We assessed indoor farming. The electricity bill decides

Published: 26 August 2026 · 19 min readIndustry assessmentsAI-native operations

We assessed indoor farming. The electricity bill decides

A modular grow-box came through our pipeline: a half-metre cube with a light inside, sold to restaurants and canteens as a monthly service rather than a machine. Hamnett Enterprises GmbH ran the full feasibility on it. One number per crop decided the case, the electricity price above which indoor growing loses money on every kilo. For lettuce that number is €0.053 per kilowatt hour. German industry paid an average of 17.2 cents in 2026 (BDEW, August 2026), and German households 37.0 cents.

Nothing here is legal or tax advice. Your Steuerberater and Anwalt decide before anything binds.

The decision

We parked it, with a capital ask of zero and a written trigger. The crop economics are real, and they are real today on German grid power, for six classes of crop. What the plan lacked was a reason we would win. A US incumbent already occupies the exact position the venture staked out and publishes its terms: $16,750 for the unit plus $375 a month on a 36-month contract per unit, with Aramark, Sodexo, Dartmouth College and American Airlines named as customers on its own site. Our model put the build at €13M to €15M of patient equity, operating break-even in the sixth year, and capital back around year seven or eight. We reopen the file when three conditions hold at once: a hardware and agronomy lead secured, at least one institutional customer committed at the modelled monthly price, and a power contract at or below €0.05 per kilowatt hour. Until then the only spend we authorise is a small probe into licensing the control software to somebody else's box.

The one law that decides everything

Outdoors the sun lights the crop for free and the rain waters it. Indoors you buy the light. That single swap is the whole industry, and reshaping the box does not repeal it.

Growing one kilo of leafy greens indoors takes roughly 15 kilowatt hours of electricity in our model, in a published range of 10 to 18 for optimised farms. At the €0.25 per kilowatt hour our model uses as a German blended commercial price, that is €3.75 of electricity against a wholesale lettuce price of about €2.20 a kilo. The electricity alone is 167% of what the crop sells for, before the building, the cooling, the water, the labour or the box. At the BDEW figure for small and medium industrial supply contracts, 17.2 cents, it is still 117%. At the household price of 37.0 cents it is 252%. There is no power tariff available in Germany at which commodity lettuce works indoors.

Run the same arithmetic on a staple grain and the gap stops being arguable. Our model shows about €6.80 of electricity to grow €0.30 of maize. Indoor grain loses in every era, at every power price, in every country. Anyone pitching indoor boxes as a calorie-security fix for wheat or rice is selling something the physics will not deliver.

The building cost piles on top. Published facility capex for controlled-environment agriculture runs €1,075 to €10,750 per square metre, of which grow-lights are 30% to 40% and the building plus air conditioning another 25% to 35%. Cooling is the line most often under-budgeted, because every watt of light you buy arrives back as heat you then pay to remove.

Break-even electricity by crop

The make-or-break number is the break-even electricity price: the crop's wholesale value per kilo, less every non-energy cost per kilo, divided by the kilowatt hours it takes to grow. Above that price, indoor loses. Below it, indoor wins.

Crop kWh per kg Wholesale €/kg Non-energy €/kg Break-even €/kWh At German grid €0.25 At cheap solar €0.05
Leafy greens (lettuce) 15 2.20 1.40 0.053 Loses, cost €5.15 Wins, cost €2.15
Fruiting veg (tomato, strawberry) 45 9.00 4.50 0.100 Loses, cost €15.75 Wins, cost €6.75
Culinary herbs (basil) 18 27.50 6.00 1.194 Wins, cost €10.50 Wins, cost €6.90
Cannabis, legal jurisdiction 350 3,000.00 400.00 7.429 Wins, cost €487.50 Wins, cost €417.50

As of August 2026. Break-even prices, energy intensities and cost lines are Hamnett Enterprises model, 2026. The German grid reference of €0.25/kWh is a blended commercial planning input and sits between the BDEW figures of 17.2 ct/kWh for small and medium industry and 37.0 ct/kWh for households (BDEW Strompreisanalyse, 21 August 2026). The cannabis row is an illustration of how value per kilo governs the arithmetic and is not a business plan. §9 KCanG permits three plants per adult at home in Germany and forbids passing the crop to third parties, and commercial cultivation is separately licensed.

The spread across that column is the entire industry in four rows. Lettuce tolerates 5.3 cents. German industry pays 17.2. Basil tolerates €1.19, roughly seven times the industrial price, which is why a herb crop absorbs an expensive grid without noticing. Fruiting vegetables sit at 10 cents, above cheap solar and below the grid, so they need a self-generated or contracted renewable supply to work at all. Cannabis tolerates €7.43, about forty-three times the industrial price, so its binding constraint is the licence rather than the meter.

