Anthropic is broadening its partnerships with major U.S. technology companies, including Microsoft (MS) and Amazon Web Services (AWS).
The company's collaboration with Microsoft and NVIDIA--part of a strategic alliance valued at approximately $45 billion--has begun delivering tangible results. At the same time, signs of tension have emerged in its relationship with Amazon, which has reportedly been linked to the U.S. government's decision to restrict foreign access to the “Fable5” family of AI models.
On July 29 (local time), Anthropic, NVIDIA, and Microsoft announced that Anthropic's Claude family of AI models is now generally available on Microsoft Azure cloud infrastructure powered by NVIDIA Blackwell Ultra GPUs. Previously, Azure customers accessed Claude through external APIs; with this launch, Claude can now run natively within Azure.
The three companies said native deployment on Azure addresses cross-border data transfer compliance concerns, making it easier for organizations that require AI services to remain within specific geographic regions to adopt Claude.
The move also enables Anthropic to expand its enterprise customer base beyond AWS and Google Cloud by serving organizations that rely on Microsoft Azure. Azure customers will have access to the flagship Claude Opus 4.8, the lightweight Claude Haiku 4.5, and Anthropic's extended reasoning capabilities.
The deployment is part of a strategic partnership announced in November last year. Under the agreement, NVIDIA and Microsoft committed investments of $10 billion and $5 billion, respectively, in Anthropic, while Anthropic agreed to purchase approximately $30 billion worth of Microsoft cloud services powered by NVIDIA hardware.
For Microsoft, adding Anthropic's models alongside its longstanding partnership with OpenAI strengthens Azure's AI portfolio and helps reduce the risk of enterprise customers migrating to competing cloud platforms.
While cooperation with Microsoft has deepened, Anthropic's relationship with Amazon appears to be facing new challenges. According to The Information, the two companies are renegotiating AI model pricing, a move that could significantly increase Amazon's costs.
Earlier this year, the companies reportedly revised parts of their agreement so that, beginning next year, Amazon will be charged based on the number of tokens processed rather than compute time.
A token is the basic unit used by AI models to process text, with one English word averaging about 1.3 tokens. Because Amazon integrates Anthropic's AI into a wide range of consumer and enterprise products--including Alexa Shopping, Kiro for coding, and Quik for workplace productivity--the pricing change could substantially increase operating costs.
Industry observers also speculate that the timing coincides with the U.S. government's restrictions on exports of Anthropic's Fable5 models. Reports suggest that Amazon's disclosure of potential security concerns played a significant role in the government's decision to prohibit foreign access to Anthropic's flagship AI models, including Mythos5 and Fable5.
Both companies denied reports of a deteriorating partnership. An Amazon spokesperson said, “Amazon and Anthropic share a broad, technology-driven partnership, and we will continue to strengthen our relationship going forward.” Anthropic likewise stated that “Amazon remains one of our most important partners.”
