OpenAI and Anthropic are competing on the front stage, while Microsoft and Amazon are going head - to - head behind the scenes.

The AI competition is no longer just a simple business cooperation but a power game around the technological high - ground in the next decade. In this competition, there are no eternal allies, only eternal capital and interests.

The AI (Artificial Intelligence) market in the United States is witnessing a confrontation between two major camps.

On the front stage are the world's two largest AI startups, OpenAI and Anthropic. Behind the scenes are two technology giants, Microsoft and Amazon, which are also the world's top two cloud providers, having long held more than 60% of the cloud market share.

How did the confrontation pattern between the two major camps come into being?

Today's AI competition is essentially a competition of computing power and models. Microsoft and Amazon need to exchange computing power and capital for the models and technologies of startups to gain a larger market. OpenAI and Anthropic need the support of the giants to grow rapidly. This is also the basis for the cooperation between the two pairs of allies, Microsoft and OpenAI, Amazon and Anthropic.

In August and September this year, OpenAI and Anthropic completed a new round of financing successively, becoming the world's third - and fourth - largest unicorns respectively (the first is SpaceX with a valuation of $400 billion, and the second is ByteDance with a valuation of $315 billion).

OpenAI has raised more than $79.7 billion in total and is valued at $300 billion. Microsoft has invested at least $13 billion in OpenAI, accounting for more than 16% of its publicly disclosed total financing. Anthropic has raised more than $31.2 billion in total and is valued at $183 billion. Amazon has invested at least $8 billion in Anthropic, accounting for more than 25% of its publicly disclosed financing.

Microsoft began investing in OpenAI in 2019, and the two initially had an exclusive cooperation relationship. The computing power for OpenAI to train and run large models is mainly provided by Microsoft. Microsoft integrates OpenAI's models into its products. According to the agreement between the two parties, Microsoft can obtain 49% of OpenAI's profits until it recovers its $13 - billion investment.

After Amazon invested in Anthropic in September 2023, the two gradually became strategic allies. Amazon promotes Anthropic's large models to its customers. Anthropic mainly trains and runs large models based on Amazon AWS and also deeply uses two self - developed AI chips of Amazon, Trainium (training chip) and Inferentia (inference chip).

OpenAI and Anthropic have driven the revenue growth and market share of Microsoft Azure and Amazon AWS. The global cloud market share of the two cloud providers has further increased from 60% in 2021 to 63% in 2023.

For Chinese technology giants and startups, the two pairs of allies, Microsoft and OpenAI, Amazon and Anthropic, must be closely monitored. Their competitive and cooperative relationships not only dominate the development direction of the global cloud computing and large - model markets but also the development direction of global artificial intelligence.

However, in the AI competition, there are no eternal allies, only eternal computing power, capital and interests. The formation and dissolution of alliances change due to market competition. After 2025, the relationship between Microsoft and OpenAI has gradually become estranged, and the cooperation between Amazon and Anthropic is not unbreakable either.

Currently, the market is concerned about several core issues.

Firstly, what has the cooperation between Microsoft and OpenAI brought to both parties?

Secondly, can Amazon and Anthropic contain Microsoft's ambition to compete for the cloud - computing hegemony?

Thirdly, what are the variables in the cooperation between Microsoft and OpenAI, Amazon and Anthropic?

Microsoft and OpenAI, on the Offensive

The in - depth cooperation between Microsoft and OpenAI is almost the starting point of this wave of large - model development.

OpenAI's commercial success in just two or three years is inseparable from Microsoft's endorsement, which has shortened OpenAI's commercialization process. Microsoft is also a beneficiary. With an investment of only $13 billion, it has leveraged OpenAI's cloud spending of tens of billions of dollars on Microsoft Azure in the next few years and made it possible to challenge Amazon AWS. OpenAI has brought not only tens of thousands of new customers to Microsoft Azure but also a brand - new, high - value growth path.

