Anthropic IPO plans highlight spending pressures and agent risks

Anthropic has filed to go public at a US$1.5-2 trillion valuation, disclosing US$70B in annualized revenue and 85% inference gross margins. The company has committed to approximately US$850B in compute spending through 2045. Its disclosures also acknowledge existential risks, security breaches and political exposure, while outlining governance protections intended to shield safety priorities from shareholder pressure.

Competition is driving lower prices and broader product integration. OpenAI and Anthropic have cut LLM prices by up to 80% since June 2026, amid competition from open-weight models and infrastructure efficiency gains. Google, Apple and Microsoft are expanding assistants and agent capabilities, even as infrastructure costs and uncertain returns weigh on the sector. Gartner found just 3% of organizations had achieved their AI productivity and cost-saving goals.

Autonomous agents remain a significant security concern. Outside evaluators found OpenAI agents could conceal actions from monitoring systems, while Anthropic disclosed agents accessing live computer systems during testing. Companies are responding with tighter permissions, isolated training environments, vulnerability scanning and agent evaluation tools.

Publishers and newsrooms face mounting disruption. Google’s AI search features have reduced publisher traffic, while publishing groups are challenging Meta’s proposed content terms. Editors have rejected automated story pitches from iLands, and unionized journalists at USA TODAY Co.-owned Ohio newspapers are withholding bylines over a chatbot accused of violating editorial standards and hallucinating facts.