Yield per watt and the cost of a gramYield per watt and the cost of a gram

Purpose and scope

Most grow-room conversations are about plants. Whether the room survives is decided somewhere less romantic: a division. All the dollars you spent in a year, over all the grams you sold. If that number is below your selling price, you have a business. If it isn't, you have an expensive hobby with a licence attached, and no amount of terpene talk changes it.

The trouble is that the industry's favourite yardsticks, grams per square metre, grams per watt, were built for other arguments. They are agronomy metrics and forum-bragging metrics, and they each quietly delete part of the bill. This paper walks through the three common denominators and what each is actually for, hedges the published benchmarks hard (because they deserve it), then builds a complete cost stack for a fictional 100 m² (1,076 ft²) room with every step of the arithmetic shown. From there: labour (the cost line that sneaks up on almost everyone), cycles per year (the hidden multiplier), quality tiers, a sensitivity tornado, and break-even thinking.

If you can divide two numbers, you can follow all of it, the entire discipline of unit economics is choosing which two numbers to divide.

Definitions

Nine terms carry the rest of the paper. Most economic arguments between growers are two people using the same word for different fractions.

Evidence and limitations

Cost-per-gram summary

g/m², g/W and g/kWh: selecting the right metric

All three metrics divide the same harvest by a different resource, and each answers a different question. The mistake is not using them. It's using one of them as the scoreboard and forgetting what it can't see.

g/m² of canopy is the agronomist's number. It compares crops, cultivars and steering decisions on the same floor plan, and it's the number most research reports. It contains no time, a nine-week cycle and a twelve-week cycle can post the same g/m² while one produces 30% more per year, and no power, no labour, no grade mix.

g/W of installed light is a relic of the lamp-shopping era, and Section 06 gives it a full autopsy. It usefully asks ‘how much crop per unit of lighting hardware’, but the denominator is nameplate watts: it ignores how long the lamps run, everything the HVAC burns, and, fatally, which fixture generation produced the watts.

g/kWh all-in (or its reciprocal, kWh per kg) divides by every kilowatt-hour through the meter, lights, HVAC, dehumidification, pumps, the lot. It is the one denominator that reconciles against a document someone actually sends you: the power bill. It's also the industry's formal benchmarking metric, Resource Innovation Institute's PowerScore scores facilities on exactly two numbers, kWh per unit of flowering canopy and grams per kWh, across 350+ producers^[1]. The spread is enormous: indoor production uses on the order of 18× the energy per gram of outdoor^[2], which is why an indoor room lives or dies on this metric while a greenhouse barely thinks about it.

Published yield benchmarks and their limits

Published cannabis yield figures are a minefield of mixed conditions, mixed denominators and outright projection. Before you benchmark against anything, look at what the honest sources actually report, and how far apart they are.

The forensic baseline. The most honest large-sample g/m² figure in the literature is also the oldest: Dutch police weighed confiscated illicit grows, and the model for the median room, 15 plants/m² under 510 W/m² of HPS, came out at 33.7 g per plant, 505 g/m²^[3]. Note the division: 505 g/m² over 510 W/m² is 0.99 g/W. That single study is almost certainly where ‘a gram per watt’ folklore comes from, a median, from HPS rooms, twenty years ago.

The controlled trials. Potter & Duncombe grew under 270, 400 and 600 W/m² of HPS and measured 0.9–1.6 g/W, with the best gram-per-watt result at the lowest wattage^[4]. More light grew more grams but fewer grams per watt: diminishing returns per unit of power, measured. Modern LED work shows the same shape from the other side, dry flower yield kept climbing roughly linearly with light intensity up to ≈1,800 µmol·m⁻²·s⁻¹ with no plateau^[8], and a follow-up at 600–1,000 µmol found each extra 100 µmol worth ≈4.6 g/plant (≈51 g/m² at ~10 plants/m²), for yields of roughly 276–447 g/m² across that range^[6]. Bugbee's group, growing high-light hemp for cannabinoids, reported 500–750 g/m² across three trials^[7].