The rule that falls out: value per kilo is the gate. A crop selling at €1 to €2 a kilo cannot carry €2 to €5 of electricity. A crop selling at €27 or €3,000 does not notice it.

When each crop crosses

Indoor growing competes against a moving benchmark. Its own cost falls as light-emitting diodes get more efficient, and our model takes about 35% fewer kilowatt hours per kilo by 2045. The field benchmark it must beat moves the other way on the climate-stress path, with fresh-vegetable prices rising about 30% by 2035 and about 70% by 2045 in real terms. Where the two lines meet is the crossover year.

Crop class Works on grid power now Grid crossover, climate-stress path Works on cheap solar now Timing verdict
Culinary and medicinal herbs Yes now Yes Now
Microgreens and propagation Yes now Yes Now
Premium and branded greens Yes now Yes Now, while the premium holds
Gourmet mushrooms Yes now n/a, almost no grow light Now
Saffron Yes now Yes Now
Cannabis, legal jurisdiction Yes now Yes Now, legally walled
Leafy greens (lettuce) No around 2045 Yes, thinly 10 to 20 years on grid, now on solar
Fruiting vegetables No around 2045 Yes 10 to 20 years on grid, now on solar
Root crops and tubers No beyond 2045 No Not within 20 years
Dwarf tree fruit No beyond 2045 Marginal Not within 20 years
Vining crops (pumpkin, melon) No beyond 2045 No Not within 20 years
Staple grains (maize, wheat, rice) No never never Never

As of August 2026, Hamnett Enterprises model, 2026. The forward years are an explicit scenario built on the IPCC climate-stress path for food prices and a published trajectory for light-emitting-diode efficacy, not a forecast. Treat the ordering as the finding and the exact years as illustrative.

The answer to "now or in twenty years" is both, and the split matters more than the timing. Six crop classes clear today even on an expensive grid, and they are enough to run a business on. Two more, lettuce and fruiting vegetables, cross around 2045 on the stress path or immediately on cheap self-generated power. Three never clear inside two decades, and the last never clears at all. The addressable crop set roughly doubles toward 2045, which makes patience a rational posture and makes the crop list, rather than the box design, the thing that decides a venture.

The awkward part of that timing work: the six crops that clear now are precisely the niches the US incumbent already serves. Better arithmetic on the crops did nothing for the question of who wins the customer.

Quality beats yield

The same crop can be grown three ways, and they are different businesses. Our model ran culinary basil at the German grid reference across four configurations.

Optimise for Yield index kWh per kg Total cost €/kg Achievable price €/kg Net contribution €/kg
Quality, taste, nutrient density 0.70 26 15.50 55.00 (chef, branded) +39.50
Balanced, standard premium 1.00 18 10.50 27.50 +17.00
Yield and throughput 1.30 14 9.50 14.00 (commodity) +4.50
Lowest cost 1.20 13 7.75 12.00 (spot) +4.25

As of August 2026, Hamnett Enterprises model, 2026, at a grid price of €0.25/kWh. Prices are planning values from published crop and channel bands, not quoted offers.

Growing for taste produces about 46% less per cube and burns about 44% more electricity per kilo, and it nets €39.50 a kilo against €4.50 for the yield-maximised configuration. The price lift, roughly double the standard premium and about four times commodity, swamps the cost lift. Every instinct that comes from agriculture says maximise yield per square metre. Indoors, with a premium channel to sell into, that instinct costs €35 a kilo.

This is where a Frankfurt or Rhein-Main venture would actually sit. Selling into restaurant kitchens and branded fresh retail in a dense, high-income catchment is the configuration the arithmetic favours. Selling by the kilo against Spanish and Dutch field production is the configuration that bankrupted the sector.

Mushrooms are the quiet standout

Mushrooms need about 3 kilowatt hours a kilo against 15 for leafy greens and 45 for fruiting vegetables, because they do not photosynthesise and need almost no grow light. The dominant cost term simply disappears. At a modest €6.00 a kilo wholesale, our model puts contribution at €2.05 a kilo on grid power and €2.65 on cheap solar, with published gross margins above 50%. That is a positive number at today's German electricity price with no premium channel, no cheap power contract and no climate scenario required.

The catch is a product one. Mushrooms need a dark, humid substrate-block regime rather than a lit hydroponic tray, so they are a separate configuration of the same box rather than one more crop on the same recipe. A venture that treats them as an afterthought will build the wrong hardware. A venture that treats them as an anchor gets the only crop in the taxonomy whose economics do not depend on the electricity price at all.

Water is the second lever

Indoor growing recirculates. Our model puts field lettuce at about 200 litres a kilo against about 12 litres indoors, a 94% saving, with thirstier crops nearer 96%. That saving has a money value, and it converts into electricity headroom: for lettuce, every €1 per cubic metre of water price buys about €0.013 per kilowatt hour of extra tolerance.

