Investors Want OpenAI and Anthropic IPOs to Deliver Transparent Numbers on AI Economics

Investors are increasingly demanding proof that multibillion-dollar AI spending delivers returns. David Fetherstonhaugh, portfolio manager at VistaShares and principal at DVX Ventures, made the point on Bloomberg Tech on October 12, 2026, in a conversation with Ed Ludlow. His central argument: the industry's most influential players, OpenAI and Anthropic, remain private companies, which is why a key part of AI economics is closed off from the market. Their future IPOs could change that.

No specific figures were given on air, and no timeline for the listings was named. This is less news about a filing than a thesis about what the market is missing.

Where AI Economics Is Dark

The AI value chain is long. At one end are chipmakers, cloud providers and data center builders, which report publicly. Their revenue and capital expenditures are visible quarter to quarter. At the other end are applications and services that are supposed to turn compute into money. In the middle sit the frontier model developers, and this is where data is scarcest.

That creates a skew. Infrastructure spending is observable, while the revenue that is supposed to justify it is known only through leaks, interviews and rare disclosures. As long as OpenAI and Anthropic do not publish audited financial statements, an investor cannot check whether demand matches the scale of the capacity that has been built and ordered.

What an IPO Would Force Them to Disclose

A public company must report its numbers to a regulator's standards, and that changes the quality of the information. For labs that spend enormous sums on training and inference, the set of metrics would be revealing:

• revenue broken down by product: subscriptions, API, enterprise contracts;
• gross margin that accounts for the cost of inference;
• commitments to buy compute from cloud providers and chipmakers;
• the cost of training new models and its share of operating expenses;
• customer concentration and dependence on a few large contracts;
• cash flow trends and the need for new capital.

Each of these lines is estimated by eye today. After a listing, it could be compared year over year and against competitors.

The Effect on the Whole Chain

Transparency from the labs matters to more than their own shareholders. Reporting from OpenAI or Anthropic would become a benchmark for suppliers of chips, cloud and energy, whose investments depend on paying demand. If inference margins turn out to be high and revenue grows faster than costs, that strengthens the case for current capex. If the opposite is true, the repricing would hit not only the labs themselves but everyone building capacity for them.

This, by Fetherstonhaugh's logic, is what the market is waiting for: not more statements about scale, but verifiable data.

What to Watch in Practice

Until the listings happen, you have to rely on indirect signals.

• Reports from cloud providers and chipmakers. Growing backlogs and comments about lab customers show how demand is growing.
• API pricing. Falling token prices alongside rising load hint at improving unit economics, while holding prices steady amid a compute shortage suggests the opposite.
• Enterprise contracts. Companies moving from pilots to large-scale deployments is a more reliable signal than the number of chatbot users.
• Terms of new funding rounds. They reveal how investors value returns, even before a stock market debut.

For teams choosing a model provider, the takeaway is down to earth: factor in the risk of changes to prices and terms until the labs' economics have been verified by public reporting. For investors, the main question stays the same: when will revenue start catching up with spending, and how much longer will that stay hidden behind private status?