecosystem for computing capacity.

Anthropic has attempted to diversify its infrastructure by using AWS Trainium, Google TPUs and Nvidia GPUs. The company has said that this multi-platform approach allows workloads to be matched with different hardware and can improve resilience.

IPO investors will focus on growth versus capital intensity

The central investment question surrounding Anthropic is whether its extraordinary revenue growth can eventually translate into sustainable profits.

The company has demonstrated the ability to scale revenue rapidly, but frontier AI development requires enormous spending on chips, data centres, electricity, engineering talent and cloud infrastructure.

A potential $2 trillion valuation would therefore require investors to look beyond current revenue and assess the size of the future AI market, customer retention, pricing power, model costs, infrastructure commitments and the competitive position of Claude.

The company's public-market story will ultimately depend on whether increasing AI adoption can outpace the cost of delivering increasingly sophisticated models.

Market Outlook

Anthropic's IPO filing highlights both sides of the AI investment story.

On one side, the company is experiencing extraordinary demand, with revenue rising roughly 12-fold in 2025 and Claude becoming a significant enterprise AI platform. Its presence across AWS, Google Cloud and Microsoft Azure gives it access to some of the world's largest corporate technology ecosystems.

On the other side, nearly half of its 2025 revenue flowed through Amazon and Google, while enormous long-term computing commitments and multibillion-dollar operating losses underline the capital required to compete at the frontier of AI.

For prospective investors, the key metrics will be revenue growth, consumption trends, customer concentration, gross margins, cloud distribution costs, infrastructure commitments and the path toward positive operating cash flow.

The IPO could therefore become an important test of how public markets value frontier AI companies whose growth prospects are enormous but whose infrastructure requirements and dependence on Big Tech remain equally significant.