Water price €/m³ Water cost saved, €/kg lettuce Extra electricity headroom €/kWh
1 0.188 +0.0125
2 0.376 +0.0251
5 0.940 +0.0627
10 1.880 +0.1253

As of August 2026, Hamnett Enterprises model, 2026, on a 94% recirculation saving for leafy greens.

At ordinary German water prices near €2 a cubic metre the saving buys 2.5 cents of headroom, which does not rescue lettuce from a 17-cent grid when it needs 5.3 cents. The lever only bites at €5 to €10 a cubic metre, the scarcity prices found in the Gulf, parts of North Africa and rationed regions, where the water saving alone can carry the case. In Hessen, water is not the argument.

The extinction event

The years 2023 to 2025 were the sector's shakeout, and the failures were funded ones. Two dated tallies of 2025 count fourteen indoor-farming and controlled-environment bankruptcies in that year, with combined historical funding across the failed companies above $1.37bn. Plenty filed for Chapter 11 protection on 24 March 2025 after raising close to $1bn from investors including SoftBank, Walmart and Bezos Expeditions, and emerged from restructuring, smaller, on 29 May 2025. Bowery shut with about $700m raised. AeroFarms filed in 2023 with more than $300m raised. AppHarvest filed with about $700m raised.

The consumer end went the same way. Scotts Miracle-Gro wound down AeroGarden, the twenty-year category leader in countertop growing, during its 2024 financial year, taking about $29m of excess and obsolete inventory write-offs, reported in its results release of 6 November 2024. A slimmer relaunch under new ownership followed in 2025, which says the brand had value and the cost structure did not.

What survived is instructive. Babylon Micro-Farms, the US incumbent in institutional managed service, sells a wall unit at $16,750 with a $375 monthly subscription on a 36-month contract per unit, covering onboarding, an account manager, auto-shipped seeds and nutrients, and a guided-growing app. It names Aramark, Sodexo, Neiman Marcus, Dartmouth College, American Airlines and Sunrise Senior Living. It reaches its customers through the catering contractors rather than selling boxes one at a time. The lesson from the graveyard is that recurring service revenue survives thin hardware margins and hardware margin alone does not.

What an AI-native operator changes and what it cannot

Start with the market size, because most of the published numbers are unusable. You will see "controlled-environment agriculture: $92bn to $122bn in 2025" and "vertical farming: $6bn to $8bn growing to $33bn to $40bn." Those count the value of the food sold, or the entire greenhouse universe. They are not a market a new equipment vendor can win. Strip out the food and keep the equipment plus control software, roughly 25% to 40% of the vertical-farming figure, then take the modular-format share of that, roughly 10% to 20%. What remains is a modular controlled-environment equipment market of about €0.15bn to €0.6bn a year globally, thirty to fifty times smaller than the headline. Every business plan we have seen in this category uses the headline.

Inside that number, the value sits in two places and neither is the cube. The first is per-crop control software: the recipe that decides light spectrum, photoperiod, nutrient dose and harvest window for each crop and each customer, and improves from fleet data. The second is fleet management: running several hundred scattered units remotely with one small team, catching a failing pump before the customer sees a dead tray. This is where an AI-native operator has real leverage, because writing and tuning several hundred crop recipes, and running the anomaly detection across a distributed fleet, is exactly the work that compresses. Our model puts minimum efficient scale at 400 to 600 installed units, the point where remote-management overhead absorbs into per-unit contribution.

What it cannot change is the photon count. A plant needs a given quantity of light to build a given quantity of tissue, and no control layer, no model, and no amount of software talent moves that. The box is a single mass-produced part with a bill of materials and a certification burden. Software is where we are strong and boxes are where we are weakest, which is why the licensing probe, and not the build, is what we authorised.

Path to yes

The configuration that works is specific, and it is worth naming precisely because "vertical farming" as a category is a bad bet while this configuration is a defensible one.

For anyone weighing an indoor-growing venture:

  • Restrict the crop list to the six classes that clear on grid power today. Herbs, microgreens and propagation, premium greens, gourmet mushrooms, saffron, and cannabis where it is licensed. Any plan containing commodity lettuce or fruiting vegetables on grid power is a plan to lose €2.95 to €6.75 a kilo.
  • Sell a monthly service, not a machine. Recurring revenue has to cross about 60% of the total, because a hardware business in this sector trades at a fraction of a service business and cannot absorb a bad quarter.
  • Secure power at or below €0.05 per kilowatt hour, which in practice means 70% or more self-generated renewable supply. Global utility-scale solar reached a weighted-average levelised cost of $0.043 per kilowatt hour in 2024 (IRENA), so the number is achievable with an asset, not with a tariff.
  • Size the capital to the working capital, not the loss. Our model reached operating break-even in the sixth year on an operating loss of about €9M, and the combined cash trough was €13M to €15M, because inventory and leased hardware tie up roughly €11M ahead of the profit and loss. Both are recoverable. Both must be funded.
  • We build feasibility studies of this shape as a decision dossier, market sizing, crop or unit economics, the option space, the path to yes and the reversal triggers, from €4,900 net: Business strategy. The financial model that sits underneath it, with the driver sheet the founder can change, starts at €3,900: Financial modelling.

