
Gelila Bekele, Contributor
· 1 min read
How AI Labs Are Evolving Ahead Of Their IPOs
Just 1% of US businesses account for 80% of spending on OpenAI and Anthropic, according to Ramp data.
If a few firms pay for most frontier AI spend, where is the rest of the work going? Increasingly, to open-weight models, which cost about 6 cents for every dollar a frontier model costs on the same benchmark tasks. Spending on open-weight models routed through Vercel’s AI Gateway grew more than seventyfold between December 2025 and September 2026, according to Vercel.
Frontier models remain the most capable and most expensive, but open-weight models keep taking on more of the work. That leaves the labs one commercial question: what work will still require a frontier model, and how much of it is there? Recognizing this, AI labs are changing both what they offer and how they charge for it while they speed toward their IPOs.
I have been examining this shift with Siddhant Sahu, an AI product lead at Amazon Web Services (AWS) who builds and launches AI infrastructure products using open and closed-source models. In his view, “customers are actively reducing spend on frontier model AI budgets for employees, as characterized by Apple & Tesla, while also migrating the workloads that can use non-frontier intelligence to open-weight models such as Coinbase. The premium that frontier AI labs currently have will only hold while the gap between commoditized intelligence and premium intelligence stays wide enough for customers to justify the higher price.” This article follows the labs’ response through five open questions.
1. Why are AI labs moving from models to products?
MORE FOR YOU
Staying at the frontier requires capital few can match. OpenAI has secured about $182 billion in committed capital and Anthropic has raised $141 billion, together about six times the combined funding of the next four labs. That capital earns a return only if customers keep paying more for frontier models than for the alternatives.
Original source
This story was published by Forbes: Innovation and written by Gelila Bekele, Contributor. SyncAI.news shows a preview; the complete article is on the publisher's site.
Read the full story on forbes.com


