QumulusAI-backed Futurum report says agentic AI boosts token use per task up to 100x
QMLS•A QumulusAI-backed Futurum Research analysis says agentic AI can increase tokens used per task 10-100x, raising cost volatility under per-token pricing. It forecasts inference spending will rise to $885 billion by 2030 from $120 billion in 2025.
1. AI cost and deployment trends
The analysis says agentic AI lifts tokens per task 10-100x, increasing cost volatility under per-token pricing. It reports that reserved or owned infrastructure accounts for 66% of AI compute consumption, versus 19% for on-demand cloud, and that 59% of 824 surveyed AI decision-makers primarily run workloads outside hyperscaler public clouds. The report forecasts inference spending of $885 billion by 2030, up from $120 billion in 2025, and favors hybrid deployments that shift predictable, high-volume inference to reserved bare-metal capacity.
