The thesis, the ask, and — deliberately — the assumptions you should test. We would rather you stress the model with us than discover its edges later.
Most farms sell one thing and live on thin spreads. This project stacks four businesses onto one shared base of land, buildings, sunlight and livestock: daily dairy cash, Eid-peaked beef, self-generated renewable energy, and globally-rentable AI compute. The same corn that feeds the cattle indirectly powers the GPUs — because the cattle's dung becomes biogas, the roof over their heads carries solar, and that electricity is sold to the world as compute at $4.5/GPU-hour. Profit is recycled into more cows, more processors and more land, compounding the loop from Tk 4.1 cr revenue in Year 1 to Tk 22.8 cr in Year 3.
Growth capital accelerates the two highest-return levers ahead of what the loan alone allows. Percentages are indicative and adjust to the final raise.
The model prices each B300-class processor at $4,500. Market pricing for top-tier AI accelerators has historically run far higher; this input reflects a specific sourcing thesis and should be validated against a live supplier quote before funds are committed. Sensitivity: at higher hardware cost, the compute line's payback lengthens but the rental rate assumption ($4.5/GPU-hr) is conservative against observed market rates, partially offsetting.
The model assumes 1 kg of Tk 80 corn yields 7 kg of hydroponic fodder, and 7 kg of fodder converts to 1 kg of live-weight gain. The sprouting yield is standard; the feed-to-gain ratio is an optimistic planning figure. Diligence step: a 90-day feeding trial on the first cohort will produce the farm's real conversion number, and the model recalculates from a single cell.
Solar is daytime-only; biogas is modest. Until battery storage is added, the compute fleet draws grid power after dark — a real cost the model carries in opex (Tk 0.77 cr by Year 3) and a reliability exposure in a load-shedding grid. The use-of-funds allocates 12% to storage precisely to close this gap.
Compute revenue assumes half of all available GPU-hours are rented. Decentralised marketplaces are growing but demand is not guaranteed; every 10 points of utilisation moves Year-3 compute revenue by roughly Tk 2.4 cr in either direction.
Biosecurity separation, vaccination schedule, retained vet, and animal-level tracking (the same system that powers the consumer traceability story doubles as the health log).
Licensed RCC design, full waterproofing membrane, corrosion-resistant fittings, and two-way ventilation so fodder-floor humidity and cattle-floor ammonia never mix. Plinth raised above surveyed flood level.
Owned bulk chiller and refrigerated van rather than third-party logistics; generator + biogas backup keep chilling through grid outages.
Premium positioning and direct channels defend price; the compute line is fully uncorrelated with Dhaka food prices, cushioning the whole.
Because compute is electricity resold at a markup, and this farm makes its own electricity from sunlight and dung. Cheap renewable power is the single biggest cost advantage in the GPU-rental market — and the farm has spare roof, land, and a story no data centre can match.
A fattening cycle is months, not years: buy a calf at Tk 80,000, feed it home-grown fodder, sell near Tk 1.9 lakh at the Eid peak. Same land, several times faster capital turns.
GPU-hours can be listed on multiple marketplaces, sold directly to local AI teams, or repurposed for rendering/research workloads. Meanwhile dairy and beef continue independently — the loop degrades gracefully rather than failing together.
Bangladesh has no functioning livestock organic certification, so the brand claims what it can prove instead: antibiotic-free, hydroponic-fed, farm-traceable — backed by live cow-cams and per-product QR, which is stronger than an uncheckable label.
The full data room — the live financial model workbook, the 3-year plan, and the operating plan — is available on request.
Request the data roomThis page is a summary of management projections prepared for discussion purposes. It is not an offer of securities, and figures include deliberately optimistic planning inputs identified above. Prospective investors should conduct independent diligence.