The Finish Line Is a Rival's Paperwork

OpenAI accelerated its $30 billion private round this week for a reason it put on the record: to lock in investors before Anthropic's IPO disclosures allow anyone to set the two companies side by side ^[1]. In this race, the finish line is a rival's paperwork, not a rival's model. The calendar has been setting the price of intelligence since August. On August 10, OpenAI cut fees on its Luna model by 80% and Anthropic introduced a high-performance model at half the price of its top system, and the record tied the move to labs chasing profitability ahead of potential public offerings ^[2]. Six weeks later, on September 22, the two synchronized another round of cuts that the reporting said reflected pressure to generate returns on investment, with Anthropic's roughly $2 trillion listing named as the forcing function ^[3]. OpenAI has stated its own economics plainly ^[4].

Intelligence has to keep getting cheaper. — OpenAI

Morgan Stanley has already drawn the diagram: cheap model access may function as a loss leader — the cheap item that pulls customers toward the profitable ones — with the margin living in cloud, storage, and managed infrastructure ^[5]. In that same window, the growth moved to the paper itself. Anthropic's filing asks the public for roughly $2 trillion of equity against $518 billion of mostly non-cancelable, decade-long compute commitments and a $42 billion loss ^[6]. The prospectus carries a take-or-pay clause — a contract that bills you even when you buy less than promised.

If our actual spend falls short, we must pay Google the difference. — Anthropic

Both labs' offerings run through the same two banks. Goldman Sachs and Morgan Stanley drafted OpenAI's prospectus and manage Anthropic's filing ^[7][6] — two of the three banks that captured a record $52 billion of equities trading revenue on the AI boom through September 30 ^[8]. The banks collect their fee whichever lab wins. The freshest number is the sobering one. A single dataset, hours old, shows realized token prices roughly halving in a month — Anthropic's realized price falling from over $1.70 in mid-September to $0.83 by October — while corporate AI spending stalls even as the volume of tokens rises ^[9]. That is context the sprint is priced against, not proof of anything yet; it is too thin and too new to carry a claim. The counterweight is real, which is what makes this a race rather than a collapse. Revenue is still growing 45 to 50 percent year over year ^[6]. Secondary demand for Anthropic shares ran three to five times oversubscribed this fall ^[10]. The bulls argue the backlog outpaces the spending and the bond market can absorb the full $5.5 trillion buildout ^[11]. By the labs' own word, the thing they meter — the token — is scheduled to keep getting cheaper. The sprint exists to fix, in advance, the price of the one thing in this industry not on that schedule: a share. And it has to be fixed before the first direct comparison reaches the market.