Percy's energy use estimate — methodology
Last updated: May 2026
Supporting documentation for the carbon-footprint question in our FAQ. If you're new to Percy, percy.garden is a better starting place.
This document explains how we estimate Percy's energy use across a full growing season. It's the methodology behind the figures referenced in our FAQ ("Does using AI mean Percy has a big carbon footprint?").
We've published this because we'd rather show the math than ask people to trust a tidy number. AI energy estimates can be flattering or unflattering depending on which assumptions you pick; here are ours, and where they could be wrong.
This isn't an energy research paper. It's a small team's best attempt to answer a question we wanted to be able to answer ourselves. The figures are estimates assembled from public research on similar models; they're not measurements, and they're not audited. We've published the methodology so anyone curious can see where the numbers come from and judge them for themselves.
What we're estimating
Percy's per-user energy footprint over a single growing season (roughly April through October in the US), measured in watt-hours of electricity consumed across the cloud services that run Percy. We don't try to estimate the energy used by your phone running the Percy app, your home Wi-Fi, or the upstream cost of model training — those are real but are common to nearly every app and not specific to Percy.
We use optimistic (high) usage assumptions deliberately. It's better to overestimate Percy's energy footprint and be honest than to lowball it.
What "an interaction" means
Throughout this document, "interaction" means a single user-initiated action that triggers at least one AI call. That covers:
- Logging an observation by voice (transcription + parsing)
- Logging an observation by text (parsing only)
- Identifying a plant from a photo
- Asking Percy a gardening question
- Reviewing or correcting an entry
A single observation can trigger several backend API calls (transcription, parsing, optional plant ID), but we count it as one interaction for the user-facing total.
Percy's model stack
Different tasks use different models, deliberately chosen to match the work:
| Task | Model | Why |
|---|---|---|
| Observation parsing, Q&A, photo moderation | A compact, efficient LLM | These tasks need fast, focused responses; a small model handles them well. |
| Plant identification follow-up | A more capable mid-tier LLM | Identifying plants accurately — especially distinguishing critical native host plants from dangerous invasives — requires reasoning over multiple ecological and visual signals. The accuracy upgrade matters ecologically. |
| Voice transcription | A compact speech-to-text model | Well-suited to short voice clips; substantially smaller than general-purpose AI. |
| Image-based plant ID | A specialized plant-identification image model | A purpose-built classifier is substantially more efficient per image than a general-purpose vision model. |
The single biggest lever on Percy's energy footprint is the choice to use a compact LLM for the high-volume tasks (parsing, Q&A) and reserve a more capable model only for the low-volume, accuracy-critical task (plant ID).
Per-interaction energy estimates
Cloud AI providers don't publish per-query energy consumption. The figures below are best-available estimates drawn from independent research on equivalent model classes, then applied to Percy's actual call patterns. We use the upper end of published ranges where the science is uncertain, to avoid undercounting.
| Interaction type | Components | Estimated energy per interaction |
|---|---|---|
| Voice observation | Speech-to-text + compact LLM parse | 0.08 – 0.15 Wh |
| Text observation | Compact LLM parse | 0.04 – 0.10 Wh |
| Plant identification (photo) | Specialized image model + mid-tier LLM + short follow-up safety checks | 0.8 – 1.8 Wh |
| Gardening question (Q&A) | Compact LLM | 0.05 – 0.15 Wh |
| Scheduled tasks (weather, summaries) | Batched, amortized across all users | ≈ 0.01 Wh |
Plant ID is the heaviest single interaction by an order of magnitude. A single plant ID involves a specialized image classifier for the species match, a mid-tier LLM for contextual identification, and short follow-up safety-and-quality checks. The combined call is roughly 10–20× more energy-intensive than a typical voice observation. We accept that tradeoff because misidentifying a critical host plant or a dangerous invasive has real consequences; we'd rather be slow and accurate here than fast and wrong.
Typical seasonal usage assumption
A representative engaged gardener over a six-month growing season (using deliberately optimistic, high-end usage numbers):
| Interaction type | Estimated count per season |
|---|---|
| Voice observations | ~80 |
| Text observations | ~40 |
| Plant identifications | ~30 |
| Gardening questions | ~40 |
| Other (corrections, reviews) | ~10 |
| Total | ~200 interactions |
For context, this assumes roughly one or two interactions per day on average across the growing season, with heavier days during planting and harvest. Light users do considerably less. The most engaged power users do more. We use the optimistic number to avoid undercounting.
Seasonal energy total
Applying the per-interaction figures to optimistic seasonal usage:
| Component | Math | Subtotal |
|---|---|---|
| Voice observations | 80 × 0.12 Wh | 9.6 Wh |
| Text observations | 40 × 0.07 Wh | 2.8 Wh |
| Plant identifications | 30 × 1.3 Wh | 39.0 Wh |
| Gardening questions | 40 × 0.10 Wh | 4.0 Wh |
| Other / batched | — | ~0.5 Wh |
| Estimated season total | ≈ 56 Wh |
We round this to approximately 50–60 watt-hours per growing season in public-facing copy. For comparison, a typical 9-watt LED bulb left on draws 9 Wh per hour — so Percy's full-season energy use is roughly equivalent to running a single LED bulb for an evening (about six hours).
Plant identification accounts for the largest single share of Percy's per-user AI energy footprint despite being only 15% of interactions. If we used a smaller model for plant ID, the total would drop significantly — but at a real ecological cost in misidentifications. This is the most consequential design tradeoff in Percy's energy profile, and it's a deliberate one.
