What a kilogram of blueberries costs on the Moon

A kilogram of freeze-dried blueberries costs a few tens of dollars at a grocery store. Delivered to the lunar surface it costs at least $770,000, and plausibly several times that. The freight is four to five orders of magnitude more than the food.

That ratio is why Artera exists. We build controlled-environment growing pods for spaceflight, on the premise that past some delivery price it becomes cheaper to grow food at the destination than to ship it there. Everything then turns on the delivery price — which, it turns out, almost nobody can cite.

We went looking for that number. Most of the figures in circulation are models wearing the costume of prices, or real prices that stopped being true a decade ago. Below is every public figure we could source, on one axis, labeled with the year each was last quoted. Then what the same kilogram costs to grow, which is where the argument gets uncomfortable.

The kilogram

One kilogram of freeze-dried blueberries is about 6.7 kg of fresh fruit with the water taken out. Blueberries are roughly 85% water; drying a kilogram of fresh fruit sheds about 0.8 kg of it and leaves about 0.15 kg of product.

You dry it because freight is sold by mass. Shipping fresh fruit means paying lunar delivery rates to move water, which is the one commodity you would rather make on site than import. Freeze-drying is not really a preservation choice — it is a response to a price. The food is dehydrated because the rocket is expensive.

Hold that kilogram in mind. It is the unit for everything that follows.

Three kinds of number wearing the same units

A published price is what a vendor will charge you today. There is exactly one at this altitude: SpaceX’s rideshare rate, $7,000 per kilogram above a 50 kg minimum, checked in 2026. It buys an insertion opportunity and nothing past it.

A model multiplies that rate by a location factor — the kilograms of vehicle you must stage in low Earth orbit to move one kilogram elsewhere. NASA’s Baseline Values and Assumptions Document publishes them: 7.2 to the lunar surface, 2.77 to 4.33 to the Martian one. Multiply and you get the grey rows.

A contract is what somebody actually paid. NASA’s Commercial Lunar Payload Services program has eight delivery task orders on the books, individually $77.0M to $322.8M. The audit that tabulates them notes that NASA’s market research found no commercial provider with capacity for payloads over 100 kg. Divide and the floor is $770,000 per kilogram — a floor, because 100 kg is the class ceiling rather than any manifested mass, and the individual instruments flown are described as ten to fifteen kilograms each. That division is ours, not NASA’s; the agency publishes mission prices and never a rate.

So for the same destination, the model says $50,400 and the invoices say at least $770,000. The method understates reality by fifteen to sixtyfold — and we only know that because the Moon is the one destination where somebody is actually buying.

Why the Moon is expensive and Mars looks cheap

The chart’s most misleading feature is that Mars appears cheaper than the Moon. That is not an error in the table. It is an assumption showing through, and unpacking it is the whole Moon-versus-Mars story.

The Moon has nothing to push against. Every meter per second of arrival velocity must be burned: roughly 0.9 km/s of orbit insertion, then about 1.9 km/s of powered descent, all of it propellant carried from Earth. Through the rocket equation that compounds exponentially, which is why the lunar factor is 7.2.

Mars has an atmosphere, and it half helps. Aerodynamic braking sheds most of the arrival velocity for free — BVAD states the assumption directly, that both architectures use “chemical propulsion and aero-braking when possible”, and only one of the two destinations makes it possible. But the Martian atmosphere is a famously bad compromise: thick enough to demand a heat shield, too thin to land heavy payloads on parachutes alone. The discount arrives bundled with an entry system whose mass the location factor never counts, because that factor is defined over propulsion assets — engines, fuel, tankage — and a heat shield is not one.

The tell is in the rows we left off the chart:

	one way to the surface	surface and back
Moon	7.2	13.8
Mars	2.77 – 4.33	9.50 – 14.83

Round trip, the two destinations converge. The gap exists only in the one-way case, because a one-way Mars delivery never buys the propellant to slow itself down or to leave again. That is correct as a statement about propellant, and silent about cost.

The two destinations also differ in a way no factor captures: cadence. The Moon is three days away, with launch windows on a roughly monthly rhythm, a second and a half of comms latency, and — decisively — a functioning market. CLPS landers are being bought right now, which is exactly why real lunar numbers exist to cite at all.

Mars is six to nine months away, with transfer windows that open about every 26 months, comms latency running to twenty minutes each way, and no cargo delivery service at any price. Nothing to buy, so nothing to quote.

That inverts the naive reading of the chart. The destination where freight looks cheapest is the destination where freight can help you least. On the Moon, a grow pod competes against a real freight line and has to beat a number. On Mars there is no number to beat, because between windows there is no delivery at all. A crop cycle measured in weeks against a resupply cycle measured in twenty-six months is not a price comparison. It is an availability argument, and availability appears nowhere in dollars per kilogram.

What the same kilogram costs to grow

Here is where an honest version of this post stops flattering its author.

We ran the production side of exactly this unit — one kilogram of blueberries, grown and picked with nobody in the loop. Scaled to the freeze-dried kilogram the chart prices, at 6.7 kg fresh:

• 1.5 to 2.9 mature plant-years. A bush yields 2.3–4.5 kg a year at maturity, so the kilogram is roughly two entire bush-years of fruit.
• 25 to 42 robot-hours of picking. Selective autonomous harvest runs about 28 seconds per fruit at 83% success, and the cost is charged per fruit, so small fruit is punished. Blueberries run 400–670 to the kilogram. Strawberries need roughly a tenth of the picking time for the same mass, purely from being bigger.
• About 53 kWh to freeze-dry, at 5.2–19.5 kWh per kilogram of water removed. Against the 500 W pod we are designing to, that is more than four days of the entire spacecraft’s power budget, for one kilogram, before any lighting.

