Break-even index / 2× RTX 3090 / Claude Sonnet 4.6 / Maryland
Will a Dual used RTX 3090 beat Claude Sonnet 4.6 in Maryland?
Break-even lands around month 19. Real, but slow — a hybrid setup (local for the routine work, cloud for the peaks) usually beats going all-in.
How to read it: OWN — the rig pays for itself inside 18 months. HYBRID — break-even lands between 18 and 36 months; real, but slow. RENT — it never pays off before the hardware is due for replacement. Every formula and threshold is public on the methodology page.
The numbers
| Build cost (Dual used RTX 3090) | $2,250 ($2,000–$2,500) |
|---|---|
| Hardware amortized (3 yr) | $62.50/mo |
| Electricity in Maryland ($0.180/kWh) | $14.58/mo (81 kWh) |
| Local total cost of ownership | $77.08/mo |
| Claude Sonnet 4.6 at the same usage | $135/mo |
| Break-even | about 19 months |
Assumes the "coding assistant" preset: ~15M tokens/month, 3 h/day active, 5 h/day idle. Claude Sonnet 4.6 priced at $3/M in, $15/M out (verified 2026-07-05). Rig ceiling: 70B at Q4, comfortably. Jargon: VRAM is the GPU memory that decides which models fit; Q4 is 4-bit quantization — a smaller, faster copy of a model with a slight quality cost.
See your verdict — then tweak every number →