Environmental impact

The honest footprint of your deck.

We paint every card with AI, and yes — AI uses energy and water. So rather than hand-wave it away, we did the math with the most cautious published numbers and set it beside the everyday things you already do. Here's what creating one personalized deck really costs the planet.

Energy

~0.3 kWh

≈ 4 hours of streaming

per full 78-card deck

Carbon

~130 g CO₂

≈ 90 minutes on Instagram

world-average electricity grid

Water

~1–3 L

≈ a couple of bottles of water

data-center cooling & power

Put it in perspective

Energy, side by side

One deck is a one-time thing. Here's how its energy compares with a few ordinary moments and chores, on the same scale.

Kilowatt-hours (kWh) — lower is lighter

Your full deck0.30
Streaming a 2-hour film0.15
An evening of TV (4 h)0.31
One load of laundry1.0
One dishwasher cycle1.5

The same energy as a few hundred web searches.

The carbon of driving under half a mile.

About three kettle boils of electricity.

For scale: producing one beef burger takes about 2,500 L of water.

How we keep it small

By design

We never re-paint a card we've already made

Our engine only generates the cards that are actually missing — nothing is wastefully re-computed. Edits touch a single card, not the whole deck.

Digital-first — nothing ships unless you ask

Your deck lives as high-resolution files by default. Paper, ink, and shipping — by far the largest part of any physical product's footprint — only happen if you choose a printed edition.

Efficient models on shared infrastructure

We use modern, optimized image models running on shared cloud data centers — not a dedicated machine humming away — so the cost of each card keeps falling as the technology improves.

And it's getting lighter every year

This isn't a fixed number — it's a shrinking one. The same deck costs a fraction of what it would have two years ago, thanks to real, measurable progress across the industry.

~10×

more energy-efficient image models than the 2023 benchmarks — and improving fast.

↓ CO₂

Electricity grids keep decarbonizing, so the same work emits less carbon each year.

−24%

Data-center water use per unit of energy fell in a single year, with new near-zero-water cooling arriving.

A deck today costs a fraction of what the same deck would have two years ago — and that number keeps dropping.

1%

We go one step further

On every single order, we direct 1% of our revenue to permanent carbon removal through Stripe Climate. It funds a portfolio of frontier technologies that pull CO₂ out of the atmosphere and lock it away — and 100% of it reaches those projects, with no fees taken along the way.

We're honest about this: 1% helps grow and scale the carbon-removal industry — we're not claiming each deck is perfectly "carbon neutral." It's a genuine contribution on top of an already small footprint.

Show your work

The math

We'd rather over-estimate honestly than cherry-pick a flattering number. Here's exactly how we got there.

How we calculated one deck's footprint

We modeled the largest deck we make — a full 78-card deck (smaller decks scale down proportionally).

  • Renders per deck: 78 finished cards, plus a free preview and the occasional re-render for quality — we assume about 100 image generations in total (a generous 25–30% overhead).
  • Energy per render: published studies put a single AI image between ~0.5 Wh (efficient modern models) and ~11 Wh (older, large 1024px models). We deliberately used the cautious cross-model average of ~2.9 Wh — not our own optimistic figure.
  • Result: about 100 × 2.9 Wh ≈ 0.3 kWh per deck → roughly 130 g CO₂ on a world-average grid (less on cleaner European grids), and around 1–3 L of water for cooling and power generation.

What we don't hide: these are estimates with real uncertainty. Training the underlying models has its own footprint, but it's spread across many millions of uses, so per-deck it's tiny. Where a range existed, we leaned toward the higher number so the story stays honest even in the worst case. Everything above comes from independent, public research.

Sources

Independent research
  1. 1Luccioni, Jernite & Strubell (2023) — Power Hungry Processing: Watts Driving the Cost of AI Deployment? arXiv:2311.16863
  2. 2MIT Technology Review (Dec 2023) — Making an image with generative AI uses as much energy as charging your phone.Read
  3. 3International Energy Agency (IEA) — The carbon footprint of streaming video (~0.077 kWh & ~36 g CO₂ per hour). Read
  4. 4Carbon Trust — European video streaming ≈ 55 g CO₂e per hour.
  5. 5Ren et al., UC Riverside — Making AI Less “Thirsty” (data-center water per query). arXiv:2304.03271
  6. 6Lawrence Berkeley National Laboratory (2024) — U.S. Data Center Energy Usage Report (water-usage effectiveness). Read
  7. 7Hannah Ritchie — Sustainability by Numbers (2025) — AI’s energy footprint in perspective. Read
  8. 8Stripe Climate & Frontier — how the carbon-removal commitment works. Read

Create something that lasts.

A personalized keepsake, made with a light touch.

Design your deck

All figures are conservative estimates drawn from the independent sources listed above.