Autonomous Grow. Smart Data. Real Value.
A modular grow and edge-compute platform in a single 40-foot footprint. 160 plants inside, a pollinator garden on the roof, and an AI compute node bolted to the end — two revenue streams on one piece of infrastructure.
Gaijin Box is a modular, off-grid capable grow and compute platform that combines licensed cultivation with edge data-center infrastructure in a single 40-foot footprint.
The idea was to have 160 plants in a standard shipping container — using exhausts, water pumps, HVAC, Raspberry Pis for condition setting and tracking, and an irrigation system to disperse water and nutritional chemicals and supplements to all plants.
Mounting a distributed AI compute node on the exterior of the Gaijin Box turns a single infrastructure investment into a dual-income asset with two fully independent revenue streams. The solar array, grid connection, smart panel, internet hookup and site lease already exist for the grow operation. Adding the compute node costs only incremental hardware — not another full infrastructure buildout.
Anyone can buy a shipping container. Not everyone can navigate all of it at once.
Cannabis cultivation licensing, compute hosting agreements, solar interconnection, commercial site leasing, and multi-party revenue sharing across multiple jurisdictions. The operator who builds this compliance playbook first owns a first-mover advantage that compounds with every unit deployed.
The whole thing, on one sheet
160 plants. Vertical racks. Full-spectrum LED. Automated nutrient delivery. Every variable tracked and controlled by the Raspberry Pi network — accessible remotely from any device.
The section drawingTwelve systems share one 40-foot shell. Take one apart at a time — click a number on the drawing, or a name in the list.
A shallow green roof — four to six inches of lightweight growing medium seeded with native flowering plants — occupying the rear third to half of the roof, co-existing with the solar array. Shipping containers are engineered to stack under tens of thousands of pounds, so the added load is negligible.
Monocrystalline panels occupy the non-garden roof area. They generate 5–8 kW peak, primarily offsetting lighting load during daylight hours. Grid-primary, solar-supplemented, battery-buffered — the architecture is honest about what an indoor grow draws.
Iron-phosphate (LiFePO₄) batteries provide 20–40 kWh of storage — enough to buffer demand peaks, smooth solar intermittency, and provide 30–60 minutes of backup for critical systems during grid outages. A Span smart panel handles load management and grid optimization.
Two opposing rack banks (left and right walls) each hold 4 vertical tiers of 20 plants — 80 per side, 160 total. Racks are modular aluminum extrusion frames with drain channels feeding the recirculating nutrient system. A 4-foot central walkway provides full-length maintenance access; it runs toward the viewer in this section and so cannot be drawn. Plant spacing is optimized for the target crop.
Commercial-grade LED grow lighting runs the full ceiling length with supplemental side-lighting for lower tiers. Spectrum, intensity and photoperiod are programmable via the Raspberry Pi controller — allowing automated vegetative-to-flowering cycle transitions. Spectrum tuning reduces wattage versus traditional HPS while improving yield quality and terpene expression.
A central nutrient reservoir feeds a drip or NFT hydroponic system via a main trunk line. Peristaltic pumps meter concentrated nutrient solutions — macros, micros, pH adjusters — based on inline EC and pH sensor readings. Each tier has pressure-compensating drip emitters ensuring even delivery to all 160 plants.
Three 100-gallon tanks with EC/pH dosing. Closed-loop recirculation recovers 85–95% of water, dramatically reducing the water footprint versus soil-based or drain-to-waste farming.
A purpose-sized mini-split maintains 70–82°F and 50–65% RH across the grow cycle. A dedicated dehumidification unit handles late-flowering transpiration loads. Intake fans with MERV-13 filtration prevent contamination.
Exhaust fans with carbon filtration manage odor — critical for cannabis compliance, and the difference between a unit a commercial landlord will host and one they won't.
Sensors report temperature, relative humidity, CO₂, VPD, EC, pH, light intensity and pump status. Each node logs data and controls actuators — HVAC, lights, nutrient pumps, exhaust — against programmable setpoints.
Multiple Raspberry Pi 5 nodes form a local mesh monitoring and control network. A web dashboard displays real-time conditions and historical trends for the leaseholder. This network is not a convenience feature — by the last chapter it has become the billing system.