The meta-analysis, and why you hedge. Backer et al. pooled the literature and found reported efficiencies of 0.31–1.97 g/W, a 6× spread, and scaled-up yield projections running from 3.4 to 3,590 g/m², a thousand-fold range driven by extrapolating small-plot numbers to areas nobody actually grew^[5]. They also found that raising installed W/m² reduced yield per watt, and that longer flowering periods raised yield per m². Both of which are denominator stories, not plant stories.

Grams per watt as a legacy lighting metric

‘A gram a watt’ was a useful rule of thumb when every serious room ran the same lamp. Under LED it has quietly become a measure of when you bought your fixtures, because the denominator changed underneath the metric.

A fixture converts watts into photons, and the exchange rate is called efficacy, in µmol of photons per joule. Double-ended HPS, the lamp the folklore was built on, delivers about 1.72 µmol/J. The best LED fixtures measured in 2020 hit ≈3.0 µmol/J (blue/red) and 2.78 (white/red), against practical ceilings around 3.4–4.1^[9]. In 2014 the best LEDs managed 1.7 — HPS parity. In one fixture generation, the photons bought per watt roughly doubled.

Now watch what that does to g/W with zero agronomy. Take the same fictional crop, 450 g/m² at 900 µmol·m⁻²·s⁻¹. Delivering 900 µmol with 1.72 µmol/J HPS takes 900 ÷ 1.72 ≈ 523 W/m²; with a 2.8 µmol/J LED it takes 900 ÷ 2.8 ≈ 321 W/m². Same photons, same plants, same grams. The HPS grower reports 450 ÷ 523 = 0.86 g/W; the LED grower reports 450 ÷ 321 = 1.40 g/W, and neither of them grew better than the other.

The efficacy shift also rewrites the buying decision. In Bugbee's lighting trials the white+red LED yielded 4.6% less per m² than HPS, and produced 27% more per dollar of electricity^[7]. Judged on g/m², the LED loses. Judged on the metric that pays bills, it wins comfortably. Same data, different denominator, opposite decision. Which is the entire argument of this paper in one experiment.

Cost components per gram

Cost per gram is built from a short, boring list. The skill isn't clever accounting. It's refusing to leave lines out. Eight lines cover a small indoor facility:

• Labour, wages plus the on-costs (leave, insurance, tax) for everyone who touches the crop, including you at a market rate.
• Energy, lights, HVAC, dehumidification, pumps, controls. All of it, off the bill, not off the fixture nameplate. For context on how dominant this line is indoors: US indoor production was estimated at 1% of national electricity a decade ago^[10], and modelled emissions run 2,283–5,184 kg CO₂e per kg of flower depending on climate^[11].
• Media + nutrients, substrate, salts, CO₂, IPM consumables.
• Rent, on gross floor area, not canopy.
• Depreciation, the fit-out and gear, spread over useful life.
• Testing + compliance, lab panels per batch plus licences, QA time, records. California data put mandatory testing alone at ≈$136 per pound (≈$0.30/g) once sampling and failure rates are counted^[12], a real line, not a rounding error.
• Packaging + consumables, bags, totes, labels, gloves.
• Other overhead, insurance, security, admin, repairs, software.

Worked example: a fictional 100 m² room

1. 1
   Fix the canopy and the light

   100 m² × 350 W/m² = 35,000 W = 35 kW installed. Sanity-check the intensity: 350 W/m² × 2.6 µmol/J = 910 µmol·m⁻²·s⁻¹, a normal LED flower target.
2. 2
   Grams per cycle

   450 g/m² × 100 m² = 45,000 g per cycle.
3. 3
   Cycles per year

   56 flower days + 7 turn days = 63 days flip-to-flip. 365 ÷ 63 = 5.8 cycles per year.
4. 4
   Grams per year

