OpenAI Tried Alphabet's IPO Trick. It Didn't Work.

Today:

Altman tried a quiet trick to clean up OpenAI's books before the IPO.

The lawyers caught it.

Here's what it tells you…

Late last year, Sam Altman tried to spin OpenAI's robotics and consumer hardware divisions off the main company.

The plan was Alphabet-style — loss-makers ring-fenced from ChatGPT before the IPO. Accounting rules killed it in the room.

Watch the S-1 when it lands: the line items he wanted hidden will be sitting right there in the filing, waiting for the analysts.

The trick failed because the lawyers can read.

The Wall Street Journal broke it on Monday.

Altman wanted the two units carved out so they could raise their own money and run independently. Both report directly to him. Insiders describe them as startups inside a startup.

The pitch: let ChatGPT shine on its own, park the experiments somewhere else.

Then accounting rules got in the way.

Even spun off, the new entities might still have to be consolidated on OpenAI's books — meaning the trick wouldn't have actually worked. So it was dropped. The point isn't that the plan failed.

It's that the plan existed.

OpenAI made $13.1bn in 2025 and is projecting $14bn in losses for 2026.

The hardware bets aren't small. A $6.5bn all-stock deal for Jony Ive's io Products in May. A 10-gigawatt custom chip partnership with Broadcom in October.

Neither will print money before the listing (and the chip deal won't print until 2029).

OpenAI is targeting a Q4 2026 IPO at up to $1tn, against a current private valuation of $852bn.

That's a 65x sales multiple. Only Palantir trades richer in the S&P 500.

Altman wanted a clean prospectus to sell the number.

He didn't get one.

So the prospectus will say it loud and clear: ChatGPT's revenue tangled up with billions in moonshot losses.

Same page.

Same column.

Pro: Hardware and robotics compound with the model. Keeping them in-house is how Altman stays in the AI race, not how he hides anything.

Con: Pre-IPO CEOs don't try to spin off their loss-making divisions when the loss-making divisions are doing fine. They do it when the losses are about to make the cover story harder to tell.

Our read: The failure of the trick is less interesting than the trick's existence. Read accordingly.

[Report] OpenAI — "Accelerating the next phase of AI"

The company confirmed a $122bn raise at an $852bn valuation, with enterprise revenue on track to match consumer by the end of 2026.

Why it matters: The pre-IPO ramp is real. So is the pressure to look profitable on paper before the listing.

$14bn loss on $13bn revenue.

[WSJ via MSN] — OpenAI explored robotics spinoff before IPO — The original scoop. The accounting detail is the story.

[CNBC] — Cerebras targets $26.6bn IPO — OpenAI's chip supplier is going first. Quieter listing, cleaner numbers.

[Motley Fool] — Why big-bang IPOs underperform — A reminder that "biggest ever" rarely means "best buy on day one."

What does the failed spinoff really tell you?

You can read all our back issue newsletters for free here.

For the love of coffee, see you tomorrow!

Enjoy your Wednesday, keep it caffeinated.

That's your Wednesday brew.

The lawyers can read. So can you.

Trust your gut.

Read yesterday’s newsletter about Gamestop’s eBay takeover here.

Thanks for reading!

Are you subscribing?

Join your crew of caffeinated sceptics today.

Be sure to get your daily Curse and Coffee fix by hitting the button below.

Open Monday to Friday.