Table of Contents
AI budgets are often submitted as a single line. They evolve dramatically across phases.
TL;DR
Year 1: ~£5K-15K total (pilot + small production). Year 2: ~£30K-80K (production at scale). Year 3: ~£100K+ (enterprise multi-region). Mostly server cost; engineering time is the bigger line item.
Three phases
- Pilot (months 1-3): hosted API + small dedicated server. ~£500-1,000/mo.
- Production (months 4-12): 1-2 dedicated GPUs + LiteLLM + monitoring. ~£2K-5K/mo.
- Scale (year 2+): multi-server, observability, eval pipeline, fine-tuning. ~£5K-15K/mo.
Real numbers
| Team / scale | Server | Hosted-API fallback | Eng time / mo |
|---|---|---|---|
| Solo founder | £169 | £100 | 20% |
| Startup, 1K users | £359 | £200 | 50% × 2 |
| Mid-size SaaS, 10K users | £900 | £500 | 100% × 2-3 |
| Enterprise | £3K-10K | £1K-3K | team of 4-8 |
Verdict
Server cost is small relative to engineering. Plan team alongside infrastructure.
Bottom line
Engineering time is the bigger budget line. See team roles.
