AI Revenue Hits 25 Billion — But the Margins Tell a Different Story
How this was made Verified AI
Every Intellegix briefing is generated from that day's broadcast and run through automated checks before it publishes — with a human paged on any flag. Here is the trail for this edition.
Quarterly AI revenue sector-wide nearly doubled to 25 billion dollars, surpassing depreciation charges for the second consecutive quarter — the milestone financial analysts had been waiting for. Yet margins remain razor-thin, explaining why big tech stocks are sliding even as top-line numbers improve. The economics of AI currently resemble a capital-intensive infrastructure build where revenue is just barely covering costs while companies bet on future margin expansion.
OpenAI published research showing its Codex agentic coding platform now handles 99.8 percent of the company's internal AI work, with non-developer adoption growing 189 times since August 2025 — a figure suggesting category expansion rather than simple user growth. DeepSeek, the Chinese AI startup that made waves with its lean, high-efficiency model, abandoned that approach entirely after closing its first outside funding round at 7.4 billion dollars, announcing plans to double its 170-person team. The company that demonstrated frontier AI could be built with minimal resources is now spending like everyone else.
Anthropic accused Alibaba of stealing AI capabilities — an allegation landing at the exact moment White House talks with Chinese counterparts are intensifying, creating diplomatic awkwardness. On the regulatory front, a House bill would require AI firms to report dangerous incidents to the Department of Commerce, establishing a mandatory incident-reporting framework analogous to those in aviation and nuclear industries. Tech giants also launched a 500-million-dollar nonprofit dedicated to retraining workers for the AI economy.
Ford's experience offered a cautionary data point. The automaker had bet that AI-based quality inspection could replace experienced engineers in manufacturing, discovered it could not, and rehired 350 veteran engineers. The episode illustrates where AI tools currently succeed — and where tacit human expertise still outperforms them. The FDA, separately, granted breakthrough device status to two generative AI radiology tools, accelerating their path toward clinical deployment. Anthropic's Claude chatbot saw its paying user base grow 75 percent since January, according to credit card transaction data, intensifying the competitive dynamics between OpenAI, Anthropic, Google, and DeepSeek even as the overall market expands.