OpenAI Shifts from Microsoft to Amazon AWS, Reshaping the Cloud AI Landscape

OpenAI, once Microsoft's closest ally, is increasingly moving toward Amazon.

On April 27, Western US time, openai and Microsoft simultaneously announced that Microsoft Azure would no longer hold exclusive rights to host and distribute OpenAI's products. The following day, on April 28, OpenAI and Amazon jointly announced that OpenAI's GPT models and Codex, its code generation platform, would become available through Amazon Web Services (AWS).

As one of the world's largest Foundation Model companies, OpenAI spends over $16 billion annually on cloud computing. It is a major customer for both Amazon and Microsoft. Given that OpenAI accounts for more than 10% of Microsoft Azure’s revenue, its strategic choices directly inFluence the competitive balance in the cloud market. With OpenAI's pivot, Microsoft Azure’s growth momentum is expected to slow, while Amazon AWS reInforces its position as the market leader.

The logic behind OpenAI’s decision is straightforward: it is starved for computing power. Relying solely on one cloud provider is no longer sustainable. On April 28, media reports revealed that OpenAI recently failed to meet its internal targets for new users and revenue. CEO Sam Altman has identified compute shortages as the biggest bottleneck restricting OpenAI’s growth, stressing the need to secure computing resources by all means necessary.

Another critical fACTor is mounting competitive pressure. OpenAI's revenue and Valuation are being overtaken by rival anthropic. By leverAGIng Amazon AWS’s platform and distribution resources, OpenAI aims to unlock new revenue streams. Additionally, to achieve greater cost efficiency, distribution reach, and strategic autonomy, OpenAI must break its single-provider dependency on Microsoft and embrace a Multi-Cloud Strategy.

This shift toward Amazon is the result of a gradual evolution that began in November 2025, marked by four key milestones:

• October 28, 2025: OpenAI and Microsoft signed a seven-year, $250 billion cloud services agreement, under which Microsoft forfeited its right of first refusal as OpenAI's compute provider. The two companies began to loosen their ties.

• November 3, 2025: OpenAI and Amazon signed a seven-year, $38 billion compute contract.

• March 1, 2026: OpenAI completed a $110billionfundinground\mathrm{.}Amazoncommittedaphasedinvestmentof$50 billion, starting with an initial $15billionandanadditional$35 billion contingent on meeting certain conditions. Both parties also added an eight-year, $100billioncomputecontract,bringingtheirtotalcomputeagreementsto$138 billion.

• April 28, 2026: Microsoft and OpenAI jointly confirmed Azure’s loss of exclusivity and the end of revenue-sharing payments from Microsoft to OpenAI for models running on Azure. Simultaneously, Amazon announced that OpenAI's foundation models would be available on AWS.

Over these six months, OpenAI gained greater flexibility and momentum through diversified partnerships. Freed from previous constraints, OpenAI is now positioned to compete head-to-head with Anthropic.

Countering Pressure from Anthropic

A key backdrop to this strategic realignment is the intensifying competition from Anthropic. Media reports on April 28 indicated that OpenAI is falling short of revenue and user growth targets during a critical pre-IPO phase. It missed monthly revenue goals multiple times earlier this year and failed to reach 1 billion weekly active users for ChatGPT by the end of last year.

OpenAI was historically the highest-revenue and highest-valued AI startup. However, since March 2026, it has lost its lead in both categories. By April 2026, OpenAI’s annualized recurring revenue (ARR) exceeded $25billion,withavaluationof$853 billion. Meanwhile, Anthropic surpassed OpenAI with an ARR exceeding $30billionasofMarch2026andavaluationreaching$1 trillion by April.

A major driver of Anthropic's rise is its claude model family’s sustained edge in coding capabilities. Coding is essential in high-value enterprise scenarios. IndiVidual users pay $20permonthforChatGPTsubscriptions,butindividualdevelopersoftenpay$200 monthly for code generation tools, and API-based usage can range from $2,000to$20,000 per user per month. With the Agent boom in 2026, the importance of coding has magnified, as agents rely on code generation to plan and execute tasks. Coding proficiency directly determines task success rates, execution time, and Token Consumption. In multi-task scenarios, differences in coding ability are exponentially amplified, leading to stark gaps in efficiency, cost, and user experience.

As a result, Anthropic’s Claude models have rapidly expanded from developer communities to broader enterprise and consumer audiences. Products like Claude Code and Claude Cowork have emerged as leading agent tools. According to Similarweb, Claude’s daily active users surged from 4 million in early 2026 to 11.3 million by early March. On the model aggregation platform OpenRouter, which reflects preferences among frontier developers, Anthropic’s Claude models accounted for 14.8% of total token consumption (37.27 trillion tokens) between January 4 and April 26, 2026, compared to 11.2% (28.24 trillion tokens) for OpenAI's GPT models.

These trends have left OpenAI at a competitive disadvantage, but the company is taking countermeasures. On April 23, OpenAI launched GPT-5.5, which surpassed Anthropic's Claude Opus 4.7 in coding benchmarks. Moreover, the Partnership with Amazon AWS is itself a strategic response. As of March 2026, AWS Bedrock platform hosted over 100,000 enterprise customers and previously served as Anthropic's largest third-party distribution channel. By joining AWS Bedrock, OpenAI effectively secures access to a key partner and sales channel that once belonged primarily to its rival.

Tilting the Balance Between Amazon and Microsoft

OpenAI operates over 50 million individual paid users and 1 million enterprise CLIents globally. The cloud provider it favors will see accelerated revenue growth. The 2026 AI Index Report from Stanford HAI estimated OpenAI’s cloud spending at roughly $16.3billionin2025,withover$10 billion paid to Microsoft Azure and several billion to Amazon AWS.

OpenAI is one of Microsoft Azure's top two customers and among Amazon AWS’s top five, representing over 10% of Azure’s revenue and roughly 5% of AWS’s revenue. From 2023 to 2025, Azure, as OpenAI's exclusive compute partner, narrowed the market share gap with AWS, attracting a wave of customers migrating to access OpenAI models. In 2025, AWS revenue reached $128.7billion(up19.7$120.4 billion (up 25.9%), with its new incremental revenue at one point surpassing AWS’s.

Now, with Microsoft losing exclusivity and Amazon gaining both a massive compute customer and the rights to sell OpenAI's models, AWS is regaining the upper hand. If OpenAI fulfills its eight-year, $138billionpurchasecommitment,itcouldgenerateover$10 billion in new annual revenue for AWS, potentially boosting AWS’s short-term revenue growth rate by over 10 percentage points. Moreover, just as OpenAI previously attracted new clients to Azure, its presence on AWS Bedrock—alongside models from Anthropic, Meta, DeepSeek, and others—will now offer enterprise customers the advanced OpenAI models they previously could only access via Azure.

By distancing itself from Microsoft and drawing closer to Amazon, OpenAI is set to slow Azure’s pursuit of AWS and help Amazon maintain its lead in the global Cloud Computing market for the foreseeable future.