The node mounts on the container's right-end exterior in a sealed, independently climate-controlled enclosure with its own liquid cooling loop, held below 40% RH and separate from the grow. Technicians never need to enter the grow space to service it.
Mounting a distributed AI compute node on the exterior turns a single infrastructure investment into a dual-income asset with two fully independent revenue streams.
The XFRA model — pioneered by Span in partnership with NVIDIA and PulteGroup — installs compact, liquid-cooled GPU compute nodes on building exteriors, tapping unused grid capacity to provide AI compute at a fraction of centralized data center cost.
Honest about power needs. Ambitious about reducing them. Grid-primary, solar-supplemented, battery-buffered.
The peak power budget, made liveEvery range below is a documented figure from the build spec. Move them and watch what the box actually asks of the grid.
| LED grow lights | 12–18 kW |
| HVAC + dehumidification | 4–8 kW |
| Compute node | 10–30 kW |
| Irrigation + controls | ~0.5 kW |
| Total peak | 26–57 kW |
| Solar output | 5–8 kW |
Peak draw and annual energy are different quantities, and the difference matters: across a full year the rooftop array is estimated to offset 10–25% of total consumption, not the instantaneous share the bar above shows at midday.
A native wildflower and pollinator garden on the container roof is more than a visual identity — it's a functional ecosystem layer with real structural, thermal, and ecological benefits.
Shipping containers are engineered to stack under tens of thousands of pounds. A shallow green roof — four to six inches of lightweight growing medium seeded with native flowering plants — adds negligible structural load. The garden occupies the rear third to half of the roof, co-existing with the solar array.
The garden is our logo.
Any commercial or industrial-zoned property with grid access and sufficient footprint. Flexibility is the product.
Cannabis is the highest-margin leaseholder crop right now, but it carries regulatory risk. A box designed to be reconfigured is a hedge the operator can actually sell.
If wholesale prices keep compressing — they are, in mature markets — or if a leaseholder loses their license, you have an empty box. A modular Gaijin Box designed to be reconfigured for saffron, microgreens, specialty mushrooms or culinary herbs gives the operator and leaseholder flexibility that pure cannabis infrastructure doesn't.
Nine candidates, rankedSort by whichever constraint actually binds. Open a row for the case.
The most expensive spice in the world by weight. The reason it's expensive is pure labor: each crocus flower produces three stigmas that must be hand-harvested during a 2–3 week window per year. Indoor vertical growing completely changes that calculus — with precise photoperiod and temperature manipulation you can force multiple flowering cycles a year, and some indoor operations are achieving 3–4 harvests annually by cycling corms through artificial dormancy.
It needs no cannabis license, faces zero regulatory friction, and the US imports nearly all of its saffron. Domestically grown, certified, premium saffron commands a serious premium with chefs, specialty grocers and pharmaceutical buyers. The plant is compact, drought-tolerant, and doesn't need aggressive lighting — lower electricity draw than cannabis, and it fits vertical racks well.
Real wasabi — not the horseradish paste served at 95% of sushi restaurants. What makes it compelling isn't the price, it's the supply gap. Authentic wasabi requires cold, clean running water, high humidity, dappled indirect light and a narrow temperature band (46–70°F): conditions nearly impossible to hold outdoors outside a handful of mountain stream environments in Japan, the Pacific Northwest and New Zealand.
A Gaijin Box is essentially a purpose-built wasabi environment. The 18–24 month wait is front-loaded, but wasabi is a perennial that keeps producing once established — stagger the plantings and a box that has been running two years becomes a perpetual income generator. If you crack controlled-environment cultivation at scale there is essentially no domestic competition.
Lion's mane, reishi and cordyceps: extraordinarily fast, low-light crops. The market for functional mushrooms is growing at over 8% annually on supplement demand, and domestic supply is severely limited.
A Mushroom Box variant — optimized for humidity, CO₂ and airflow rather than lighting — would have dramatically lower electricity costs than the cannabis-configured unit and could reach positive cash flow faster. Lower barrier to entry, no license, faster cash flow, lower per-unit revenue ceiling.
The sleeper category. High-end restaurant supply is chronically undersupplied with ultra-fresh exotic herbs — micro-shiso, Vietnamese coriander, specialty basils, edible flowers — which wholesale far above commodity herbs.
