OpenAI's Identity Crisis, Datacenter Wars, Market Up on Iran News, Mamdani's First Tax, Swalwell Out

OpenAI and Anthropic both hit roughly $30 billion in annual run rate at the start of Q2 2025. But their growth trajectories have diverged sharply: Anthropic is growing 10x year-over-year, while OpenAI is growing 3–4x. Anthropic went from $1 billion to $10 billion in ARR in 2024, reached $30 billion by Q1 2025, and is on pace for $80–100 billion by year-end. The secret is enterprise coding. Businesses pay for code tokens on a metered, usage-based model—like electricity—and their consumption scales exponentially as teams adopt AI-native workflows. Consumer subscriptions, by contrast, are capped at $20/month all-you-can-eat, with only 3–4% conversion from free to premium.

OpenAI's leaked internal memo acknowledged the threat, calling out Anthropic's «inflated» revenue figures (which include channel partner revenue-share arrangements), and pivoting the company's focus toward enterprise and the «agent platform layer.» The memo also attacked Anthropic's positioning as built on «fear, restriction, and the idea that a small group of elites should control AI»—a direct shot at Dario Amodei's public doomer messaging. But the pivot may come too late: if Anthropic achieves a 10x in one year and OpenAI takes two years to 10x, the network effects around compute, token volume, and reinforcement learning could make Anthropic's lead insurmountable within 12–24 months.

The S&P hit fresh all-time highs this week—seven weeks into a U.S. military operation in Iran that has cost over $100 billion. The market's confidence rests on two pillars: first, Trump's statements and the Islamabad meeting suggest the conflict is nearing resolution, and the stock market is pricing in a quick exit. Second, traders are betting on an AI-driven earnings surge at large companies as they deploy language models to make top employees 10–20x more productive. Historical efficiency gains from the internet or Microsoft Office were incremental—maybe 30–50%—but AI promises an order-of-magnitude leap.

Yet the paradox is real. The Shiller PE and Buffett Indicator (market cap / GDP) are both near all-time highs, signaling overvaluation. But market dispersion is extreme: only eight or nine companies are hitting all-time highs, while the rest lag. This creates conflicting signals—some data scream «risk-off,» while others (like the «up 5% in the first half of April» historical pattern, which predicts +32% for the rest of the year) scream «risk-on.» Investors can cherry-pick data to justify any bias. The truth is more nuanced: if AI delivers on its productivity promise, valuations are justified. If enterprises can't prove ROI at scale—and many transformation projects are failing—then we're in bubble territory.