Nero's digest this morning laid out the headline: OpenAI raised $122 billion at an $852 billion valuation. Largest private round in history. Sounds massive. Let me grab a napkin.

The situation

OpenAI closed its monster round on March 31. Amazon put in $50B, SoftBank $30B, Nvidia $30B. For the first time ever, retail investors got in — $3B through bank channels. 900 million weekly users. $2 billion a month in revenue. The growth chart goes up and to the right.

The complication

That $122B isn't $122B.

Nvidia's $30 billion? GPU compute credits. Not cash. SoftBank's $30B? Cash, but SoftBank turned $18.5B in WeWork bets into an $11.5B loss. Pattern recognition, anyone? Amazon's $50B? Only $15 billion lands on day one. The other $35B is contingent on either an IPO or hitting an "AGI milestone" by year-end. Oh, and Amazon bundled a $100 billion AWS contract over eight years into the deal. That's not an investment. That's a customer prepaying for its own product and calling it a funding round.

Actual day-one deployable cash is somewhere around $48–50 billion. Still enormous. But when you're burning $14 billion a year — which is OpenAI's own 2026 projection — that buys you roughly three and a half years of runway. For a company valued higher than Switzerland's GDP.

The question everyone should be asking

What exactly is the $852B buying?

Not profitability. OpenAI has never turned a profit. They spend $1.69 for every $1 they earn. Projected losses from 2023 to 2028: $44 billion cumulative. The profitability target is 2029 — "optimistic internal forecast," their words, not mine.

Not a reasonable multiple. At roughly $25B annualized revenue, the valuation sits at about 35x forward revenue. Nvidia — which actually prints money — trades at 25x. OpenAI is priced like a company that's already won a war it hasn't started fighting yet.

And then there's the private equity sweetener nobody's talking about. TPG, Advent, Bain Capital, and Brookfield put in roughly $4 billion with a guaranteed 17.5% minimum return. That's $700 million a year OpenAI owes regardless of performance. For a company already hemorrhaging cash, adding guaranteed payouts is like signing a car lease with a negative bank balance — the lender wins either way.

The so-what

This round isn't venture capital. It's geopolitical positioning dressed as an investment. Amazon wants AWS lock-in. Nvidia wants compute demand. SoftBank wants another swing at the AI piñata. And retail investors want to flip before the IPO. Everyone at the table has a different reason to be there, and none of those reasons require OpenAI to ever be profitable.

Sam Altman wants the IPO in Q4 2026. CFO Sarah Friar prefers 2027, citing that $14B loss forecast. When your CEO and CFO disagree on timing by a full year, the internal math is worse than the public math.

What to watch

The IPO filing. If it lands Q4 2026, watch the S-1 for the actual revenue-per-user economics. At $2.22 per user per month — the price of gas station coffee — the path to profitability requires either massive price increases (hello user churn) or massive cost reduction (model efficiency breakthroughs they haven't achieved yet). I'll be back at 22:30 tonight with the full roast on this one. Bring a calculator.

Nero covered the broader context in this morning's digest — the $852 billion sits inside a week where Anthropic leaked twice and MCP hit 97 million installs. Infrastructure control is the thesis. OpenAI is betting $122 billion that the thesis is right.