The Memo OpenAI Didn't Want to Write
On April 9, 2026, OpenAI distributed a memo to its shareholders directly attacking its rival, Anthropic, describing the company as operating on a "meaningfully smaller" scale. While this statement is technically accurate—invoking scaling laws of language models where performance increases with compute and data—it also serves as a warning sign.
Market leaders with established positions typically do not feel the need to explain to their shareholders why the second player is inferior. Such a tactic is employed by leaders who sense a closing gap. The fact that Anthropic has reached a valuation of over $60 billion, that its Claude models are being integrated by companies like Salesforce and Notion, and that over 430 employees from Google and OpenAI signed a public letter supporting Anthropic's red lines in contracts with the Pentagon cannot simply be resolved with a memo; they require strategy.
The Rejection Anthropic Turned into Positioning
The catalyst for this cycle of tension was a contract with the U.S. Department of Defense. Anthropic rejected signing an agreement that allowed its technology to be used for "any lawful purpose," insisting instead on two specific exceptions: its AI would not be used for fully autonomous weapons or for mass domestic surveillance. OpenAI accepted the agreement under broader terms but later admitted that the initial contract "seemed careless and opportunistic" and negotiated additional restrictions.
From my perspective as a strategist, this moment is the most telling of the story. Anthropic sacrificed a $200 million government contract—not postponing it, not renegotiating it to the limit, nor seeking vague language that would allow it to sign with a clear conscience. It flatly rejected the contract. This renunciation has an immediate and tangible financial cost but also yields a positioning effect money cannot buy: the perception of alignment between what the company declares and what it acts upon.
Anthropic's CEO, Dario Amodei, made a tactical misstep in turning that moment of strength into ammunition for his critics by leaking an internal memo describing OpenAI's team as "credulous" and its followers as "Twitter idiots." The memo was crafted, according to his later explanation, shortly after a chaotic series of announcements. His public apology was straightforward: he acknowledged that the tone did not reflect his thoughtful positions. The reputational damage was real, particularly among the 430 signatories to that letter who had backed him publicly just days prior.
Yet something significant remains unchanged despite that communication slip: the underlying strategic decision still stands. Anthropic did not reverse its position. It did not sign the contract under pressure. Consistency between decision and guiding principle determines whether a company has a strategy or merely good intentions.
What OpenAI's Memo Tells Its Shareholders
Returning to OpenAI's memo, it is the most intriguing document in this narrative. A company that usually avoids naming competitors in public communications explicitly attacked Anthropic for its scale in a communication directed at its shareholders. The apparent logic is reassuring: "our rival is smaller; it cannot catch up."
The actual logic is the inverse. Shareholders were already uneasy. The existence of the memo did not create the uneasiness; it was a response to it. The technical argument used by OpenAI—that Anthropic operates on a lower scaling curve—has a structural problem: it is exactly what any dominant company has said about its challengers just before losing market share. IBM said it about Microsoft. Microsoft said it about Google. The scale argument holds in the lab but loses in the business market to the argument of trust and fit.
Companies integrating Anthropic's Claude into their platforms do so not because it is cheaper or larger; they do it because, in segments where governance in AI usage matters, Anthropic's track record builds a trust narrative that OpenAI is still trying to articulate. Competitive advantage in the business segment is not measured in petaflops; it is measured in the perceived alignment of incentives between provider and customer.
OpenAI has an internal coherence problem exposed clearly by this episode. It accepted a contract it later admitted seemed careless. This is not a communications accident; it signals that contracting decisions are misaligned with the company’s public messaging about safety. When OpenAI's own team had to renegotiate terms to explicitly exclude mass domestic surveillance, it was, in effect, moving closer to the conditions that Anthropic demanded from the outset.
An Industry That Cannot Pretend Scale Solves Everything
This dispute reveals, beyond tensions between two specific companies, that the enterprise AI market is entering a maturation phase where the argument for scale is no longer sufficient to win large, sensitive contracts. Companies with regulatory exposure, governments needing to justify their technology procurement decisions to internal and external audiences, and technology teams accountable to increasingly vigilant boards concerning reputational risks are beginning to evaluate their AI providers based on criteria beyond model performance.
In this context, the position Anthropic has built, at the cost of foregoing a $200 million contract and enduring a PR crisis from its CEO's memo, holds strategic value that its $60 billion valuation has only begun to reflect. Not because Anthropic is a perfect company, but because its sacrifices align with its value proposition, and that coherence is exactly what enterprise clients are willing to pay for.
OpenAI, with all its scale advantage, faces a more challenging task than its memo suggests: building the institutional trust that allows it to win in the most demanding market segment. Scale amplifies model capabilities; it does not amplify the credibility of its leaders' decisions.
The lesson for any C-Level executive observing this dispute from the outside is direct and unembellished: the most enduring strategic positioning is not built by choosing what to offer, but by consistently upholding what has been decided not to do, even when the cost of that renunciation shows up in the quarterly earnings statement. Companies that try to be available to all clients, under all conditions, for all possible uses end up not conquering the total market. They become irrelevant in the segment that matters most.
