Meta's $2‑per‑Million‑Token Pricing Redefines AI Cost Structure

Meta Platforms announced that its Business Agent service will move to a per‑token pricing model of $2 per million tokens, shifting enterprise software economics from a per‑seat or per‑conversation basis to charging for AI output volume. The change creates a 40‑50× price advantage over competing per‑conversation models and will affect companies that run high‑volume customer‑engagement workflows.

At the same time, the rapid decline in foundation‑model inference costs is driving a valuation split in the healthcare AI sector. Deeply integrated health‑AI platforms that own proprietary data, EHR write‑back capabilities and regulatory clearances continue to command high revenue multiples (8‑20× ARR), while thin AI wrappers and point solutions without such defensibility are seeing multiples compress to 1‑3.5× ARR. Investors and acquirers are now rigorously distinguishing between platform‑level AI defenses and superficial AI applications.