Following the analysis from our latest podcast, where we broke down Anthropic’s pivot away from the public sector and OpenAI’s aggressive move to fill the vacuum, we’re bringing you our characteristic “bad maths” on who actually made the smarter play.
The logic here is a classic strategic trade-off: one path leads to a government-backed monopoly, while the other bets on the shifting loyalties of the consumer market. With high-profile figures like Katy Perry publicly making the jump to Claude, and hordes of non-space-adventurers in hot pursuit, this is starting to look expensive for OpenAI.
Here is the breakdown of the numbers:
The “War” Chest
Let’s start with the easiest variable: the Pentagon. The US Department of Defense recently unveiled its 2026 fiscal budget, earmarking $66 billion for IT. Within that, $305 million is slated for software licenses and $507 million specifically for AI development.
Even with a generous estimate, a general-purpose AI tool might capture 20% of those combined pools—roughly $160 million. This sits just under the $200 million ceiling Anthropic disclosed back in 2025. On paper, that is now OpenAI’s to lose. But in a market where we talk in trillions, $200 million is a rounding error. The real question is what OpenAI sacrificed to get it.
The Cost of Selling Out
OpenAI has been bleeding sentiment for months. The decision to integrate ads earlier this year was a turning point, a signal of defeat that Altman derisively called a last straw for his company just a few years ago. Simultaneously, Anthropic’s refusal to play ball with the Pentagon has sent Claude to the top of the US App Store.
To see why this matters, we have to look at the “bad maths” of subscriptions:
- OpenAI Revenue: Estimated at $13 billion.
- Subscriber Revenue: Roughly $9 billion of that comes from individual paid users.
- The potential loss: If just 2% of those subscribers find OpenAI’s new military ties or ad-bloat distasteful and switch to Claude, OpenAI loses $180 million in recurring annual revenue.
In this light, Anthropic hasn’t just taken the moral high ground; they’ve taken the more lucrative one. Our internal tech tracker suggests a 2% defection rate is highly conservative. The drift is likely much larger.
Too Big to Fail
If the math favors Anthropic, why is Sam Altman smiling?
Most of the attention is on the pentagon, but in practice, Anthropic is now persona non grata for a lot of other government departments that will also be investing in AI. That makes the potential government contract value much higher than detailed here – but we’d still argue it will struggle to keep up with consumer spending.
But of more consequence is a deeper play by OpenAI. By embedding its models into the bedrock of national security and critical infrastructure, it’s attempting to become a systemically important institution. This is a calculated survival move. With many industry pundits predicting OpenAI could face a liquidity crisis or bankruptcy by 2027 due to astronomical compute costs, becoming the Pentagon’s primary brain makes them very difficult to liquidate.
Anthropic is winning the hearts of the market, but OpenAI is trying to become the state. Anthropic has the better balance sheet today; OpenAI is betting that by 2027, the US government won’t have a choice but to keep their lights on.
