Leave the Lights On: How Vertical Farms Cut Electricity Costs Without New Hardware

Last updated: August 7, 2026

A team in China grew butter leaf lettuce under a light schedule that reads like a typo: 120 hours of light, then 60 hours of darkness, over and over. Almost every vertical farm in the world runs sixteen hours on and eight hours off, and against that normal schedule the plants on the strange one ended up between 34% and 83% heavier. The lights were on for longer, the lettuce came out bigger, and the electricity spent per kilo went down.

I read that three times before I believed it.

A vertical farm’s electricity bill comes from two places: the lamps that stand in for sunlight, and the cooling system that pulls their heat back out. Together they take the biggest slice of what it costs to stay open. Cutting that bill without buying anything means changing when the lamps run, how low they hang, and what grows under them.

I have put these five in the order of how much you actually have to change to use them, starting with the one that costs nothing but a different number in the software, and ending with the one that decides where you put the building. The order matters, because the cheapest ideas turn out to be the ones nobody in the industry is working on, and the size of that gap is the most interesting thing I found.

Table of contents

1. Plants cannot tell the time
2. The cheapest kilowatt is the one you buy at 3am
3. You are paying for the hypotenuse
4. Some plants are simply better at this
5. Someone else is throwing away exactly what you need
6. The box nobody has opened
7. FAQ

Plants cannot tell the time

All of this rests on one fact about plants, and the fact is that a plant does not keep score the way a light meter does. What it reacts to is the total amount of light that lands on it across a whole day, and within pretty wide limits it does not much care whether that light arrives as a short bright blast or a long soft one. Growers have a name for the daily total, the daily light integral, and once you know the plant is counting that and nothing else, the whole bill is suddenly up for negotiation.

Jason Lanoue and his colleagues at Agriculture and Agri-Food Canada put the idea to work on microgreens and published what happened in Frontiers in Plant Science in 2022. Running the lamps for more than eighteen hours a day but turned down delivered the same daily total, the crop was as good or better, and because dimmer lamps can be smaller and fewer, the kit costs less to buy as well as less to run. Their paper says where the money comes from in plain terms: the long gentle schedule moves the electricity out of the expensive daytime hours and into the cheap ones.

Then there is the result I opened with. Xiao-li Chen and co-authors, writing in Scientific Reports, pushed past the 24-hour day completely and ran lettuce on cycles of 24 hours light and 12 dark, then 48 and 24, then 96 and 48, and finally 120 and 60. Every one of those odd cycles beat the normal one, and the longer the cycle got the bigger the win, which is not what you would expect from something that grew up on a spinning planet.

What I like about this whole family of ideas is that none of it needs a purchase order. A farm already running LEDs on a timer can try a longer cycle next week, on one rack, and find out what its own crop does in its own building.

The cheapest kilowatt is the one you buy at 3am

If the plant genuinely cannot tell the time, then the grower gets to decide what time it is, and that decision is worth real money anywhere electricity is priced by the hour. Dafni Despoina Avgoustaki and George Xydis ran the numbers in Biosystems Engineering in 2021 and made the case for moving a vertical farm’s lighting into the cheaper hours. On their model the farm stops being a building that takes whatever power it wants whenever it wants it, and starts being something the grid can actually work with.

It scales up further than I expected. Vahid Arabzadeh, Panu Miettinen and Titta Kotilainen took the idea to a city farm sitting on a grid with a lot of wind power, where prices swing hard with the weather (Applied Energy, 2023). Akshay Ajagekar, Benjamin Decardi-Nelson and Fengqi You went further still in the same journal, teaching a machine-learning system to run a whole network of greenhouses together, working out which building should pull power at which moment.

This is also the one place where I found a company instead of a paper. Sollum Technologies, based in Montreal, has a live patent filed in March 2023 for handling what the power industry calls load shedding, meaning the moments when the grid is stretched and the utility needs its biggest customers to pull back. Their system stays connected to the utility, and when a cutback is announced it reshuffles the lighting across several farms so the crops take the hit without being hurt by it. The grid gets its relief and the grower keeps the harvest, and the entire negotiation happens in software.

You are paying for the hypotenuse

Light leaves a lamp in a cone. Lift the lamp higher and the cone opens wider, so more of the light you just paid for lands on the walkway, the shelf frame and your shoes instead of on a leaf. Drop the lamp and the cone closes up. Pythagorize that lamp, and a piece of your bill comes back for free.

