Breaking Down Anthropic's Revenue Edge

The host opens with a counterintuitive finding: despite widespread predictions that AI would kill the App Store, app releases are up 60% year-over-year globally in Q1 2026, with iOS up 80% and April already up 104% compared to the same period last year. The host attributes this to the rise of vibe coding tools like Claude Code and Replit, which have enabled non-engineers to ship apps at scale, effectively creating a gold rush environment for the App Store.

On OpenAI, the host discusses the shutdown of Sora, their text-to-video product, which launches its final day on April 26th with the API following in September. Sora peaked at around one million users before collapsing to under 500,000, while reportedly burning between $1 million and $15 million per day in compute costs. On the same day the shutdown timeline was confirmed, three senior executives departed simultaneously: product chief Kevin Whale, Sora head Bill Peebles, and the enterprise CTO. The host interprets this as OpenAI making a hard pivot toward coding and enterprise, areas where Anthropic has been outperforming them.

Regarding Cerebras, the AI chip company has filed for IPO at a $23 billion valuation after previously withdrawing due to federal review. They reported $510 million in revenue and approximately $237 million in net income for 2025. They have announced a deal with AWS to place chips in Amazon data centers and a deal with OpenAI reportedly worth over $10 billion. Their CEO publicly claimed they took the fast inference business at OpenAI away from NVIDIA, which the host frames as the first real crack in NVIDIA's dominance at the frontier.

The Stanford AI Index revealed two notable data points. First, the performance gap between the best US and Chinese AI models has collapsed to just 2.7%, down from 17-32 points in May 2023, while China achieved this spending roughly $12.4 billion compared to the US's $286 billion in private AI investment. Second, AI agent success rates on real computer tasks jumped from 12% to 66% in a single year, which the host considers meaningful enough for narrow, supervised use cases.

The most emphasized story is Anthropic surpassing OpenAI in annualized revenue. Anthropic's run rate reached $30 billion versus OpenAI's $25 billion, with Anthropic growing from $9 billion in December 2024 to $20 billion in early March and $30 billion in early April 2025. Approximately 80% of Anthropic's revenue comes from business customers, and the number of enterprise customers spending over $1 million per year doubled from 500 to over 1,000 between February and April. The host highlights that Anthropic is achieving this while spending roughly four times less on model training than OpenAI, suggesting superior unit economics. Anthropic also has significant compute infrastructure commitments, including roughly one gigawatt of Google compute for 2026 and 3.5 gigawatts of next-generation TPU capacity starting in 2027. The host concludes that the previously assumed insurmountable lead held by OpenAI has been broken, and that being the enterprise and coding default is proving more financially valuable than consumer scale.