What else does Percy run on?
The numbers above cover the AI services that do the heaviest work. Percy also runs on a small set of supporting cloud services for hosting, data storage, scheduled jobs, and operations. We estimated each one separately to make sure nothing meaningful was being hidden in a catch-all line:
| Service category | Role | Estimated season energy per user |
|---|---|---|
| Database, storage, and backend compute | Stores observations, runs the functions that process them, holds photos | ~3–6 Wh |
| Daily weather and daylight data pulls | One pull per active garden per day across the growing season | ~5–10 Wh |
| Website hosting and content delivery | Serves the marketing site and app shell on each visit | <0.5 Wh |
| Encrypted daily backups | Off-site backup of your observations for redundancy | ~0.2 Wh |
| Transactional email | A few account emails per user per year | <0.1 Wh |
| Privacy-respecting analytics | Counts visits to the marketing site | <0.1 Wh |
| Error tracking | Fires only when something breaks | negligible |
| Subtotal | ~10–15 Wh |
Adding this to the ~56 Wh from AI brings Percy's total estimated seasonal energy footprint to roughly 65–75 Wh per user, or about 7–8 hours of a 9-watt LED bulb — still well within an evening.
These figures are necessarily rougher than the AI estimates. Shared infrastructure like a managed database has a fixed baseline cost regardless of how many users it serves, so the per-user figure goes down as we grow. Today's per-user infrastructure cost at our current user count is higher than it would be at scale; the figures above reflect today's count, which is the more honest framing for now.
Converting to carbon
To translate watt-hours into greenhouse gas emissions, we apply the US national average grid carbon intensity from the EPA's eGRID dataset:
- US grid average: approximately 370 g CO₂-equivalent per kWh (EPA eGRID, most recent published year)
- Percy seasonal energy: ~70 Wh = 0.070 kWh
- Percy seasonal carbon: ~26 g CO₂-equivalent per user per season
Regional variation matters: a user on a coal-heavy grid (some Midwest and Mountain West areas) might be at 600+ g/kWh, while a user on a hydro- or nuclear-heavy grid (Pacific Northwest, parts of New England) could be at 50–100 g/kWh. We use the national average for the public-facing figure.
The garden center comparison
We say a round trip to the garden center produces "hundreds of times more carbon than Percy uses in an entire season." Here's that math:
- EPA estimates the average new gasoline-powered passenger vehicle emits roughly 400 g CO₂ per mile (US fleet average, EPA Automotive Trends Report)
- A 10-mile round trip: 4,000 g CO₂
- Versus Percy's seasonal ~26 g CO₂
- Ratio: ~155× more carbon for one garden center trip than Percy's entire season
A 15-mile round trip is roughly 230× Percy's seasonal footprint. The "hundreds of times" claim holds across a reasonable range of common garden-center distances, though we'd describe it as "roughly a hundred and fifty times" or "in the low hundreds" rather than implying an order of magnitude more than that.
Sources of uncertainty
Honest places where this estimate could be off:
- Per-query AI energy is the largest unknown. Providers don't publish it, and independent estimates vary by an order of magnitude. We use figures from published academic research on equivalent model classes (MLPerf benchmarks; independent measurements of comparable LLMs and image classifiers).
- Mid-tier LLMs specifically have wide reported energy ranges. Estimates for mid-tier reasoning models span roughly 0.3 to 3 Wh per query depending on input length, output length, and serving infrastructure. We use a midpoint that reflects the way Percy structures these calls to keep active inference modest per query.
- Model efficiency is improving. Today's models are several times more energy-efficient than the same class of model two years ago. If our model providers ship efficiency gains, these numbers come down.
- The 200-interactions assumption is deliberately high. A heavy power user might log 400+ interactions a season; a typical user logs closer to 100–150. Our copy uses the optimistic number, not the average.
- Training cost is not included. The energy used to train the AI models Percy depends on is real, but it's spread across all users of those models worldwide (often billions of queries). The marginal training cost per Percy user is small but nonzero. We exclude it because we can't measure it cleanly.
- Embodied energy of phones and infrastructure is excluded. This is standard practice for software carbon estimates, but it's worth knowing.
If any of these inputs are materially wrong in either direction, we'd want to know. Email us at hello@percy.garden if you spot an error.
Sources
- Grid carbon intensity: US Environmental Protection Agency, Emissions & Generation Resource Integrated Database (eGRID). https://www.epa.gov/egrid
- Vehicle emissions: US Environmental Protection Agency, Automotive Trends Report. https://www.epa.gov/automotive-trends
- Electricity reference data: US Energy Information Administration. https://www.eia.gov
- AI inference energy ranges: Independent academic measurements of small and mid-tier LLMs and image-classification models, published in peer-reviewed venues (MLPerf benchmarks; published energy-per-inference studies). Provider-specific per-query figures are not publicly disclosed by AI providers.
Treating this as a snapshot
These estimates reflect our best understanding as of the last updated date at the top of this page. Several of the underlying inputs — model efficiency, grid carbon intensity, our own typical usage — shift over time. We'll revise this when we get cause to (significant model changes, new published per-query figures from providers, material EPA updates, or a reader pointing out an error), but we make no promise about how often that will happen. If the date at the top is more than six months old, treat the numbers as directional rather than current.
Spotted an error or have access to better data than we used? Email us at hello@percy.garden — we read every one.
Questions or corrections: hello@percy.garden