   45,000 g × 5.8 = 261,000 g = 261 kg per year.
5. 5
   Lighting energy

   35 kW × 12 h × 56 days = 23,520 kWh per cycle of lighting.
6. 6
   All-in energy

   23,520 × 2.2 = 51,744 kWh per cycle → × 5.8 ≈ 300,000 kWh per year. Cross-check: 300,000 ÷ 261 kg ≈ 1,150 kWh per kg, efficient-end for indoor; plenty of real rooms run 2–4× this^[2].
7. 7
   Price the energy

   300,000 kWh × $0.20 = $60,000 per year.
8. 8
   Add the rest of the stack

   Labour $200,000 · rent $60,000 · testing + compliance $46,000 · depreciation $43,000 ($300,000 ÷ 7) · other overhead $40,000 · media + nutrients $26,000 · packaging $20,000. With energy: $495,000 per year.
9. 9
   Divide

   $495,000 ÷ 261,000 g = $1.90 per finished gram. That is the room's real scoreboard, everything else in this paper is a way of moving it.

Now score the same room on every denominator from Section 04, so you can see what each lens would have told you:

Labour costs

Ask a new grower what indoor production costs and they'll talk about power. The fictional room's power bill is $0.23 a gram. Its people are $0.77 — the largest line by a factor of three, and the one most plans either omit or price at zero because ‘I'll do it myself’.

Start with the honest division: $200,000 of payroll over 261 kg is $766 per kg. At a loaded $25/hour that's ≈31 hours of paid time per finished kilogram. Where does it go? Mostly one place: hand trimming. Industry throughput for a hand trimmer is roughly 0.45–1.4 kg (1–3 lb) of dried flower per 8-hour shift, at $15–20/hour or $100–200 per shift piece-rate^[13]. Run the division: that's ≈6–18 hours per kg for trim alone. Call it 10 — at $25/hour loaded, $250 per kg, $0.25 per gram, just for trimming. The scissors out-cost the electricity.

Notice the gap: tasks sum to ≈18 h/kg but payroll says ≈31. The missing 13 hours are real work that never touches a bud, mothers and veg care, meetings, cleaning, records, sick days, and plain idle time between tasks. That gap is utilisation, and it's why headcount models built from task lists always come in under the real payroll. Budget from payroll; use task minutes to find what to fix.

• Measure before you buy. A trim machine at 9–18 kg/h (20–40 lb/h)^[13] looks unanswerable next to 0.9 kg/shift (2 lb/shift), but weigh the grade impact on your product and your buyer before the capex (Sections 11 and 14).
• Smooth the spikes. Harvest weeks need 3× the hands of week 3 of flower. Staggered rooms (Section 10) turn a hiring problem into a scheduling one.
• Price the founder. If your own hours enter at $0, every bad room you'll ever build will look profitable on paper.

Annual crop cycles

Everything you produce in a year is grams-per-cycle × cycles-per-year. The industry obsesses over the first term and lets the second one rot. Turn time, the days between harvesting one crop and flipping the next, multiplies everything.

The arithmetic is brutal because it's a division that compounds. At a 7-day turn the room runs 5.8 cycles and ships 261,000 g. Let the turn drift to 21 days, a slow clean here, a late clone batch there, a week waiting on a parts order, and it's 4.7 cycles and 213,300 g. Two extra weeks per turn costs 47,700 g a year: at a $2.20 blended price, over $100,000 of revenue, for zero saved cost. No nutrient program on earth moves the needle like that.

Backer's meta-analysis found longer flowering raised yield per m²^[5], and that's exactly the trade to price properly: an extra week of flower must earn more grams than the same week would earn as a fresh cycle. At 45,000 g per cycle, a 63-day flip earns ≈714 g per calendar day; a 70-day flip has to yield ≈50,000 g per cycle, 11% more, just to tie. Run that division before you extend ripening, not after.

