Palo Alto Networks CEO Narayan Murthy has stated that AI token efficiency needs to improve dramatically, dropping costs by up to 20% over the next 12 months and by 90% by the following year.

  • Token efficiency must drop to as much as 20% within 12 months.
  • A further reduction to 90% is required by the subsequent year.

The article highlights these aggressive cost reduction targets for AI inference without detailing specific technical methods.