The business model is different: instead of wholesale, establish direct accounts with 15–20 high-end restaurants and become their exclusive supplier of herbs they can't reliably source anywhere else. A relationship business with high switching costs once a chef builds a menu around your product. Fast cycles also mean fast cash flow.
The highest-margin leaseholder crop and the reason the box exists. About 40% of legal cannabis is indoor-grown, and an estimated 15,000+ cultivation licenses are active in the US with more states legalizing.
It also carries every regulatory cost the others don't: state licensing, compliance reporting, CCTV and biometric access, crop insurance, and odor control that is not optional. Wholesale price compression in mature markets is the single biggest threat to leaseholder economics — which is the entire argument for this chapter.
Up to $2,000/kg for premium Japanese-market export grade — the second most expensive mushroom in the world after truffles, which can't be cultivated in containers at all.
The catch is that matsutake form a symbiotic relationship with tree roots in the wild, which makes them notoriously difficult to cultivate. Several Japanese and Korean research programs have made progress on indoor methods, but this is a research bet, not a lease-ready crop.
Fresh turmeric wholesale isn't spectacular, but certified organic, specialty or heirloom fresh turmeric reaches a real premium through direct channels — and pharmaceutical-grade curcumin extract is in surging demand from the supplement industry.
It grows fast, loves the humidity and warmth a container already holds, produces dense rhizomes that pack well into vertical systems, is legally unrestricted everywhere, and the domestic supply chain is almost entirely dependent on imports. Not the highest ceiling — potentially the most operationally simple.
Genuinely one of the highest-value crops by weight grown in North America, with wild-simulated roots fetching more still. The problem is time: peak market value takes 5–8 years, which is a poor fit for a lease that has to be paid monthly.
But it could work as an equity crop for the operator. Instead of leasing the space, the operator plants ginseng, tends it with minimal labor, and waits — while the compute revenue covers operating costs during the grow. By year 5–6 a single box could represent $80,000–$150,000 in harvest value. The compute node essentially pays you to wait.
Harder and slower, but the upside is enormous if you solve the biology — the US imports virtually 100% of its vanilla. The vine takes 2–3 years from planting to first harvest, requires hand pollination of every flower, and then a curing process that takes another 3–6 months outside the container.
Capital tied up for three years with no return is a serious problem for the lease model specifically: the leaseholder needs income to pay their monthly lease. Viable as a long-term play for an owner-operator — a "Vanilla Box" sold rather than leased.
All costs are borne by the operator and amortized into the lease. The single most important capital decision is who buys the GPUs.
Capital expenditure — one unit| Line item | Low | High | Midpoint |
|---|---|---|---|
| Container & site | |||
| 40ft ISO container — used, one-trip preferred | $3,500 | $7,000 | $5,000 |
| Delivery & crane placement — distance and site access dependent | $1,500 | $5,000 | $3,000 |
| Site prep — pad, utility hookup, trenching | $5,000 | $15,000 | $8,000 |
| Permits & licensing — structural, electrical, zoning | $2,000 | $10,000 | $5,000 |
| Interior build-out | |||
| Insulation & wall finishing — closed-cell spray foam | $4,000 | $9,000 | $6,000 |
| Vertical grow racks — 4 tiers × 2 banks, 160-plant capacity | $14,000 | $24,000 | $18,000 |
| Full-spectrum LED array — commercial tier, dimmable | $18,000 | $32,000 | $24,000 |
| Hydroponic irrigation + plumbing — recirculating drip/NFT | $8,000 | $18,000 | $12,000 |
| Reservoir + dosing pumps — 3 × 100gal, EC/pH dosing | $3,000 | $6,500 | $4,500 |
| HVAC + dehumidification — mini-split + flowering-load stage | $8,000 | $16,000 | $11,000 |
| Exhaust + carbon filtration — odor compliance | $2,500 | $6,000 | $4,000 |
| Raspberry Pi control network + sensors | $2,000 | $5,000 | $3,000 |
| Security system — biometric, CCTV, alarms | $4,000 | $9,000 | $6,000 |
| Electrical panel & distribution — 200A+ service | $7,000 | $14,000 | $10,000 |
| Power & sustainability | |||
| Rooftop solar array — 5–8 kW monocrystalline | $8,000 | $15,000 | $11,000 |
| LiFePO₄ battery bank — 20–40 kWh | $14,000 | $28,000 | $20,000 |
| Span smart panel + interconnection | $5,000 | $8,500 | $6,500 |
| Roof garden | |||
| Membrane + growing medium — EPDM, 4–6" perlite/coco | $2,500 | $6,000 | $4,000 |
| Plant material + drip irrigation — regional species | $800 | $2,500 | $1,500 |
| Compute node | |||
| Exterior enclosure + mounting — sealed, liquid cooling plumbing | $3,000 | $7,000 | $5,000 |
| Total CAPEX — GPU hardware partner-provided | $117,300 | $242,500 | $167,500 |
| GPU compute hardware — 4–8× NVIDIA Blackwell, if operator-provided | $120,000 | $320,000 | $200,000 |
| Total CAPEX — operator-provided GPUs | $237,300 | $562,500 | $367,500 |
The operator owns the box and the infrastructure. That's it. The leaseholder pays for everything related to growing, including the grow electricity.
