Running a personal AI assistant sounds expensive. API bills spiraling out of control, hundred-euro invoices. I put it off for months.
Then I just did it. My first month? €19. Less than Netflix and Spotify combined.
Here is the breakdown:
The Two Costs
Your own AI assistant has two cost components:
- Hosting (the server it runs on)
- AI API (the language model)
With managed hosting + Google Gemini free tier:
- Hosting: €19/month
- Gemini Free API: €0 (20-50 requests/day)
- Total: €19/month
The API Fear is Overblown
VentureBeat calculated you would need to process 74,000 pages of text daily to hit $180 in API costs. That is physically impossible for one person.
Real usage? Most people do 20-50 requests a day. Draft emails, check calendar, summarize research, quick questions. All covered by free tiers.
What About DIY?
A €6 VPS sounds cheaper. But factor in:
- 15+ hours initial setup
- Docker, SSL, firewalls, reverse proxy
- Ongoing maintenance
- Risk of losing data (no backups configured)
If you value your time at €50/hour, DIY costs €3,000-5,000 in the first year. Managed hosting costs €228.
Budget API Options
Google Gemini Free: 20-50 requests/day, zero cost
DeepSeek: $0.27 per million tokens (95% cheaper than GPT-4)
OpenRouter: Test multiple models, pay only what you use
Full breakdown with comparisons: https://clawhosters.com/blog/posts/own-ai-assistant-costs-clawhosters
Top comments (0)