OpenAI was founded in 2015, and Microsoft made two investments in 2019 and 2023. Starting early gave OpenAI a first - mover advantage. As of the end of July this year, OpenAI's annualized revenue (monthly revenue × 12. Software companies usually charge by annual/monthly subscriptions, so this statistical method is applicable) was about $12 billion, a year - on - year increase of more than 100%. As of the end of August this year, Anthropic's annualized revenue was about $5 billion.

Currently, OpenAI's financing scale, valuation scale, and revenue scale are all higher than those of Anthropic. OpenAI's revenue is 2.4 times that of Anthropic, its financing scale is 2.6 times that of Anthropic, and its valuation is 1.6 times that of Anthropic.

Microsoft not only provides funds and computing power to OpenAI but also builds a sales channel. Microsoft has long been deeply involved in the enterprise and government markets and has an enterprise sales network established over decades, which enables OpenAI's models to reach millions of customers globally. Applications such as Microsoft 365 Copilot (Word, Excel, Teams AI assistant) and GitHub Copilot under Microsoft have quickly made OpenAI's models office tools used by hundreds of millions of users worldwide.

OpenAI has given Microsoft Azure the opportunity to challenge Amazon AWS's dominant position in the cloud - computing market. It is not only a major customer of Microsoft Azure but also has activated Microsoft's cloud - computing business.

OpenAI's model training and inference consume a huge amount of computing resources, and most of these costs are paid to Microsoft. Media reports suggest that OpenAI is expected to pay up to $13 billion in cloud costs to Microsoft Azure in 2025. Both Morgan Stanley and Goldman Sachs predicted in July this year that Microsoft Azure's revenue would exceed $100 billion in 2025. Based on this data, OpenAI's share in Microsoft Azure's total revenue may exceed 10%.

New customers have started to migrate to Microsoft Azure to use OpenAI's models, which has brought in a large amount of new revenue. In a past earnings conference call, Microsoft's management disclosed in early 2024 that more than one - third of the 53,000 AI customers of Microsoft Azure were new customers. According to the information disclosed by Microsoft's management, the contribution of AI to the growth of Microsoft Azure has increased from 3 percentage points in the third quarter of 2023 to 16 percentage points in the second quarter of 2025.

After the cooperation between OpenAI and Microsoft, all of Microsoft's products have integrated GPT - series models and have been upgraded. The large models have helped Microsoft Azure attract more enterprise customers and promoted the sales of basic cloud products such as computing, storage, networking, and databases, forming a positive cycle of "selling multiple products to one customer". The revenue growth rate and market share of Microsoft Azure have rapidly increased, continuously narrowing the gap with Amazon AWS.

Microsoft has always hidden Microsoft Azure's revenue under the complex statistical category of "Microsoft Intelligent Cloud". A senior executive of a Chinese cloud provider who knows Microsoft well once told Caijing that from a statistical strategy perspective, "Microsoft Intelligent Cloud" is a financial category used by Microsoft over the years to cover up the revenue gap between Microsoft Azure and Amazon AWS.

Microsoft Intelligent Cloud's revenue in fiscal year 2025 (July 1, 2024 - June 30, 2025) was $106.3 billion, approaching that of Amazon AWS. However, a long - standing view in the cloud - computing industry is that its revenue quality lags behind that of Amazon AWS. Because "Microsoft Intelligent Cloud" includes not only Microsoft Azure but also traditional businesses such as server - related software products, GitHub cloud services, Nuance healthcare cloud services, and digital consulting services.

After the successful bet on AI, Microsoft Azure has truly emerged. Microsoft disclosed Microsoft Azure's revenue for the first time in July this year - $75 billion in fiscal year 2025, a year - on - year increase of 34%, with new revenue of $25.5 billion. During the same period, Amazon AWS's revenue was $116.4 billion, a year - on - year increase of 18%, with new revenue of $21 billion. This means that Microsoft Azure's new revenue in the past fiscal year has exceeded that of Amazon AWS, and the revenue gap between the two is rapidly narrowing.

The gap in their market shares is also narrowing. Data from international market research firm Gartner shows that Microsoft Azure's market share has gradually increased in the past three years, and the gap with Amazon AWS has narrowed to 12%.