| Operator cost | Per box / mo | 60 boxes / mo | Annual |
|---|---|---|---|
| Property host fee — site lease | $300 | $18,000 | $216,000 |
| Insurance + maintenance reserve | $350 | $21,000 | $252,000 |
| Internet — compute uptime requirement | $80 | $4,800 | $57,600 |
| Monitoring platform | $70 | $4,200 | $50,400 |
| Admin overhead — scaled across the fleet | $200 | $12,000 | $144,000 |
| Total operator OpEx | $1,000 | $60,000 | $720,000 |
The defensible lease structure is not a flat fee at all. It's a percentage of gross crop revenue — verified automatically by the box's own monitoring system.
Why the flat lease brokeThe lease price was reverse-engineered from the operator's desired margin rather than forward-engineered from what the leaseholder can actually support. That's backwards — and it only became obvious when the same $4,500/month rate was run against a crop other than cannabis.
Saffron at optimized density and four forced cycles grosses about $50,400 per box per year. The flat lease alone is $54,000. Wasabi grosses $26,400 against the same $54,000. Both are underwater before a single input cost is counted. A vacant box generates zero revenue for anyone, so pricing the leaseholder out is not a rounding error — it's the whole model failing.
The methodLease = 20% of the leaseholder's gross crop revenue, audited quarterly against actual yield data from the monitoring system already built into every box. Compute revenue belongs entirely to the operator — separate from the crop lease, and the operator's guaranteed income floor. Per-box gross figures below are modelled for the tristate market: dense restaurant culture, high wholesale prices, legal cannabis in all three states, and almost no local competition for premium wasabi or saffron.
| Crop | Gross | Lease @ 20% | Leaseholder keeps | Their costs | Net |
|---|---|---|---|---|---|
| Cannabis | $10,000 | $2,000 | $8,000 | $4,500 | $3,500 |
| Saffron — D2C | $10,267 | $2,053 | $8,213 | $2,200 | $6,013 |
| Wasabi — direct restaurant | $6,167 | $1,233 | $4,933 | $2,200 | $2,733 |
Compute hosting adds $2,500 per box per month to the operator, on top of the lease. Operator OpEx is $1,000 per box per month.
Guarantees operator revenue scales with leaseholder success. No one profits from a struggling grower.
Eliminates the pricing-out problem entirely. The lease is always affordable because it is always proportional.
Makes the monitoring infrastructure a contractual necessity rather than a nice-to-have.
Gives the operator a natural incentive to help leaseholders succeed — better yields mean higher lease revenue, so agronomic support, preferential nutrient purchasing and crop optimization consulting all belong in the lease package.
The yield data is the billing system.
Chapter two's sensor network stops being a feature the moment the lease is written as a percentage. It becomes the instrument of record for every invoice — which is also why the operator, not the leaseholder, maintains it.