For a restaurant, canteen or facility manager in Rhein-Main being sold one of these units: ask for the crop list and the kilowatt hours per kilo before the sustainability slide. A unit growing herbs and microgreens for your own kitchen pencils. The same unit growing salad to replace what a wholesaler delivers does not.

Where we could be wrong

  • Energy intensity is the assumption everything rests on, and the published range is wide enough to break the conclusion. One 2026 industry comparison puts a vertical farm at 150 to 350 kilowatt hours per kilo of lettuce against 1 to 5 for open field, ten to twenty times our 15. At merely three times our figure, basil's break-even falls from €1.19 to about €0.40 per kilowatt hour, still comfortably above the grid, while lettuce's falls from €0.053 to about €0.018 and the 2045 crossover disappears entirely. The high-value crops survive that revision. The timing story does not.
  • Our €0.25 per kilowatt hour grid input is above what a commercial site now pays. BDEW puts small and medium industrial supply at 17.2 cents in 2026, down 1.6 cents on the year. At 17.2 cents, lettuce electricity is 117% of wholesale rather than 167%, still underwater, and fruiting vegetables still lose about €3.24 a kilo. If a site secured a 10-cent contract, fruiting vegetables would cross now and the workable crop set would widen from six classes to seven.
  • The entire quality case rests on basil holding €27.50 a kilo wholesale and €55 into a chef channel. If those compress to the €14 commodity price, contribution falls from €17.00 to €3.50 a kilo and the monthly lease stops covering the hardware. Premium erosion, not electricity, is the failure mode we would expect to hit first.
  • We assumed a lease of €650 a unit a month. The incumbent publishes $375 a month against a $16,750 unit price, a different split between capital and service. If a European entrant has to match that split rather than the all-in lease, the revenue line that our model is most sensitive to moves by about a quarter, which is roughly €7.8M on the sixth-year revenue.

FAQ

Why do vertical farms go bankrupt?

Because the electricity to grow a cheap crop costs more than the crop sells for. Growing a kilo of lettuce indoors takes about 15 kilowatt hours, which is €2.58 at the 2026 German industrial price of 17.2 cents, against a wholesale price near €2.20. Every head is sold at a loss before rent, labour or the building. Two dated tallies count fourteen indoor-farming bankruptcies in 2025 alone, with more than $1.37bn of historical funding behind the failed companies. The pattern was building at €100m scale on a promise of profitability at €10m scale, with venture capital expecting a five to seven year return on what is a fifteen to twenty year infrastructure payback.

Which crops make money indoors?

Six classes clear on German grid power today: culinary and medicinal herbs, microgreens and propagation material, premium branded greens, gourmet mushrooms, saffron, and cannabis in jurisdictions that license it. All of them are worth enough per kilo that electricity is a small share of their value. Mushrooms clear for a different reason, needing about 3 kilowatt hours a kilo instead of 15, because they need almost no light. Lettuce, tomatoes and strawberries need cheap self-generated power or a wait to about 2045. Root crops, tubers, dwarf tree fruit and vining crops do not clear within twenty years, and staple grains never do.

Does cheap solar change the answer?

For two crop classes, yes, and it changes them completely. At €0.05 per kilowatt hour, lettuce moves from a €2.95 loss a kilo to roughly break-even and fruiting vegetables from a €6.75 loss to a €2.25 gain. Global utility-scale solar reached $0.043 per kilowatt hour on a weighted average in 2024 (IRENA), so that price exists where you own the generation. It does not exist as a German grid tariff, which is why practitioners treat grid dependence as fatal and 70% or more self-generated renewable supply as the working standard. Cheap power does nothing for staple grains at any price.

What is a managed-service model in agriculture?

The customer leases equipment and pays a fixed monthly fee, and the vendor owns the hardware, installs it, ships the seeds and nutrients, and runs the growing remotely through software. The customer harvests. The US incumbent in this lane publishes $16,750 for the unit plus $375 a month over a 36-month contract, covering onboarding, an account manager, auto-shipped supplies and a guided-growing app. The model matters because it converts thin one-off hardware margin into recurring gross profit, which is what a buyer of the business eventually pays a multiple for. It also consumes cash, since the vendor funds the hardware sitting in every customer's building.

Further reading

Sources

Nothing here is legal or tax advice. Your Steuerberater and Anwalt decide before anything binds.

← All posts