Microsoft CEO Satya Nadella explained the reason for Microsoft Azure's rapid growth in an earnings conference call after the fiscal year 2025 report: AI has accelerated the migration of traditional businesses to the cloud, led to the rapid development of cloud - native and AI applications, and also increased the consumption of computing power. The three factors complement each other, driving the growth of Microsoft Azure.

A strategic planning expert from a leading Chinese cloud provider told Caijing in May this year that Microsoft's financial reports have shown that the direct revenue brought by AI is limited, but it has increased the demand for computing - power consumption, data storage, and transmission. Therefore, the revenue growth of the four major components of cloud computing (computing, storage, networking, and databases) is relatively high. It doesn't make much sense to strictly distinguish between "AI business" and "non - AI business" because all AI businesses will ultimately run on the cloud, driving the consumption of cloud resources.

Another significant advantage of Microsoft in the cloud - computing field is its "well - rounded" product line. Collectively known as "Microsoft Cloud" (including Microsoft Azure + software for productivity processes), it had a revenue of $168.9 billion in fiscal year 2025, a year - on - year increase of 26.3%.

Since the traditional IT era, Microsoft has spent more than 40 years building a complete To - B business line. It has public cloud, hybrid cloud, and private cloud capabilities, as well as integration consulting and service capabilities. At the IaaS (Infrastructure as a Service) layer, there is Microsoft Azure; at the PaaS (Platform as a Service) layer, there are products such as Dynamic ERP and the Power Platform development platform; at the SaaS (Software as a Service) layer, there are collaborative office products such as Teams and Office 365.

In the AI era, these businesses can still meet the needs of almost any enterprise customer. Microsoft is also using large models to transform these products to achieve AI transformation and business growth.

Amazon and Anthropic, Fighting Back with a Tight Bond

The aggressive offensive of Microsoft and OpenAI has put Amazon AWS under intense competitive pressure. Wall Street has even started discussing the "cloud - computing crisis" of Amazon AWS. Amazon has not remained idle and has quickly launched a powerful counter - attack.

After OpenAI emerged in early 2023, Amazon invested $4 billion in Anthropic in September 2023 and an additional $4 billion in November 2024. Amazon needs a "bridgehead" in the competition with Microsoft.

In the face of the fierce offensive of Microsoft Azure, the cooperation between Amazon AWS and Anthropic first plays a strategic defensive role, stabilizing potential customers who might otherwise leave. More importantly, it has attracted a group of new customers who have diverse requirements for model selection or are reluctant to choose the Microsoft ecosystem due to reasons such as data privacy, AI security, and cost.

Anthropic has rapidly grown into the most powerful competitor to OpenAI in just two years. Although there are gaps in its financing scale, valuation scale, and revenue scale compared with OpenAI, its differentiated advantage lies in targeting the To - B (business - to - business) market. 80% of Anthropic's revenue, that is, $4 billion, comes from enterprise subscriptions and API (Application Programming Interface) calls. It is also growing faster than OpenAI. In January 2025, Anthropic's annualized revenue was only $1 billion, but it climbed to $5 billion by the end of August, a growth of 400%.

Anthropic's precise positioning has enabled it to keep up with OpenAI in the competition and even take the lead in some important niche markets.

The founding team of Anthropic left OpenAI because they believed that OpenAI did not pay enough attention to AI security when pursuing Artificial General Intelligence (AGI). Anthropic's strategic positioning at its inception was "reliable, interpretable, and controllable", which has attracted industries under strict supervision such as finance, law, healthcare, and government.

Anthropic can always launch a comparable model within a month or even a few weeks after OpenAI releases a new model. Anthropic also has a "killer feature" - code - generation ability. The code generated by models such as Claude 3.7 Sonnet, Claude Sonnet 4, and Claude Opus 4.1 of Anthropic is stable and has few errors, and has a good reputation among programmers. Most programmers who use independent code - generation tools such as Cursor prefer Claude.