Economics and Profitability

A vertical farm's initial investment is decided by "how many years to recoup it"

“I want to start a vertical farm. About how much should I budget for the initial investment?” There is no single figure I can give as a market rate in answer to this. Even when we all say “vertical farm,” once the number of shelf tiers, the degree of automation, and who you sell to differ, the money you need shifts literally by orders of magnitude. And yet you have to put an estimate in front of your boss. So most people start by hunting for “the market rate.” But in everything I have seen on site, the initial investment is not the number you decide first. It is the number that only comes into view by working backward, once you have drawn out “how many years to recoup it.” This article is about that order.

The initial investment only has meaning paired with a payback period

“How much do you need to start a vertical farm?” Search it and you get figures spread wide apart, from tens of millions of yen to over a billion. How much should I budget, then? — and you’re left at a loss. But there is room to pause for a moment over the very way that “how much?” question is framed.

In all the launches I have watched, the difference between a cheap factory and an expensive one was not simply a difference of scale. The cheap one looks like the better deal, and yet something nags. Line up, say, a small factory you can build for tens of millions against a large one costing several hundred million, and there are patterns where “cheap = a good buy” does not hold. Comparing on the figure alone doesn’t sit right.

That nagging feeling is probably right. On the figure alone the cheaper one looks like the better deal, but the initial investment is not a number you can judge good or bad on its own. Even a small factory at tens of millions can end up with weak insulation and HVAC driving up the electricity bill, or harvest volume that never ramps up and leaves the cost per plant stuck high—and so “built it cheap but it doesn’t add up” happens after all. Conversely, if a several-hundred-million-yen factory can cut labor through automation and hold a high utilization rate, then once you include payback, that one can be the rational choice. In other words, the initial investment is a number that first takes on meaning not on its own but paired with “how many years it takes to recoup.” Cheaper is not the right answer. Over how many years is that figure designed to come back to you—only once you see that far can you judge whether it is high or low.

“Built it cheap but it doesn’t add up” is not a rare story. It shows up in the numbers, too. In field surveys of large-scale domestic protected cultivation and vertical farms, a certain share of operators have stayed in the red year after year. The older the figure, the harsher: there is even a report that as of 2017, after cumulative subsidies of roughly 50 billion yen had been poured in, 75% were still in the red (see 1). That said, this 75% is an old figure from 2017, the worst period, and does not represent the present. In the most recent FY2025 survey, overall 64% are profitable or break even—or, put the other way, the share in the red has fallen to about 36% (see 2).

Here is one caveat that, as a matter of field sense, cannot be left out. This deficit rate blurs the reality when you aggregate across facility types. Even within that same latest survey, Greenhouse and hybrid types are each over 70% profitable or breaking even, whereas PFAL (the closed-environment method that grows under LED lighting alone) is about 50% profitable or breaking even—meaning even now, nearly half are in the red (see 2). The question “what share of vertical farms are in the red?” has no answer unless you separate the types. It is PFAL that needs heavy investment and is hard to recoup, and the tension I am about to discuss—what to cut and what not to cut—mainly has this PFAL in mind. Because the aggregate is limited to large-scale domestic facilities, this is not to say “your project will land on this rate too,” but at the very least there is a clear structure in which simply making the figure cheap does not make it viable.

The figure only has meaning paired with a payback period

So how is that figure decided? In what I have seen, it is not so much that people who chose a cheap factory “didn’t think about payback,” but that they never set, at the outset, over how many years they intended to recoup it. The people who chose the expensive factory, conversely, fix the payback period first, so the figure follows. If so, is the figure something that emerges as a result?

Half of that is true, and half is the reverse. In practice, “how many years do I want to recoup it in” and “how much can I afford to spend” are decided at the same time. Even if you fix the payback period first, the moment the site’s electricity rate, the crop’s unit price, or the buyer changes, the same figure yields a different payback period. In quite a few projects, the subsidy allocation cap or the borrowing ceiling fixes the ceiling on the figure first. The figure does not simply fall out as a result; it is decided going back and forth together with the payback period, the sales channel, and the operating cost. What trips up the person who chose a cheap factory is fixing the figure first and then settling the payback period afterward with “it’ll work out somehow.” So what I want to say is not the flat assertion that “the figure is decided by working backward,” but something more mundane. Comparing figures without looking at the payback period is meaningless, and the figure only has meaning paired with the view of over how many years and up to how much you can recoup.

