Ramp data shows AI spending flattening just as Anthropic nears its IPO

Spending-management platform Ramp's transaction data shows enterprise AI usage continuing to climb while the money spent on it has flattened, according to a TechCrunch report published on 6 November. The finding lands as Anthropic approaches its IPO, a moment when public investors will be asked to price AI growth.

The mechanism the report points to is price: list prices have fallen and the heaviest spenders have become more disciplined about cost control, so the same volume of usage generates a smaller bill. For vendors charging by token, that breaks the link between usage growth and revenue growth.

Ramp's numbers come from corporate card and expense flows, which makes them a read on what companies actually pay rather than what vendors report as bookings. That is a useful cross-check on industry claims, though the sample skews toward firms that use Ramp for spend management.

The implication for the IPO story matters: if usage can grow while spend stalls, the hypergrowth assumption underlying current valuations needs a different justification, either new categories of usage that carry higher prices or a shift from per-token billing to outcome-based pricing.

Vendors have already been moving in that direction, bundling agents and workflow products priced by seat or by task rather than by tokens consumed. Whether buyers accept those framings is the open question the data raises.

Watch three things: revenue and gross margin disclosure in Anthropic's filing, whether other spend platforms publish similar curves, and whether agent products create a new billing unit that reverses the flattening. Until then, the gap between usage and spend is the number to argue about.