{
 "title": "On-Prem AI Inference Break-Even 2026: Own vs Rent UK",
 "url": "https://www.servnetuk.com/research/own-vs-rent-ai-inference-break-even-2026",
 "publisher": "Servnet Research",
 "license": "CC BY 4.0 — free to reuse with attribution",
 "collected": "2026-07-19",
 "records": [
  {
   "metric": "On-prem annual cost (8-GPU A100, UK, 2026)",
   "value": "£75,000–£90,000/year",
   "source": "Technolynx"
  },
  {
   "metric": "Cloud annual cost (8-GPU A100, UK, 2026)",
   "value": "£200,000–£240,000/year",
   "source": "Technolynx"
  },
  {
   "metric": "Break-even utilisation (on-demand vs on-prem)",
   "value": "40–60%",
   "source": "Technolynx"
  },
  {
   "metric": "Break-even utilisation (reserved vs on-prem)",
   "value": "60–80%",
   "source": "Technolynx"
  },
  {
   "metric": "Cost advantage (on-prem vs cloud IaaS, per million tokens)",
   "value": "8× lower",
   "source": "Sesame Disk"
  },
  {
   "metric": "Cost advantage (on-prem vs GenAI API, per million tokens)",
   "value": "18× lower",
   "source": "Tilkal"
  },
  {
   "metric": "Real break-even utilisation range",
   "value": "55–75%",
   "source": "YouTube (Ep.013)"
  },
  {
   "metric": "Break-even tokens/day (local pays off)",
   "value": "5M+ tokens/day",
   "source": "MindStudio AI"
  }
 ]
}