Even when you pin down the payback period, in the case of a vertical farm that period itself is something you have to set fairly long. By one estimate cited in a CEA (controlled-environment agriculture) review, such facilities typically take 5 to 7 years to reach profitability, and a vertical farm’s initial cost is estimated at roughly 2 to 3 times that of a Greenhouse (see 3). This is a secondary figure where the review cited an outside industry estimate, and it is overseas and of a different method, so it does not map directly onto your own project—but the relationship itself, that “the heavier the initial investment, the longer the time until it comes back,” squares with field sense as a reason to pin down the payback period first.

Where the line falls between costs you can cut and costs you cannot

When it comes time to build the budget, the judgment of “this can be cut” and “this must not be cut” always arises. Starting cheap is not in itself a bad thing. But get the way you cut wrong and it bites later. So how do you tell where that line falls?

The first thing that matters greatly is “whether it can be added later.” On the shell side—insulation, airtightness, HVAC capacity, and the basic structure of the building—if you skimp at the outset, efficiency is hard to restore even by retrofitting later, and it keeps getting carved into your monthly operating cost in the form of the electricity bill. In the sites I have seen, too, there were almost no cases where a factory that cut here managed to recover later. This directly pushes down payback speed, so it is on the protect side. Conversely, the number of cultivation shelf tiers, additional lighting, and the level of automation in packaging machines and conveyance leave room to add on in step with sales once operation is up and running. Rather than maxing out from the start, filling in once demand is visible reduces the capital you leave idle, and payback actually gets faster.

That said, this “can be added later” comes with one big condition. You can add shelf tiers, pile on lighting, or load automation later only if you secured, at the first-building stage, the headroom in power supply capacity, communications standards, ceiling height, and HVAC to absorb it. The number of shelf tiers tops out at the ceiling height, and adding lighting or automation rebounds onto power supply capacity and HVAC load. In other words, the upper limit of what can be added later is baked into the very shell side I told you to “protect.” So to put the line precisely: within the headroom of power supply capacity, communications standards, physical space (ceiling height), and HVAC that you secured in the first building, you fill in the “content”—automation, shelving, lighting—later. Put the other way, if you did not take that headroom at the outset, “can be added later” does not hold. Underestimate this and trim the first building too far, and when you go to expand you hit a wall: the power supply is short, the controls won’t connect (this ties directly to the procurement discussion further down).

Cutting the shell or HVAC bites every month as the electricity bill—this read is backed up from the cost-structure side as well. In PFAL, one quantitative study reports that the electricity bill accounts for roughly 25% of total production cost (see 4). This is not a sensitivity of “cut the shell and the electricity bill rises by this much,” but the composition ratio of the electricity bill within running costs; still, that lighting and HVAC make up a large chunk of running cost does not change. Skimp at the outset on the shell side that supports that chunk, and it tends to stay loaded onto your monthly fixed costs—that is the order I have seen play out time and again on site.

On the other hand, the view that “automation can just be added sparingly later” is one to be cautious about. In one retrofit (after-the-fact remodeling of an existing facility) simulation, investing heavily in automation held the impact on ROI to about 2% even when labor cost doubled. But with the minimum investment (a minimal-automation configuration), the same labor-cost increase drops it by about 29% (see 5). This is an estimate from a specific model that optimized a Korean remodeling case at a single size, so there is a range, but the implication runs rather in the direction that “investing heavily in automation up front is stronger against later labor-cost increases.” This study also points out that facilities are not designed to take in new technology after the fact, and that the wrong retrofit tends to become an irrecoverable loss. So putting automation into the category of things to simply cut—“start thin for now and add later”—is not something I recommend from my experience either. If you are going to add later, take the room to add (power, communications, space) first in the first building—here too the order is the same.

Work back from the sales channel to the equipment cost

We have looked at “cut or don’t cut,” but there is something that should be placed ahead of it. The point was that if the buyer changes, the payback period moves too. How much you can ultimately spend on equipment cannot be worked back until what you make and who you sell it to is settled. Put up an impressive box with no sales channel in sight, and it becomes a problem before “cut or don’t cut” even enters into it.

