Transcription Service: Cost at 100 Hours/Month
What does it cost to run transcription service at 100 hours/month? Self-hosted dedicated GPU vs API provider pricing.
Monthly Cost Comparison at 100 hours/month
| Provider | Monthly Cost | Pricing Model | vs GigaGPU |
|---|---|---|---|
| GigaGPU (RTX 5060 Ti) | £119/mo | Fixed | — |
| OpenAI Whisper API | £60/mo | Per-hours | API is cheaper at this volume |
| AWS Transcribe | £144/mo | Per-hours | 52% cheaper with GigaGPU |
| Google Cloud STT | £96/mo | Per-hours | 28% cheaper with GigaGPU |
The Honest Numbers at Low Volume
A small clinic digitising patient consultations. A podcast studio batch-processing interview recordings. At 100 hours/month, the raw API cost from OpenAI Whisper (£60/mo) actually undercuts a dedicated RTX 5060 Ti at £119/month. So why consider self-hosting at all?
Because the API price only tells half the story. AWS Transcribe charges £144/mo and Google Cloud STT hits £96/mo for the same volume. More importantly, API pricing shifts without warning — and at 100 hours you are already close to the crossover point where any rate increase tips the balance toward dedicated hardware.
Annual savings potential: Up to £900 per year compared to the most expensive API option, assuming consistent 100 hours/month usage.
Why Teams Choose Dedicated Hardware Even at 100 Hours
- Data stays on your server: Medical, legal, and financial transcription requires data never leaving your infrastructure. No third-party data processing agreements to negotiate.
- No rate limits or throttling: API providers impose concurrency limits that force queuing during peak hours. A dedicated GPU processes your backlog on your schedule.
- Budget ceiling: £119 is £119 regardless of whether you process 80 hours or 130 hours in a given month. API costs fluctuate with every minute of audio.
- Model flexibility: Fine-tune Whisper on domain-specific terminology — medical jargon, legal citations, regional accents — impossible with hosted APIs.
When the API Still Wins
- Sporadic usage below 80 hours: If your volume fluctuates heavily and often dips below 80 hours, OpenAI Whisper API at £0.60/hour keeps costs lower.
- No ops capacity: Small teams without DevOps expertise benefit from zero server management.
- Multi-language prototyping: Testing across dozens of languages before committing to a production pipeline.
Hardware Recommendation
The RTX 5060 Ti at £119/month handles 100 hours/month of Whisper Large v3 inference comfortably, with headroom for 20-30% burst capacity. GigaGPU servers ship pre-configured with CUDA, Docker, and popular inference frameworks — deploy Whisper in under 15 minutes.
Lock In Fixed Pricing for Transcription
Get predictable transcription costs with a dedicated GPU — no per-hour fees, no surprise invoices at month-end.
