Agentic AI Can Raise Token Use per Task Up to 100 Times, Accelerating the Shift Away From Per-Token Pricing, Futurum Research Finds
QumulusAI announced the release of a Futurum Research report it sponsored, finding that agentic AI can increase token consumption per task by 10 to 100 times versus a simple inference call, pressuring per-token pricing economics. The report, based on a survey of 824 AI decision-makers, finds reserved and owned infrastructure already account for 66% of AI compute consumption and forecasts inference spending growing from $120 billion in 2025 to $885 billion by 2030.
Key figures
- Survey respondents
- 824
- Inference spend 2025
- $120 billion
- On demand cloud share
- 19%
- Token increase per task
- 10 to 100 times vs simple inference call
- Ai first cloud growth 2026
- 107.1%
- Reserved owned infra share
- 66%
- Ai first cloud 2030 forecast
- $375.2 billion
- Inference spend 2030 forecast
- $885 billion
- Workloads outside hyperscalers
- 59%
- Agent reasoning inference growth 2026
- 219%
AI analysis
Red flags1 · Pro
The rest of the AI analysis, red flags and sentiment are part of Signal8 Pro.
AI-generated analysis of a public disclosure. Not investment advice; verify against the original document.