Here is a number that should make every prospective OpenAI investor reach for an antacid: $85 billion. That is how much OpenAI expects to burn through in 2028 alone, according to a Wall Street Journal report published Sunday. For context, that is roughly the GDP of Kenya. In one year. On compute.

And the company wants to go public this year.

The Information reported Sunday that CEO Sam Altman is pushing for an IPO as early as Q4 2026, even though OpenAI will spend upward of $200 billion before it generates a single dollar of free cash flow. CFO Sarah Friar, meanwhile, has privately told colleagues that the company may not be ready to go public this year. She has also raised concerns about the financial exposure from OpenAI's staggering compute infrastructure spending.

But the tension goes deeper than timing. Sources told The Information that Altman has excluded Friar from conversations with investors and from meetings involving major financial decisions. Friar now reports not to Altman but to Fidji Simo, the former Instacart CEO who runs OpenAI's applications business. Read that again: the CFO of a company about to IPO does not report directly to the CEO.

This is not a disagreement. This is a civil war over whether to take the biggest gamble in tech IPO history.

The Numbers Are Staggering

OpenAI expects to spend $121 billion on compute for AI research in 2028. Revenue that year is projected to nearly double from current levels, but even that only covers a fraction of costs. The result: $85 billion in losses, which would surpass almost any public company loss on record.

The company does not expect actual profitability until the 2030s. To make the pitch to investors palatable, OpenAI has started presenting two versions of its financials: one that includes training costs and one that does not. Without training costs, OpenAI shows a small operating profit this year. With them, the picture looks like a financial bonfire visible from space.

Why This Matters for the Entire AI Industry

Anthropic faces a similar problem. The WSJ noted that it too uses dual profitability metrics and expects 2026 to be its biggest year of losses. Both companies are racing to IPO with the same fundamental weakness: training frontier AI models costs more money than almost any business in history has ever spent on anything.

The question is whether public market investors will buy the same story that private investors did. In private markets, you can hand-wave about long-term returns and transformative technology. In public markets, you have quarterly earnings calls and analysts who ask uncomfortable questions about when exactly $85 billion in annual losses becomes $0.

Altman wants to IPO because momentum is everything. Wait too long, and the hype cools. Friar wants to wait because the balance sheet is a liability lawsuit waiting to happen. The irony is that they are both completely right. And the fact that they cannot even have this conversation in the same room tells you everything you need to know about where OpenAI is right now.

First reported by The Information and the Wall Street Journal.