Two labs. One asked for $30 billion. The other showed the $518 billion bill.

I was reading the news between study sessions last night and I had to do a double take on the numbers. Not because they were wrong. Because they were that big, in the same week, from the two companies whose products I use every day.

OpenAI is out asking for at least $30 billion more, privately, at a valuation around $1.4 trillion. That's up 64% from March, when they raised $122 billion at $852 billion and everyone assumed that was the last round before the IPO. Turns out the IPO isn't happening in 2026. Sam Altman told Fortune it's "an ill-advised moment to go public" given AI safety concerns, so the company needs a bridge round to a 2027 listing instead. Meanwhile the money coming in is real: Bloomberg reports annualized revenue hit $40 billion in August, up 70% since July, and coding products are a big part of the story. TechCrunch

Same week, Anthropic went the opposite direction. Reuters reviewed their IPO prospectus and the numbers are wild: targeting a valuation above $2 trillion, more than double the $965 billion private valuation from May. 2025 revenue grew 12-fold to nearly $4.6 billion. And a $42 billion net loss, though about $34 billion of that is a non-cash accounting charge tied to convertible financing. The real number is the operating loss: just over $8 billion. PYMNTS

But the line that got me wasn't the loss. It was this: Anthropic has signed $518 billion in multiyear infrastructure commitments. About 80% of that is non-cancelable or payable no matter what. $111.1 billion with Google. $110 billion with Amazon. $31.4 billion with Microsoft. $161.2 billion in Broadcom equipment leases. These are contracts that say "we will spend this money whether we need it or not," stretching seven to ten years out.

That's what a public filing buys you. Truth. When you stay private, investors get the revenue curve and a promise. When you file, everyone gets the receipts. I wrote last week about how the AI boom is being financed on credit, and this is the document version of that same story. Except now the debt has a name and a term sheet attached.

The part that hit me

Two companies. Same product family. Opposite answer to the same question.

OpenAI looked at 2026 and said no. The stated reason is safety, that listing now would be "ill-advised." Maybe that's true. It's also convenient, because going public means showing your hand, and OpenAI's hand reportedly includes an internal projection of $278 billion in cumulative negative free cash flow through 2030. A private round at $1.4 trillion lets them raise the money without a public auditor or a public argument about that burn.

Anthropic chose the hard door. Public markets, public numbers, public scrutiny. Nearly a quarter of their revenue comes from just two customers who could walk. The investor roadshow reportedly starts the week of November 9, with trading before Thanksgiving. Stock Moguls

I don't know which path is smarter. I'm a WGU student, not a finance guy. But I know which one I trust more. The one that shows the bill.

Why I keep coming back to the boring part

Here's what keeps nagging me. Bloomberg credited OpenAI's revenue jump to a strategic refocus on coding. Coding assistants, agent frameworks, metered API calls. That's the product paying for this whole circus. That's the same tool I open every night when I push code.

So the picture looks like this: developers pay for coding products, that revenue justifies valuations in the trillions, and the money gets turned into decade-long compute contracts that are mostly non-cancelable. The intelligence layer is priced like it's the business. The infrastructure contracts are priced like they're the business. Somewhere between those two beliefs is a gap, and it's measured in hundreds of billions of dollars.

I'm applying to IT roles right now. I can't ignore the gap. It tells me the unglamorous work, the compute, the networking, the people who keep the racks alive, isn't going anywhere. You don't sign $518 billion in non-cancelable contracts and then not need the humans to run the thing.

The numbers are absurd. But for once, they're published. And published numbers are something a student can actually learn from.