· AI
The AI Spending "Stall" Everyone Is Citing Is a Price Cut, Not a Pullback
A slowdown in dollars spent and a slowdown in AI use are two different facts, and this week's coverage keeps treating them as one.
Business spending on AI tokens grew a combined 1.5 percent across August and September, after climbing 42 percent in July alone, according to TechCrunch's reporting on data from the corporate card company Ramp. That ran days after reports that Anthropic's IPO investors have marked down their expected valuation from roughly 2 trillion dollars to the 1.5 to 1.6 trillion dollar range, with Blackstone reportedly telling investors to expect a junk-grade credit rating after listing, Stocktwits reported on November 3, citing Bloomberg and the Financial Times. Put those next to each other and the obvious read is that enterprise AI demand is cooling right when the biggest AI vendor needs to prove it is not.
That is not what Ramp's own numbers show. I pulled Ramp's AI Index directly. Token volume rose 37.1 percent in the four weeks ending October 23 versus the same window a month earlier, while total token spending rose only 8.9 percent over that period. The blended price businesses paid per million tokens across OpenAI and Anthropic fell 26.8 percent in the same stretch, from $0.76 to $0.56. OpenAI's blended price alone dropped 42.4 percent. Paid AI adoption among all businesses in Ramp's sample kept rising too, to 55.9 percent in September from 55.2 percent in August. None of that is a company pulling back on AI. It is a company using more of it for a bill that is not keeping pace, because the per-unit price is falling out from under the usage.
The steelman for the "stall" reading deserves a real hearing. TechCrunch's piece ties the flat spending directly to price cuts: OpenAI reportedly halved API prices for its GPT-6 Sol and Luna models in September, and Anthropic cut prices on Fable 5.1 and Opus 5.5 around the same time. Cutting your price and then reporting flat revenue growth is not a demand problem you can wave away, it is margin compression a company chose to accept, and a company staring down a 1.5 to 1.6 trillion dollar IPO and a possible junk rating has every reason to worry about what that compression does to its growth story. Ramp's own index shows how concentrated that revenue base is: the top 1 percent of AI-paying firms spent roughly $9,568 per employee per month in September, versus about $16 for the median firm, a gap of more than 600 to 1. A revenue base that lopsided is also a fragile one. If a handful of the biggest accounts in that top tier tighten their budgets in the same quarter, the vendor feels it immediately. That is a legitimate risk for the vendors and for anyone pricing their equity.
But it is a vendor-side risk, and the coverage keeps importing it into a customer-side conclusion that the data does not support. If you spent $1,000 on tokens last month and $900 this month while running more requests through the model, you did not cut back on AI. You got a volume discount nobody had to negotiate for you, because two well-funded labs are undercutting each other's prices to keep your business. That is what price competition is supposed to do, and the fact that it shows up as "stalled revenue" on a vendor's income statement is the vendor's problem to solve, not evidence that the technology stopped being worth buying.
The number worth watching at your own company is not Anthropic's valuation. It is whether your AI bill is still climbing in lockstep with your usage. If it is, you are not capturing the price war that is visibly happening in Ramp's own data, you are paying list price while your vendor quietly cuts rates for everyone willing to ask, switch models, or route routine work to a cheaper tier. That is the exact kind of model and vendor decision MojoAI gets brought in to sort out: which workloads still need the expensive model, which ones never did, and whether your current contract reflects a market where blended prices just fell more than a quarter in two months. If you have not checked that lately, get in touch and let's look at what you are actually paying for.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.