Model Lifecycle TCO
Estimate $/token and total cost across training, fine‑tuning, and inference — see how utilization, power price, and hardware choices move your unit economics.
KPIs
Token Cost ($/1k)
Unit cost per 1,000 tokens across the model lifecycle, including amortized training/fine‑tuning and inference OPEX.
Higher is worse
Obfuscated preview — sign in to view exact values
USD/1k tokens
Lifecycle Total Cost
Total cost over the planning horizon: training + fine‑tuning + inference.
Higher is worse
Obfuscated preview — sign in to view exact values
USD (millions)
Training Cost
Total cost of training runs for the next major model release.
Higher is worse
Obfuscated preview — sign in to view exact values
USD (millions)
Fine‑Tuning Cost
Total cost of fine‑tuning and alignment runs for this release window.
Higher is worse
Obfuscated preview — sign in to view exact values
USD (millions)
Inference Cost
Monthly OPEX of serving inference, including power, hardware amortization, and platform overhead.
Higher is worse
Obfuscated preview — sign in to view exact values
USD (millions)
TCO Index
Normalized 0–1 index summarizing lifecycle total cost for scenario comparison (higher = worse).
Higher is worse
Internal Factors
Utilization Rate
Share of accelerator wall‑time spent doing useful work (active/available).
Higher is better
Power Price ($/MWh)
Blended electricity price per MWh after tariffs and hedges.
Higher is worse
Power System Overhead Index (from PUE)
Facility power overhead fraction derived from Power Usage Effectiveness (PUE); 0 = perfect, higher = worse.
Higher is worse
Accelerator Installed Base
Count of deployed accelerators available for training/inference.
Average Sequence Length
Average total tokens per request (prompt + completion).
Higher is worse
Energy Consumption
Monthly facility‑level energy usage for the AI fleet.
Higher is worse
Levers
Utilization Target
Scheduling/placement target for fleet utilization subject to SLAs.
Power Hedge Share
Share of expected load under fixed/hedged electricity pricing.
PUE Improvement CAPEX
Capital expenditure aimed at lowering facility overhead (cooling/power path).
Obfuscated preview — sign in to view exact values
USD (millions)
Accelerator Generation Choice
Chosen accelerator generation for the fleet (e.g., H100, B200, MI300X) with implied perf/W and memory profile.
Model Release Cadence
Days between major model releases; ties training budget to roadmap.
Context Length Limit
Platform cap on total tokens per request to bound inference cost tails.