AI Bubble Debate: Nvidia Reduces Risk, Anthropic Performs Strongly

Nvidia and OpenAI are close to a deal to finance a huge data center in Ohio, but the conditions have been significantly adjusted. Instead of the originally planned $250 billion guarantee, Nvidia would initially guarantee only about $120 billion, according to the Wall Street Journal, due to investor concerns about Nvidia's risk exposure. The guarantee covers the first construction phase with about five gigawatts of capacity. OpenAI is simultaneously negotiating a lease for the entire 10-gigawatt project, developed by SB Energy, a SoftBank subsidiary. Additionally, Nvidia is discussing separate financing for OpenAI's chip purchases of up to $350 billion. The reduced deal may embolden critics warning of an AI bubble, as Nvidia's risk halving under investor pressure could signal caution even among the biggest winners. At the same time, Anthropic provides a strong counterargument: its revenue more than doubled in one quarter, from $4.73 billion in Q1 to over $11.5 billion in Q2, a 14x year-over-year growth, reported by Reuters. For 2028, Anthropic forecasts revenue of around $190–200 billion, far exceeding the annual run rate of about $45 billion communicated in May. Anthropic claims its revenue has more than decupled each year for three years. If these figures hold, it would signal that demand for proprietary AI services continues to grow massively despite political headwinds and Chinese competition. Anthropic is reportedly planning an IPO at a valuation of nearly $1 trillion in late September or early October. Financial services firm Ramp recently measured a slight flattening in demand for Anthropic tokens.