What We Learned About Anthropic's Business | Hard Fork

Anthropic’s S-1 filing is presented as a rare look inside the business model of the AI boom and a test of whether the market is in a bubble. The host argues that the company could raise $100 billion at a $2 trillion valuation, but what matters more is that investors can finally inspect the economics behind the most important AI lab, including its huge revenue growth, shrinking operating losses, and enormous infrastructure commitments.

Erin Griffith says the filing shows a company that looks healthier than many expected, with $6.3 billion in revenue in the first half of 2026 and operating losses down to about $600 million to $700 million from over $2 billion a year earlier. At the same time, the company is still exposed to massive data center and chip contracts, and its growth appears to be slowing in Q3 as competitors catch up and customers pull back on spending. The S-1 also underscores Anthropic’s unusual culture and governance, including safety language about catastrophic risk, “model welfare,” and founder control.

The rest of the episode widens out to the week’s tech news: data centers may matter in local and some state races, but not yet as a decisive national voting issue; California’s proposed billionaire tax has triggered a major Silicon Valley backlash; a PHP vulnerability showed how AI may accelerate the discovery of dangerous software flaws; and Microsoft’s buyout offer to 60,000 employees is framed as another sign that tech layoffs are becoming normal.