Who owns whatAmbiguity in responsibility is where deals fall apart and legal exposure accumulates. These are the load-bearing rows.
| Obligation | Operator | Leaseholder | Property host |
|---|---|---|---|
| Container, buildout, solar, compute node | Primary | — | — |
| Site prep & utility connection | Shared | — | Primary |
| Cultivation license & state reporting | — | Primary | — |
| Zoning approval | Shared | — | Primary |
| Compute hosting agreement | Primary | — | — |
| HVAC, LED, irrigation, node & software maintenance | Primary | — | — |
| Roof garden maintenance | Primary | — | — |
| Crop management, labor, harvest & sale | — | Primary | — |
| Electricity — grow portion | — | Primary | — |
| Electricity — compute portion | Reimbursed | — | — |
| Lease payment | Receives | Pays | — |
| Property hosting fee | Pays | — | Receives |
| Compute hosting revenue | Receives | — | — |
| General site security | — | — | Primary |
Gaijin Box sits at the intersection of three growing markets — and the combination is largely unaddressed by any single integrated product.
All licensed cultivators in legal markets, all CEA operators seeking modular infrastructure, and all commercial property operators who could host distributed edge compute. Roughly 40,000 potential units globally over ten years across cannabis-legal US states, Canada, Germany, the Netherlands and emerging EU markets. $2.4B/yr recurring revenue TAM. Assumes full penetration, which is theoretical.
US-only, cannabis-legal states with active commercial licensing. Top 12 markets: California, Colorado, Michigan, Illinois, Massachusetts, Nevada, Oregon, Washington, New York, New Jersey, Arizona, Maryland. Assuming 20% of licensed cultivators would consider a turnkey container lease over building their own facility: ~8,000 licenses × 20% = ~1,600 units, plus compute hosting on those same units. $96M/yr recurring.
Realistic near-term capture with a small team, limited capital and a single-state focus — Colorado or Michigan as a launch market, for their established licensing infrastructure and favourable zoning. 25–50 units in years 1–3 on a waitlist model, with 1–2 state expansions by year three. 25 units at $167.5k CAPEX, $672K/yr recurring.
Cannabis moving Schedule I → III would expand banking access, reduce compliance burden and bring institutional capital into the sector — inflating demand for professional grow infrastructure.
GPU demand is expected to outpace centralized data center buildout for the foreseeable future. Distributed compute is a structural response to that gap, not a trend.
Container farms are increasingly accepted in urban planning frameworks, with several cities actively incentivizing them as food security infrastructure. Regulatory friction is declining.
No company currently combines licensed grow infrastructure with distributed edge compute in a single commercial lease product.
Legal cannabis wholesale has dropped significantly in mature markets. Leaseholder economics deteriorate if that continues — which is what the crop-diversification chapter exists to answer.
The entire compute pillar depends on securing a real hosting agreement. Without it the business reverts to a standard container grow lease — which works, but with thinner margins and longer payback.
At $150–175k per unit, scaling to 25 units requires $3.75–4.4M. Revenue-based financing against the lease income stream may be the most accessible early path.
Every new state market requires a new compliance playbook — zoning, licensing, electrical codes and data infrastructure rules all differ. Multi-state expansion is operationally taxing for a small team.
The reference sheet. All specifications subject to engineering review.
| Base unit | 40ft ISO standard |
| Exterior | 40' × 8' × 8'6" |
| Interior | 39'5" × 7'9" × 7'10" |
| Interior volume | 2,390 cu ft |
| Tare weight | 8,380 lbs |
| Max gross weight | 67,196 lbs |
| Plant capacity | 160 plants |
| Rack config | 2 banks × 4 tiers × 20 |
| Lighting | Full-spectrum LED, programmable |
| Hydroponics | Recirculating drip / NFT |
| Water efficiency | 85–95% recirculation |
| Climate range | 65–85°F / 45–70% RH |
| Controls | Raspberry Pi 5 + web dashboard |
| Node type | Exterior sealed enclosure |
| GPU | NVIDIA Blackwell Server Edition |
| Cooling | Liquid-cooled, sealed |
| Node draw | 10–30 kW |
| Humidity isolation | <40% RH, separate from grow |
| Access | Exterior only |
| Branding | Gaijin Edge module |
| Architecture | Grid-primary / solar-supplemented |
| Solar capacity | 5–8 kW peak |
| Battery storage | 20–40 kWh LiFePO₄ |
| Smart panel | Span |
| Peak combined draw | 26–57 kW |
| Solar offset est. | 10–25% of consumption |
| Green roof | 4–6" substrate, native wildflowers |