OpenAI Challenges Anthropic: A Numbers Battle Ahead of the IPO

OpenAI and Anthropic: A Rivalry Intensified by an Internal Memo

A recently leaked internal document from OpenAI sheds light on the growing tension between two giants of artificial intelligence. This memo, written by Denise Dresser, Chief New Revenue Officer of OpenAI, directly targets Anthropic, a major competitor. As both companies prepare to go public, their rivalry is escalating, with accusations that go beyond mere business strategies.

The memo was revealed by The Verge this weekend. Denise Dresser harshly criticizes Anthropic, not only outlining OpenAI's priorities for the upcoming quarter but also dedicating a significant portion of the document to attacking its rival.

Contested Revenues: A Battle of Numbers

At the heart of this confrontation lies a divergence over reported revenues. Anthropic recently announced an annualized revenue of $30 billion, an impressive figure compared to the $9 billion projected for the end of 2025. However, Dresser claims that this amount is inflated by about $8 billion.

The controversy hinges on differences in how revenues are accounted for. Anthropic includes commissions from platforms like AWS, Azure, and Google Cloud in its figures, which constitutes gross revenue. In contrast, OpenAI reports its sales on a net basis, deducting partner shares before recording them in its accounts.

While both approaches comply with U.S. accounting standards, they lead to different rankings. Anthropic, viewing its cloud platforms as mere distribution channels, sees itself as the primary player in these transactions. If the net method is applied, its revenues would drop to $22 billion, below the $24-25 billion claimed by OpenAI.

Furthermore, the memo highlights a significant difference in infrastructure. An investor document forecasts that OpenAI will reach a computing capacity of 30 gigawatts by 2030, compared to only 7 to 8 gigawatts for Anthropic by 2027. Dresser describes this lag as a "strategic fault" already evident in Anthropic's products.

Opposing Visions for the Future of AI

Beyond the numbers, the memo reveals an ideological clash. Dresser accuses Anthropic of promoting a vision of AI based on fear and restriction, where a small group of elites would control the technology. In contrast, OpenAI positions itself as a champion of accessibility, advocating for powerful systems with appropriate safeguards.

This criticism is not new. Sam Altman, the CEO of OpenAI, had previously stated that Anthropic sells expensive products to a privileged clientele. The memo reinforces this line of attack as both companies approach their planned IPOs. OpenAI aims for a valuation of one trillion dollars by the fourth quarter of 2026, while Anthropic could go public as early as October.

However, Dresser acknowledges that Anthropic currently dominates the enterprise code segment. According to data from Ramp, Anthropic holds 30.6% of the paid AI customer market, compared to 35.2% for OpenAI. A trend reversal could occur soon.

Strategies and Internal Tensions at OpenAI

The memo also highlights internal tensions within OpenAI. Dresser admits that the historic partnership with Microsoft has limited the company's ability to reach certain businesses. However, she presents the new agreement with Amazon, which includes up to $50 billion in investment, and the Bedrock platform as solutions to this issue.

The timing of this memo's release on a Sunday is not coincidental. It aims to establish a narrative before the markets seize upon it, underscoring the strategic importance of this communication for OpenAI.