AI has a real cost.
We offset it ten times over.

We measure what our AI takes from the world. Energy. Water. The mining behind every chip. Then we donate enough to leave it better than we found it.

The Standard

10×

Every month, we donate enough to offset ten times the environmental cost of every NebulaSEO user’s AI usage.

Per location · per month

What it actually costs.

Every location on NebulaSEO runs on AI all month: the posts and review replies we draft, the audits and fixes we run, and every conversation their chat widget and phone agent handle for their own customers. The physical footprint of one location’s month, estimated from published per-inference figures for the AI infrastructure we run on:

Carbon Emission

0.00 kg CO2e

roughly equivalent to driving a gas car two and a half miles

Water Consumption

0.0 L H2O

roughly a large soda bottle, drawn from data-center cooling

Material Extraction

0 g Material

roughly an AA battery’s weight, amortized across each chip’s lifecycle

Sources
  • Carbon: derived from Luccioni et al., “Estimating the Carbon Footprint of BLOOM” (JMLR 2023). 1.5 gCO₂e per inference, applied across a location’s full monthly AI usage, including the conversations its chat widget and phone agent handle. We use this figure rather than the lower per-prompt numbers some providers now publish, because we would rather overstate our cost than understate it.
  • Water: derived from Li et al., “Making AI Less ‘Thirsty’” (UC Riverside, 2023). On-site cooling water per prompt, conservative end of published range.
  • Material: amortized GPU lifecycle estimate. No peer-reviewed per-query figure exists; the number reflects rare-earth and bulk-mineral content of an AI-class GPU divided across its expected inference volume. Reported here in good faith and revised as better data emerges.

Per person, small.
Across thousands of users, a problem.

Allocation

Donations are split across three organizations, chosen because their work measurably undoes the specific damage our AI causes.

Climeworks

70% · Carbon
What they do

Direct-air-capture plants in Iceland that pull CO₂ from ambient air and mineralize it permanently into basalt rock, 1,000m underground. Climeworks reports its removals are verified and audited by DNV.

Why them

AI is mostly an energy-and-emissions problem. Climeworks doesn't store carbon temporarily, it removes it from the atmosphere. That's the only thing that actually undoes the damage.

The Freshwater Trust

15% · Water
What they do

Restores measurable freshwater habitats in the western US: riparian planting, irrigation upgrades, watershed projects. Impact is measured in gallons returned to depleted rivers.

Why them

Every AI generation cools a data center somewhere. The Freshwater Trust puts measurably more water back than NebulaSEO's infrastructure consumes.

Pure Earth

15% · Mining
What they do

Cleans up the world’s worst toxic pollution sites: lead smelters, mercury mining waste, chromium plants. Pure Earth reports 120+ sites remediated across 50+ countries.

Why them

Every GPU contains rare-earth metals pulled from mines whose downstream damage poisons real communities. Pure Earth cleans up the same supply chain that builds our infrastructure.

The build itself

Built by humans, with AI assistance.

Every decision in this product was made by a person. A person scoped each feature, reviewed the work, tested it, and chose what shipped. And through all of it we used AI every single day. Plenty of companies do the same and stay quiet about it. We’d rather say it plainly and pay for it honestly.

Made by people

Human judgment at every gate: what to build, what counts as done, what never ships. The AI drafted; people decided.

AI assistance

Millions of AI tokens went into writing, testing, and re-testing this platform. That usage drew real electricity and real water, the same costs we account for above.

Offset in full

We donated enough to cover the entire build’s AI usage, not only our customers’. The receipt will be published with the rest of the year’s donation documents on the date at the bottom of this page.

We refuse to:

  • Call ourselves “net-zero” or “carbon-neutral.” Both are accounting labels that hide more than they reveal.
  • Pay for cheap, unverified carbon credits. Paper certificates from forestry projects that often never plant a tree or double-count the same forest twice.
  • Hide what our AI infrastructure actually costs the planet. The numbers above are estimates from published research, shown with their sources, and ours to defend.
  • Overstate our role. We are doing our share. We are not single-handedly solving climate change, and we will never claim to be.

“The greatest threat to our planet is the belief that someone else will save it.”

Robert Swan
Polar explorer · Environmental advocate
First Annual ReceiptDecember 31, 2026Itemized donations and offset verification will be published here on the date above.