And then the finding that matters most, which is that we cannot grow it at all. Blueberry takes 6–8 years to reach full production, requires 800–1,500 chilling hours in a narrow band around freezing, and stands 1.5–2.4 m tall against the 0.65 m cell our pod actually has. Worse, the chill is not a one-time cost: sustained warmth negates accumulated chilling, so a chamber held at the productive constant setpoint every other crop wants will actively suppress fruiting. Those are properties of the species, not of our hardware. Our own evidence points at strawberry instead — 0.25 m tall, 77.88 g of fresh mass per square metre per day, with an actual robotic-harvest literature behind it.

We kept blueberries as the example anyway, because the shape of the answer is the lesson: a delivery price tells you what you are allowed to spend, not what you are able to build. Seven hundred and seventy thousand dollars of freight headroom does not conjure a bush that fruits without a winter.

Why we went looking

Our own financial model carried $250,000 per kilogram to the Gateway. It was labeled a sourced estimate and attributed to the lunar logistics contract’s scope. Fetching the citation for an email, we could not construct one — not because the source was hard to find, but because the arithmetic does not exist. The Gateway Logistics Services contract publishes a $7 billion ceiling across all providers and all missions over a twelve-year ordering period. No per-mission value. No manifested mass. Nothing that survives division.

An empty cell in a financial model does not stay empty. It fills with something that sounds researched, and then it renders onto a page and reads as fact to everyone including the people who put it there. Ours is now marked a placeholder.

The destination has moved besides. NASA’s Inspector General recorded in June 2026 that the agency reformulated the Artemis campaign that February, and stated the consequence plainly: rather than using the Gateway as a staging location for lunar surface missions, NASA intends to shift toward a permanent Moon base. The Gateway’s habitation module is among four systems worth $5.9 billion that the agency has announced plans to cancel or repurpose, with a stop-work order issued to its contractor. So the destination with no price attached is also the destination whose anchor customer has left. Those are probably the same fact.

What to ask for instead

Cost per kilogram is a category error beyond low Earth orbit. Nobody sells a kilogram there. They sell a mission, and the mass is a consequence of the mission’s design rather than the unit of sale — which is why every honest figure above is a launch rate that stops at LEO, a model with an architecture baked into it, or a contract value that somebody divided.

When a number arrives claiming otherwise, three questions strip it down: what mass does it divide, what does the money buy, and what year was it quoted? Most figures in circulation fail the third before they reach the first.

Artera is pre-flight. What we have today is the Earth Analogue Experiment — one instrumented node and a gravimetric watering cell — which is a long way from a pod on the lunar surface. But the arithmetic above is what the company is for, and publishing the version that kills our own crop choice seems a better use of it than publishing the version that doesn’t.

---

The figures and their sources

Delivery

Destination	Figure	Kind	Cited from	Source
Low Earth orbit (SSO)	$7,000/kg, $350k floor	Published price	2026	SpaceX rideshare rate, retrieved 2026-07-16
Mars orbit	$15,120 – $23,590	Calculation	2022 × 2026	BVAD Rev 2 Table 3-20, factors 2.16–3.37, × the SpaceX rate
Mars surface	$19,390 – $30,310	Calculation	2022 × 2026	BVAD Rev 2 Table 3-20, factors 2.77–4.33, × the SpaceX rate
Lunar surface	$50,400	Calculation	2022 × 2026	BVAD Rev 2 Table 3-20, factor 7.2, × the SpaceX rate
Lunar orbit	$51,100	Calculation	2022 × 2026	BVAD Rev 2 Table 3-20, factor 7.3, × the SpaceX rate
Lunar surface	$1,200,000	Vendor list price	2016	Astrobotic press release, 17 November 2016
Lunar surface	at least $770,000 – $3,200,000	Contract-derived calculation	2024	NASA OIG IG-24-013, task orders $77.0M–$322.8M against a sub-100 kg payload class
NRHO / Gateway	none	—	—	Gateway Logistics Services publishes only a $7B program ceiling

Government contract values are what NASA paid specific vendors under firm-fixed-price task orders that included agency-directed changes; the same audit attributes $208.2 million in cost growth, 82 percent of it to those changes. A vendor list price establishes an offer, not a transaction. The location-factor figures are modeled comparisons and were never intended as tariffs. None of these is a quote available to a third party.

Production

Per kilogram of fresh fruit, from our own resolved research spike; the freeze-dried figures above multiply these by 6.7.

Quantity	Value	Source
Water fraction	~85%	1 kg fresh sheds ~0.8 kg water, yielding ~0.15 kg dried
Yield at maturity	2.3 – 4.5 kg per plant-year	Iowa State Extension
Time to full production	6 – 8 years	OSU Extension EC 1304
Chilling requirement	800 – 1,500 h (northern highbush)	OSU Extension EC 1304; MSU Extension P3067
Mature height	1.5 – 2.4 m	OSU Extension EC 1304
Selective harvest rate	28.2 s per fruit, 83% success	Robofruit, Parsa et al. 2023
Freeze-drying energy	5.2 – 19.5 kWh per kg water removed	MDPI Sustainability, 2021
Strawberry productivity	77.88 g fw/m²·d, 0.25 m tall	BVAD Rev 2

Fruit-per-kilogram counts assume 1.5–2.5 g per blueberry, which is an internal assumption and not sourced. We make no nutrition, shelf-life, or water-recovery claim here; those remain open in our own records. The plant-year and robot-hour figures for the dried kilogram are derived, not measured.