Anthropic Just Filed to Go Public. The Numbers Are a Warning About Your AI Budget.

Anthropic filed paperwork this week to go public, and the numbers in that filing are a more useful planning tool than anything in its product announcements: $16.3 billion in first-half 2026 revenue, an $80.7 billion net loss, and more than $518 billion in compute commitments already on the books. If your team runs on one AI vendor with no fallback priced out and no number that would make you look for one, this is the week to build both.

The filing nobody outside finance read closely

Anthropic's draft S-1, filed Oct 24, showed revenue up roughly 15x year-over-year alongside that $80.7 billion loss — a gap that's growing in both directions at once. Two days earlier, CNBC reported OpenAI is separately raising at a $1.25 trillion valuation. Both companies are proving enterprise demand is real while burning cash faster than that demand is paying for, on a bet that scale eventually closes the gap. AI subscription pricing is still subsidised — this filing is the first time outsiders have seen exactly how large that subsidy is, in one company's own numbers.

What to do differently Monday morning

None of this changes what Claude or GPT does for your team today. It changes what you should assume about price a year from now. A company committing over half a trillion dollars to future compute while losing $80 billion in six months will eventually need a bigger raise, a price increase, or a change to free-tier terms — and no team wants to learn which one by having a workflow break overnight. The fix isn't switching vendors preemptively. It's knowing, before the next invoice, what a price increase would cost you and what you'd do about it.

Priya: the budget line that didn't assume anything

Priya runs operations for a 90-person logistics company where dispatch summaries and customer email drafts run through Claude roughly forty times a day. When she set the workflow up last year, she also priced out Mistral Large at a comparable tier — not because she planned to switch, but because she wanted a real number to compare against if Anthropic's pricing ever moved. When the IPO filing made the rounds this month, the question she asked wasn't whether to worry. It was whether this month's bill was still inside the range she'd already planned for. It was, so she didn't have to do anything — the budget line had assumed volatility going in, and the filing just confirmed that was the right assumption.

Dennis: the one-tool shop with nothing to compare against

Dennis runs a four-person marketing consultancy and does everything through it — strategy decks, ad copy, research summaries — on a single AI subscription billed to his personal card. He's never priced a second tool, never checked what the subscription costs at list price instead of its current promotional rate, and has no plan for this week if that price doubled. The filing doesn't create his risk; it just makes it visible. A four-person shop with one vendor and no ceiling is more exposed to a pricing shift than Priya's 90-person company with a comparison already costed out.

The one thing

Anthropic's filing isn't a reason to drop Claude, and OpenAI's valuation isn't a reason to switch to it. It's a reason to know, for every AI tool your team depends on, what you'd do the week the price changes — because at $80.7 billion in losses in six months, somebody eventually has to ask for more.