The starting point that governs payback speed is the sales channel. What you make, who you sell it to, and at what price—that is, the farm-gate price and the volume a buyer will absorb—sets the ceiling on monthly sales. Only once that ceiling is set does “over how many years and up to how much you can recoup” come out of the remainder after subtracting operating cost, and the equipment’s upper limit is set to fit within that payback period. The equipment cost is a number worked back from the sales channel; in order, the sales channel comes first and the figure comes after. Production density and scale are the same; they cannot be decided independently. How many shelf tiers to stack against a contracted volume, how many square meters to build, is decided to match the volume and price the buyer takes. Build only a box with no sales channel in sight, and because there is no outlet to match the capacity, fine equipment sits idle while only the fixed costs go out every month. This is the state, ahead of “cut or don’t cut,” where the very calculation of payback does not hold together. In order: first lock in one sure sales channel, even a thin one, set density and scale to match that volume and price, and last, confirm whether the equipment cost fits within the payback period—building in this direction was, in all I have seen on site, the most failure-resistant way to put it together.

That the sales channel governs payback is pointed out on the research side too. In a study dealing with domestic vertical farm vegetables, the difficulty of adjusting supply and demand (because, like an industrial product, it is produced to plan, it is hard to align supply with demand) is cited as a factor pushing down profitability, and securing flexible sales destinations is raised as an effective lever (see 6). The sales channel is the “securing of an outlet” and, at the same time, is backed up as an element that bears on profitability.

How much the selling price bears on payback shows up starkly in the break-even sensitivity. In one model estimate, the minimum scale at which PFAL lettuce becomes profitable was put, depending on conditions, at on the order of tens of square meters (38 m²). But a mere 20% drop in the selling price is estimated to send that break-even scale leaping all at once to over a thousand square meters (about 1,700 m²) (see 7). With the same equipment, a small move in the selling price changes the “scale at which it holds up” wholesale. This is what it means to say you cannot work back the equipment cost without first locking in the sales channel and the price.

Economies of scale improve the economics. But you need stable sales channels

Here some may have a nagging objection. “The bigger you build, the lower the construction cost per unit”—the so-called economies of scale. If so, isn’t building boldly large the better deal in the end? Doesn’t it clash with the “sales channel first” order so far?

To state the conclusion: economies of scale are real, and they do indeed improve the economics. With the same specification, the bigger you build, the lower the unit cost of the shell and HVAC, and that speeds payback. The study that produced the earlier estimate, too, positions scale as “an important factor deciding whether the economics hold up” (see 7). So “scale cannot save the break-even” is an overstatement; the correct direction is that scale improves both unit cost and break-even.

That said, there is a weakness that does not vanish even when you go large: fragility to price and contract shocks. The earlier sensitivity—“a 20% drop in the selling price sends the break-even scale leaping from tens of square meters to over a thousand” (see 7)—does not vanish even when you scale up. If anything, the larger you build, the larger the capacity that sits idle when the selling price collapses. So whether you can turn scale’s cheapness into a tailwind for payback hinges on whether you have a sales channel that will steadily absorb that large capacity. If the buyer is stable, the unit-cost drop that comes with scaling up directly speeds payback. Conversely, if you build large first because the unit cost falls and put off the outlet, a small move in price or contract leaves the whole big box idle. Scale is a powerful means of raising the economics, but the weakness to price and contract instability does not vanish with scale, so it only becomes a tailwind once stable sales channels accompany it—that is the relationship between the two.

This economy of scale itself shows up clearly in the estimates. In the same study analyzing vertical farm construction cost, the average scale elasticity is about −0.17, meaning that expanding scale 100-fold is estimated to lower construction cost per unit by about 55% (see 7). “The bigger you build, the cheaper it gets” is real as a number, too. And the study itself, while envisioning that as economies of scale advance, larger vertical farms could become the norm, places the securing of stable transactions (buyers) as the premise for that to hold. Scale does improve the economics, for sure, and you need stable sales channels so as not to drop that effect—it cuts two ways.

A small-scale case of a different method also shows a threshold where the economics flip with the combination of scale and labor cost. In a financial estimate for one hydroponics unit (a different method called GREENBOX), the economics broke down only under the combination where the skilled-labor wage exceeded $19 an hour and the number of units installed was under 300 (see 8). The threshold numbers are specific to that technology, but it is one example of the structure in which scale works as a premise for the economics: “once scale falls below a floor, the weight of labor cost bites all at once and it stops holding up.”

How to buy—the procurement trap that stands before payback

We have looked at “how much you spend and over how many years to recoup it.” But to build the factory to that blueprint, there is one more gate you have to clear: “from whom, and how, you buy.” However correct the payback calculation, get the way you procure wrong and you lose time and money before you ever start cultivating.

