It is rare for a technology company to give the public a glimpse under its own hood. OpenAI has managed to do so twice in recent weeks. First, an internal memo intended for investors was leaked, in which it sharply attacked Anthropic. Then it emerged that Chief Revenue Officer Denise Dresser had sent employees a four-page letter in which she was not afraid to include a few brutally honest lines about Microsoft. Both memos were ultimately read by the entire technology industry and media outlets such as CNBC.
Anthropic and Microsoft are a thorn in their side
Let's start at the beginning. In early April, OpenAI sent an internal memo to its investors. According to CNBC, the company described Anthropic as a company that "operates on a significantly smaller compute curve" and is running into a shortage of computing power. In plain English: Anthropic allegedly does not have enough servers to keep pace over the long term.
OpenAI plans to reach 30 gigawatts of computing capacity by 2030. And Anthropic? According to OpenAI's estimates, it should have around 7 to 8 gigawatts by the end of 2027. In OpenAI's words, "our lead is significant and continues to grow."
Another leaked memo from Denise Dresser, whom OpenAI poached from Slack for the position of Chief Revenue Officer only last December, was sent to employees and ran to four pages, including the sentence: "Our partnership with Microsoft has been foundational to our success. But it has also limited our ability to reach businesses where they are."
Microsoft has poured more than $13 billion into OpenAI since 2019. It was Microsoft that provided OpenAI with computing infrastructure when ChatGPT did not even exist yet. And now its own protégé is publicly, albeit indirectly, describing it as an obstacle. Microsoft declined to comment on the matter.
Amazon to the rescue?
In the same letter, Dresser presented Amazon as a key ally for the future. At the end of February 2026, Amazon announced an investment of up to $50 billion in OpenAI as part of a strategic partnership. The companies want to make OpenAI models available through Amazon Bedrock, a platform already used by thousands of businesses around the world. "Since we announced this partnership, customer interest has been nothing short of astounding," Dresser wrote.
But this partnership comes with a catch. Microsoft is considering legal action. Its contract with OpenAI states that access to all models must go through Azure, Microsoft's cloud platform. Amazon and OpenAI devised a technical workaround to this rule, known as the Stateful Runtime Environment. However, Microsoft engineers claim that this solution does not comply with the contract.
Dresser versus Anthropic
The memo also addressed its direct competitor. Anthropic recently announced that its annualized revenue had exceeded $30 billion a year. Dresser responded with a considerable degree of skepticism. She claims that this figure is inflated by roughly $8 billion due to the way revenue from partnerships with Amazon and Google is accounted for.
Anthropic reportedly includes amounts flowing through cloud partners in its revenue as though it were the principal seller. OpenAI, by contrast, records only the net margin after deducting Microsoft's share. Dresser considers her approach more appropriate, especially in view of the planned IPO (stock market listing). Anthropic defends its approach as fully compliant with GAAP accounting rules. It argues that it genuinely acts as the principal party in these transactions, rather than merely as an intermediary.
Dresser acknowledges in the memo that Anthropic has gained a lead in the enterprise customer segment. At the HumanX conference in San Francisco, which took place last week, Anthropic's Claude model was the hottest topic, according to CNBC.
OpenAI is responding with its own strategy. Enterprise customers account for 40% of the company's revenue, and Dresser wants enterprise revenue to equal consumer revenue by the end of 2026. The strategy is clear: stop operating as a company with separate products and offer a unified platform with multiple entry points. "Let's stop thinking like a company with separate product lines. Let's think like a platform," Dresser wrote. She also took a jab at the rival: "Anthropic is built on fear and restriction. A positive vision will move us forward."
Behind all of this is also a purely pragmatic question. OpenAI is currently valued at more than $850 billion, while Anthropic is valued at $380 billion. Both companies are preparing to go public, probably later this year. And investors understandably want to know which of them has the more sustainable business model.
OpenAI knows that dependence on a single partner (namely Microsoft) is a weak point. Hence Amazon, hence the aggressive tone toward Anthropic, and perhaps hence the "deliberately" leaked internal memos.