Fatemeh Sheibani, Mike Bourget and Robert Morrow gave this a name and a measurement in Frontiers in Plant Science in 2023, calling it close-canopy lighting. Their starting point is that the industry has spent a decade making LEDs more efficient while a serious share of that hard-won light still spills past the edges of the tray, which means everyone has been polishing the lamp and ignoring where it hangs.

The reason this is newly possible is genuinely nice. Old grow lamps were hot on the face, so putting one near a leaf would cook it, and growers kept their distance for good reason. Modern LED grow lights stay cool where the light comes out and they dim smoothly, so the safe distance shrank and nobody quite noticed that the old spacing rules had retired along with the old lamps.

Some plants are simply better at this

Theekshana Jayalath and Marc van Iersel grew mizuna and lettuce side by side under six different light levels and published the comparison in Plants in 2021. They measured two things: how wide each crop spreads its leaves, which decides how much light it catches in the first place, and how well it turns the light it catches into actual plant. The two crops came out clearly different on both, and the authors say the obvious thing out loud, which is that choosing more efficient crops is one of the tools available for bringing the cost down.

That turns picking a crop into an engineering choice. A plant that spreads wide early catches light that would otherwise hit an empty tray, and a plant that uses what it catches well turns more of it into something you can sell, so two crops with the same price on the shelf can carry very different electricity bills behind them.

The extreme version is to grow something that needs no light at all, which is exactly what mushroom farms do, and it is why the crop that grows in the dark has running costs that leafy greens can only dream about.

Someone else is throwing away exactly what you need

The last one is the biggest, because you decide it before the building goes up. Takuya Togawa, Tsuyoshi Fujita and Liang Dong worked out whether you could heat an indoor farm with waste heat piped over from a power station in the Journal of Cleaner Production, and their answer turns on how far apart the two buildings are, because warm water loses its warmth over distance and takes the whole idea with it.

The same trick works on the other thing a farm buys by the tonne. A. Thomson, G.W. Price and colleagues looked at catching the carbon dioxide that comes off composting and feeding it to the plants, which would replace bottled CO2 with a gas that a composting site is currently paying somebody to let out into the air.

The box nobody has opened

I started this research with a different question, which was whether you could grow several crops together inside a vertical farm and have them help each other, the way farmers have mixed crops in fields for centuries. So I went looking properly, through the science and through the patents, and the box turned out to be empty. A search for mixed crops grown indoors returns exactly one paper, and a search for anyone testing plant varieties for how little electricity they need returns nothing at all.

That surprised me more than the lettuce did.

It is also good news rather than bad. There are several hundred papers on how a vertical farm can use less power, barely a dozen on how it could buy that power more cleverly, and essentially none on whether the plants themselves could be arranged to do part of the job. Nobody has tried this and found that it fails, which is a completely different thing from a dead end, and it is sitting in a corner of the field where one rack and one growing cycle would be enough to run a real test.

I want to give that its own article. For now, the useful part is that the five ideas above are already written up, already available to a farm that is standing today, and none of them is waiting on anyone to invent anything.

FAQ

Lorenzo Russo makes FoodLore from Sardinia, Italy. Former pasta maker, current food tech obsessive. Currently trying to explain to his flatmate why the kitchen light needs to stay on for five days.

Sources: Chen et al. (2022), Scientific Reports 12, doi.org/10.1038/s41598-022-10681-3 · Lanoue et al. (2022), Frontiers in Plant Science 13, 983222 · Avgoustaki & Xydis (2021), Biosystems Engineering 211, doi.org/10.1016/j.biosystemseng.2021.09.006 · Arabzadeh, Miettinen & Kotilainen (2023), Applied Energy 331, 120416 · Ajagekar, Decardi-Nelson & You (2024), Applied Energy, 122349 · Sheibani, Bourget & Morrow (2023), Frontiers in Plant Science 14, 1215919 · Jayalath & van Iersel (2021), Plants 10, 704 · Togawa, Fujita & Dong (2014), Journal of Cleaner Production, doi.org/10.1016/j.jclepro.2014.06.010 · Thomson, Price et al. (2022), Journal of Cleaner Production, 130051 · Olvera-Gonzalez et al. (2021), Energies 14, 1603 · Engler & Krarti (2021), Renewable and Sustainable Energy Reviews 141, 110786 · Sollum Technologies, patent NL2034401B1