HVAC, LED, cultivation shelving, the nutrient solution system, the control panel, piping, electrical equipment. Take estimates for each separately and combine cheap units, and the initial cost looks like it goes down. I, too, have seen this “fragmented procurement” many times on site, and it was the classic pattern of trying to come in cheap and ending up costing more. A vertical farm is not a collection of equipment but a single, mutually dependent production system.

What actually happens with fragmented procurement is, roughly, the following four. First, the problem of equipment not connecting to one another. The environmental control system and the LED lighting each work on their own, but the communication protocols don’t match and the two won’t talk to each other. The control side says “ask the lighting maker,” the lighting side comes back with “it’s a problem with the control settings.” Second, power supply capacity. Each maker calculates only the power its own product needs, so when you sum all the equipment, the power supply facility’s capacity turns out to be short. What I wrote earlier—“take power supply headroom in the first building”—is precisely to avoid this. Third, the spread in delivery times. The cultivation shelving arrives but the control panel doesn’t. The construction contractor is on standby, and cost goes out while you wait. Fourth, the “not our problem” finger-pointing when trouble hits. When equipment won’t run, responsibility goes vague, and in the end the client either isolates the cause themselves or pays a separate integration contractor.

Joe Swartz of the U.S. hydroponics maker AmHydro describes this kind of fragmented procurement as “the fastest way to lose time and money.” He also says that cases where a missing part is first found only at the installation stage happen frequently. It matches my own sense completely.

So the realistic course is to procure, at least for the first building, on the integrated side. To carry the inter-equipment compatibility verification and the setting of responsibility boundaries on the client’s side takes commensurate technical knowledge and time. Step into fragmented procurement before that has accumulated, and the problems just listed reproduce as is. From the second building on, after you have built up operating experience, optimize individual units—this order makes more sense. What I said on the payback-design side, “secure in the first building the room to add later,” and what I say on the procurement side, “go integrated for the first building,” are two sides of the same thing. Because you lock down the power, communications, and space headroom first within integrated procurement, you can add the content later.

If you entrust it all to one party, there is one more viewpoint to hold from the start: preparing for withdrawal risk. There is no guarantee that the contractor you entrusted the launch to will keep doing vertical-farm design, construction, and maintenance going forward. Business withdrawal, a change of direction, a staff reassignment—the place you rely on disappearing right when maintenance is needed does happen in reality. So always keep the blueprint and the procurement specification in-house, and choose, as far as possible, equipment of open standards or industry-standard specifications. At the point of ordering, confirm “can we run it ourselves if this contractor disappears, can we hand it off to another contractor?” How to scrutinize the contents of procurement, and how to use integration versus fragmentation, is itself a judgment in its own right. #c0003_procurement

Decide by the direction of payback, not the market’s minimum figure

Building on the flow so far, let me draw one line. When you talk like this, you come to want to know the minimum figure: “how much, after all, can I start from?” But from the view I have conveyed today, a single number—“a minimum of X yen”—that strips out the premises of the project’s sales channel, production density, and automation is hardly material for judgment. The same goes for “tricks to do it cheap” lined up without speaking of operational risk; from the view of payback speed, that is the entry point to be most wary of. Concrete estimates of figures and payback periods are a domain to be worked out not with generalities but with your own project’s business plan and, if needed, with the help of a specialist.

That said, payback speed is not always the right answer. When the funds you can use have an immovable ceiling and additional procurement is hard, or when you decide to treat it as a short-term trial run to test demand, a judgment that prioritizes holding total cost down first is entirely possible. Which to optimize for—total-cost minimization or payback speed? The question that tells them apart is just one: whether the premise of recouping that investment in full stands. If the premise stands, optimize for payback speed; as long as it doesn’t, hold total cost down and see how it goes. Which you take is decided not by the market rate but by the conditions of your own funds and sales channel.

Finally, let me look back at how the view has changed up to here. At the start, you were comparing “how much is needed, is cheaper the better deal” by the figure itself. But in reality, first the sales channel sets the ceiling on volume and price that can be sold, from there you subtract over how many years to get it back, and last the upper limit of the equipment cost emerges. The figure is the number that emerges last, after you have drawn the picture of how it gets recouped. And even within the same act of saving, cutting the shell or HVAC pushes payback further out, while adding shelving or automation later—if you took the room to absorb it in the first building—speeds payback; the two are exact opposites. From now on, look not at the size of the figure but at which direction the cost you cut is facing.