The AI Bubble and the Infrastructure That Stays 0 ▲ zoia.org 11 minutes ago · Politics · hide Anthropic S-1 filing (and subsequent S-1/A) has finally revealed part of the company’s cost structure. The company reported $16.3 billion in revenue in the first half of 2026, with an implied gross margin of roughly 85% after cost of revenue. It would look like an extraordinary business if R&D expenses—primarily training and research compute—hadn’t reached $9.7 billion in the same period, and if the company hadn’t disclosed roughly $900 billion in committed compute and infrastructure spend. Are we in a bubble? I think we are. It’s not only about how much of the AI boom is just nVIDIA’s and other providers’ cash being recycled. Frontier labs need to discover a way to train new models at a significantly lower price. The IPO is a liquidity event that will allow investors to cash out, but it won’t solve the underlying business model problems. The Railroad Bubble After the US Civil War, US railroads mileage roughly doubled, from about 35,000 miles in 1865 to more than 70,000 by 1872. High… Commenting on Bubbles has been removed. Why? If you want to talk about this post, write to the author, use the comments on their blog if it has them, or reply with a "RE:" post on your own blog. Suggestions? feedback@bubbles.town