The vertical farm was never the answer. The boiler is.

An industry compilation counted fourteen indoor-farming bankruptcies and closures worldwide in 2025. The question was never whether to grow food indoors. It is what, and next to what.

We wrote recently that American farming is shrinking while still needing about 70,800 people a year, and that more than half of farmers are 55 or older.¹ The obvious next question is what to do about it. So this piece is the other half: given a problem, what would actually fix it.

The first move is the one most solution pieces skip.

This piece is about the second problem. The first one needs land trusts, lease-to-own, incubator farms and estate structures that let a young farmer take over the cash flow years before the deed — and that is a subject of its own.

Start with the graveyard

Any honest answer here has to begin with the fact that a great many people have already tried the glamorous version, raising well over a billion and a half dollars combined before failing.

The failure mode is the same every time, and it is not bad luck. A vertical farm replaces two free inputs — sunlight and rain — with two expensive ones: electricity and engineering. That trade only works if the crop is worth enough per kilogram to pay the difference. Most food is not.

What Canada actually grows in glass, and how much it yields

Canada has one of the largest greenhouse vegetable sectors in the world, so we do not have to guess. Here is what a square metre produced in 2024, from national production and area figures.⁴

Yields are our arithmetic: production divided by harvested area, national averages across operations of every level of technology.⁴

So what would a cold city of 250,000 actually need?

Canadians had 61.8 kg of fresh vegetables available per person in 2025, excluding potatoes — the seventh consecutive annual decline.⁵ For a city of 250,000 that is about 15,450 tonnes a year.

In the north, heat is the whole game

Lighting gets the attention. In a cold climate, heating is what decides whether the building survives its second winter. Three sources change the arithmetic, and which one is right depends entirely on where you are standing.

• Biomass, where there is a sawmill. A common boiler-sizing rule: a biomass boiler sized to 50–60% of daily peak demand, backed by a buffer tank or a fossil-fuel peaking boiler for the coldest hours, covers 80–90% of annual heat need — because peak-hour demand is a small share of the year's total hours.⁶ Northern Ontario has forestry residue and the mills that make it. That is a local fuel with a local price.
• Geothermal, where the alternative is flown in. Nothing has been built yet, but a modelled design study for Resolute Bay, Nunavut, estimated geothermally-heated greenhouse vegetables could be produced at up to half the delivered cost of the same vegetables flown in — the study calls itself a proof-of-concept, and the modelled cost excludes labour, water, soil and seeds. Heat was the problem the design solved; it still needs diesel-fuelled lighting.⁷ The further from a highway, the better this looks — because the competitor is an aircraft.
• Waste heat, where something hot already exists. A data centre rejects nearly all the electricity it consumes as low-grade heat, and liquid-cooled AI hardware specifically returns it at 55–65°C — hot enough to be useful (air-cooled halls reject heat at a lower 25–40°C).⁸ Researchers have modelled exactly this pairing.⁹ A greenhouse is a better customer for that heat than a district network, because it wants heat in winter and it can use the carbon dioxide from a flue too.
• Gas cogeneration, the option this piece almost skipped. Biomass and geothermal both solve heat but leave winter lighting as a separate, unsolved bill — and lighting is exactly the cost this piece blames for killing vertical farms. Combined heat and power (a gas engine on site) is already standard practice in Leamington-style greenhouses: one flame supplies electricity for lights, heat for the building, and carbon dioxide for the crop, together. Any cold city with a gas line is a candidate for this before geothermal, which needs a purpose-built well.

Two things not to do, and one nobody mentions

Do not put root vegetables under lights

Carrots, beets and potatoes are heavy, cheap and slow. They are the worst possible crop to grow with purchased photons. Grow them in a field in summer and store them. Modern controlled storage beats winter lighting so decisively that it is not a close call — and it is unglamorous enough that nobody pitches it at a conference.

Do not build the tall version first

A greenhouse uses free sunlight and pays for supplemental light and heat. A vertical farm pays for all of its light. The companies above proved what happens when you take the second path for crops that cannot carry it.