1. 1
   Define flip-to-flip

   Flower-in to flower-in, in days, on the whiteboard. If it isn't measured it will drift, nobody notices a turn stretching one day per cycle.
2. 2
   Pre-stage the turn

   Repair list closed, room consumables staged, clean crew booked, before harvest morning. The turn is a pit stop, not a project.
3. 3
   Keep veg ahead of flower

   The most common turn-killer is clones that aren't ready. Veg capacity must run one full flip ahead of the flower room's calendar.
4. 4
   Stagger if you can

   Four small rooms flipping in rotation give the same annual cycles as one big room, but level the trim labour and turn a crop failure into a 25% event instead of 100%.

Quality premiums and yield volume

Cost per gram is half the story; the cheque depends on the price per gram, and price is tiered. US spot-market averages in early 2024 ran ≈$1,378/lb for indoor flower (≈$3.04/g), $725/lb greenhouse (≈$1.60/g) and $418/lb outdoor (≈$0.92/g)^[14] — a 3× spread on production method alone, before grade tiers within each method split further into A-flower, B/smalls and trim, each with its own price.

This is why blended price, not headline price, belongs in the model, and why chasing top-shelf changes the whole equation rather than one line of it. Compare two strategies for the fictional room, which sits near break-even at a $1.90 blended price:

Cost-per-gram sensitivity

Before spending a dollar to improve the room, ask the model which dial is worth touching. The method: take the fictional baseline ($1.90/g), move one input at a time across a plausible swing, hold everything else, and recompute. Plot the results widest-first and you get a tornado:

Read the order, because it's the whole strategy. A 20% yield move swings cost per gram four times further than halving-or-adding-half to the entire nutrient budget. The two biggest bars, yield and labour, are grower skill and process design. The bars vendors talk about most (power price, capex, bottles) are the small ones. And note what the swing sizes hide: a 20% yield swing is one bad pest cycle or one steering mistake, while a 50% power-price swing requires renegotiating with a utility. The big bars are also the easy ones to move, in both directions.

Break-even analysis

Break-even is the yield, price or cycle count where profit crosses zero, and knowing where it sits turns vague anxiety into specific targets. Three divisions, same fictional room:

• Break-even price at 450 g/m² and 5.8 cycles: $495,000 ÷ 261,000 g = $1.90/g blended. Below that cheque, every gram ships at a loss.
• Break-even yield at a $2.20 blended price: $495,000 ÷ $2.20 = 225,000 g → ÷ (100 m² × 5.8) = ≈388 g/m² per cycle. That's the floor under a bad run.
• Break-even cycles at $2.20 and 450 g/m²: 225,000 ÷ 45,000 = 5.0 cycles → flip-to-flip must stay under 365 ÷ 5.0 = 73 days. The calendar has a red line.

Two habits make break-even thinking useful rather than depressing. First, compute it per constraint (a price floor, a yield floor, a calendar ceiling) so every team member owns a number they can actually influence. Second, recompute after every change: costs creep, prices sag, and last year's comfortable margin can become this year's break-even without a single dramatic event. Falling wholesale prices have been the norm in maturing markets^[14], build the model expecting the band to move down, not up.

Common unit-economics mistakes

Every one of these is survivable once and fatal as a habit. All of them are denominators or missing lines. None of them is agronomy.

Troubleshooting

Symptoms first, causes second, same as diagnosing a sick plant, except the sensor is the bank statement and the lag is a full quarter.

Unit-economics control variables

What to actually do this week, in order:

1. Build your own cost stack from the last twelve months of real spending, all eight lines, founder hours priced at market rate.
2. Divide by grams sold in the same twelve months. Write the $/g answer somewhere prominent and slightly uncomfortable.
3. Put flip-to-flip days on the whiteboard and start the streak.
4. Time one full trim shift and one full harvest day, your two biggest labour blocks, before considering any machine.
5. Run the tornado with your numbers and re-rank your wishlist by span, not by excitement.
6. Recompute quarterly. Costs creep, prices sag, and the model is only honest while it's fresh.

And keep the humility the benchmarks force on you: the published record spans 0.31–1.97 g/W^[5] and hundreds of g/m² between honest studies^[6][7]. Nobody else's number, including the fictional room's $1.90 — is your number. The method is portable; the answers never are.