Running AI in the cloud consumes energy, which translates into carbon emissions. To understand what it would take to offset that impact, we can look at the Amazon rainforest’s ability to capture and store CO₂.

A single GPU cloud instance emits roughly 18 tonnes of CO₂ per year. Offsetting this requires about 0.09 acres of rainforest—equivalent to $360–540 annually at common carbon credit prices. For eight GPUs, the footprint rises to 144 tonnes per year, or about 0.71 acres of forest, costing $2,880–4,320 annually.

Viewed monthly, the numbers are easier to grasp. One GPU produces about 1.5 tonnes of CO₂ per month. That can be offset by protecting or restoring about 0.018 acres of rainforest (roughly 780 square feet), at a cost of $30–45 per month. For eight GPUs, the monthly footprint is about 12 tonnes of CO₂, requiring 0.148 acres (around 6,450 square feet) of rainforest, at $240–360 per month.

In practical terms, each GPU instance requires the protection of a small patch of rainforest each month—the size of a basketball court. Eight GPUs require an area closer to three-quarters of a football field. By linking cloud usage to tangible offsets, we make Net Zero commitments clear, measurable, and achievable.