OpenAI and Anthropic IPOs Reshape Big Tech's AI Trade

OpenAI and Anthropic have filed confidential preparations for initial public offerings. SpaceX plans to list this month. The three companies, each valued in the hundreds of billions, are set to bring direct ownership of frontier AI and infrastructure to public markets.

For years, public-market AI exposure meant buying a few hyperscalers. Microsoft invested over $13 billion in OpenAI and embedded its models across Azure and Office. Alphabet and Amazon committed billions to Anthropic. Nvidia’s data center segment became tied to AI training and inference demand. Investors owned the infrastructure layer, not the model builders themselves.

Anthropic was valued at $965 billion in private markets, CNBC reported. The company operates at an annualized revenue run rate approaching $50 billion, though compute and training costs remain opaque. OpenAI generates billions in revenue and reaches hundreds of millions of weekly users, Reuters reported. SpaceX, while outside pure AI software, has been valued between $1.5 trillion and $2 trillion in secondary transactions. CEO Elon Musk told The Globe and Mail that AI represents SpaceX’s largest growth opportunity.

The IPOs would end the indirect-exposure structure. Investors could buy equity in model developers directly. That changes the risk profile. Public markets require financial disclosure. Private funding rounds obscured how much of OpenAI’s revenue goes to compute costs and how dependent it is on Microsoft’s cloud.

Benzinga described the circular nature of the AI ecosystem. Hyperscalers fund model developers. Those developers drive compute demand, which feeds back into hyperscaler revenue. In private markets, that loop supported valuations across the stack. Public markets subject it to quarterly scrutiny. A weak earnings report at Anthropic could ripple back to Amazon’s cloud revenue expectations.

Current holders of MSFT, AMZN, GOOGL, and NVDA face a potential revaluation. The premium these stocks carry for AI exposure could compress if direct alternatives emerge. MSFT’s Alpha Score sits at 58, reflecting uncertainty about the shift. NVDA at 66 still benefits from compute demand.

The biggest risk is that the IPOs reveal the model builders’ economics are worse than expected. If margins are thin after compute costs, valuations may correct. That would affect not just the new stocks but also the hyperscalers that hold stakes.

SpaceX’s IPO is expected this month. OpenAI and Anthropic have only filed confidential paperwork; a public offering could come in 2025 or 2026. The S-1 documents, once released, will provide the first clear look at revenue and cost structures. The percentage of OpenAI’s revenue that flows to Microsoft for compute will be one of the most closely watched numbers.