Realtime AI News
Ramp data: AI usage soars while spending stalls as tokens turn into a commodity
Ramp's spending data shows AI token outlays grew just 1.5% across August and September combined, far below July's 42% month-over-month jump, even as token volumes kept rising. The numbers point to commoditization: price competition and customer-side cost control are separating usage growth from revenue growth.

Businesses are using more AI than ever, but the bills are not keeping up. Data compiled by Ramp shows AI token spending rose just 1.5 percent across August and September combined, down from a 42 percent month-over-month increase in July.
Ramp's index draws on real spending and invoice data from a large base of corporate customers, which makes it one of the higher-frequency reads on how fast businesses are actually adopting AI. What stands out in the latest numbers is the divergence from usage: token volumes kept climbing.
In other words, tokens are turning into a commodity. Unit prices are being competed down, discounts are becoming routine, and using more no longer means paying proportionally more. For model companies whose growth story assumes usage converts linearly into revenue, that is an uncomfortable signal.
Two forces are driving it. On one side, model providers are competing on price and handing out enterprise discounts to lock in accounts. On the other, customers are getting thrifty — capping budgets for long-running tasks, routing lighter requests to cheaper models, and deciding internally where expensive models are actually allowed.
The report is also a caution against reading one month as a trend: October data showed spending growth recovering, with token spend rising again. The shape of the curve looks less like demand collapsing and more like the end of hypergrowth and the start of a price-sensitive steady state.
For investors, the real question is where adoption spreads next. Heavy spending is concentrated among a small set of high-growth technology companies; if that cohort plateaus, it is unclear whether healthcare, retail and other traditional industries will spend enough to make up the difference.
What to watch is when the price war plateaus and whether AI budgets graduate from experimental line items into fixed costs. If usage keeps rising while spending stays flat, the valuation math for AI businesses has to change.
Why it matters
Decoupling usage from spend means AI revenue no longer grows automatically with adoption, pushing pricing pressure straight into model companies' valuations and fundraising plans.
Nearby Updates
All11/08, 05:35
Google adds prompt-based revisions to its Gemini Omni video model
Trend Hunter reported on November 7 that Google has added prompt-based revisions to Gemini Omni, letting users adjust generated video through natural-language instructions. The change pushes enterprise video generation from one-shot output toward conversational iteration, bringing editing closer to a text-style workflow.
11/08, 05:17
Memo claims a Singapore-based executive was recruited to spy on Nvidia's CEO
TechCrunch reported on November 7 that a memo circulated earlier this year claimed China's Ministry of State Security recruited a female technology executive to spy on Nvidia CEO Jensen Huang. Nvidia told the New York Times that Huang's interactions with the executive were "not unusual" and that the company does not share classified information with customers.
11/08, 03:14
NVIDIA Claims Vera CPU Doubles Agent Speed of AMD's 256-Core EPYC Venice
NVIDIA has published comparison data claiming its Vera CPU runs AI agent workloads twice as fast as AMD's 256-core EPYC Venice, pitching agent capacity per watt as a second metric. The numbers shift the data center CPU argument from core counts and throughput toward agent latency and energy efficiency.
11/08, 02:37
Coinbase's Brian Armstrong runs his work through a network of AI agents
Coinbase CEO Brian Armstrong said on November 7 that he now runs his work through a network of always-on AI agents, with a lead agent for each focus area overseen by a chief of staff agent and backed by sub-agents that write and test code. He still reviews every change himself, and the setup is personal rather than a company-wide rollout.