Anthropic IPO Roadshow Focuses on Compute Investment, Also Emphasizes AI Safety

Anthropic released an approximately 42-minute initial public offering roadshow video on November 10. Chief Executive Officer Dario Amodei mentioned shortly after the opening the risk that AI could be used to create biological weapons, while also calling AI a highly promising technology and emphasizing that the company values safety. Business Insider separately used ChatGPT and Claude to analyze the video transcript and found that besides "Claude" and "model," "compute" and related terms appeared 49 times, "customer" appeared 42 times, and "safety" and related terms appeared 27 times. According to the themes summarized by the media, model and training and compute and infrastructure ranked at the top, while themes such as safety, alignment, trust, and risk appeared a combined 55 times. The video also includes introductions from customers such as Bristol Myers Squibb and Accenture, attempting to demonstrate Claude's commercial uses. Anthropic said in the video that safety will not hinder business growth, and that trustworthy models instead help apply research results to real-world scenarios. As the company seeks financing through an initial public offering, its massive compute commitments are also a key focus for investors: its latest prospectus lists total spending committed under purchase and lease agreements at about US$900.9 billion, most of which is related to compute and infrastructure, with about US$451.6 billion expected to be incurred by the end of 2030.