Sprouts need no light at all, and mushrooms need none either

Sprouts — broccoli, alfalfa, mung — are grown in the dark, in water, in trays, in about a week. There is no lighting bill because there are no lights. Microgreens need a little light and about two weeks. Canada already grows microgreens and shoots commercially, though the sector is small and sprout production has actually shrunk since 2020, not grown.⁴ Health Canada advises people who are pregnant, young, elderly or immunocompromised to avoid raw sprouts specifically — a real constraint on how big this slice can get. Mushrooms are the larger, already-proven version of the same idea: grown year-round in closed, dark, climate-controlled buildings, with no lighting thesis required at all, and Canada already produces roughly 500 times as much mushroom tonnage as sprout tonnage.⁴ A rented industrial unit and a few thousand dollars starts either one.

The ladder

Short-term steps that lead into the long-horizon structure, rather than requiring it:

What we cannot see from here

• The yields in this piece are ANNUAL averages, dominated by summer light. Canada does not publish monthly greenhouse output, so we cannot show what a glasshouse at Sudbury's latitude actually produces in December and January specifically. Unlit fruiting crops produce little in deep winter; year-round production needs supplemental lighting — the same purchased-photon cost this piece blames for killing vertical farms, arriving in exactly the months a cold-city project targets. Solving heat does not by itself solve this.
• For a road-connected city, the real competitor may be a truck, not a plane. Canada's greenhouse sector exports the majority of what it grows and also imports greenhouse vegetables, mostly from Mexico, into the same market. A northern greenhouse competing for a city's grocery shelf is competing with produce that already travels efficiently by road, not only with the most expensive flown-in case. This does not apply the same way to communities that genuinely have no road, like Resolute Bay.
• Local electricity and fuel prices decide everything, and they are not national numbers. The same building can be viable in one town and hopeless ninety minutes away.
• Our yields are national averages across every level of technology. A modern Leamington glasshouse beats them; a hoop house does not reach them.
• We have not costed labour, which is the second-largest line after energy and the hardest to find in a small northern town.
• Availability is not consumption. The 61.8 kg figure is food available per person, before waste in homes and stores. Actual eating is lower, so our area estimate is, if anything, generous.

Explain it to a friend

Sources

1. Our own piece, Growth is 3.4% of the jobs, from US Bureau of Labor Statistics Employment Projections table 1.10 and CPS table 11b.
2. TechCrunch, Vertical farming company Plenty files for bankruptcy after raising nearly $1B, 24 March 2025. — techcrunch.com
3. Reporting on Bowery Farming's liquidation and the 2025 bankruptcy count. Figures for combined funding across failed indoor-farming companies are trade-press estimates, not audited totals — treat them as an order of magnitude. — fertilizerdaily.com
4. Agriculture and Agri-Food Canada, Statistical Overview of the Canadian Greenhouse Vegetable Industry, 2024 edition (published October 2025), tables 1.3 and 1.5, sourced in turn from Statistics Canada table 32-10-0456-01. Yields are our own division of production by harvested area. — agriculture.canada.ca (PDF)
5. Statistics Canada, The Daily: Food availability, 2025, released 28 May 2026: fresh vegetables excluding potatoes, 61.8 kg per person, down 3.3% from 2024. — statcan.gc.ca
6. Review of greenhouse heating in cold regions: biomass boilers meeting 50–60% of daily peak and 80–90% of annual heat demand. — Greenhouse Canada
7. Geothermal energy for food production in Canada's remote northern communities, Energies 12(21), 2019, including the Resolute Bay design case. — doi.org/10.3390/en12214058
8. Our own piece, The AI build-out makes two things nobody is bidding for, on data-centre waste heat and its price.
9. École de technologie supérieure (ÉTS Montréal), modelling of greenhouse heating from data-centre server waste heat. — etsmtl.ca
10. Statistics Canada, Table 32-10-0054-01, Food available in Canada: lettuce 7.24 kg and spinach 0.97 kg per person, 2025. — statcan.gc.ca. Kale, chard, arugula and other leafy greens are not broken out in this table, so lettuce plus spinach is a floor on the true leafy-greens share, not the whole of it.

Figures retrieved 24 September 2026. Statistics Canada material is used under its open licence; this is not an endorsement by Statistics Canada. Corrections appear in our corrections log.