I. The Matrix That Leaked
The entire human food chain is built on the principle of conversion: animals eat what is not fully edible for humans — in order to then become food for humans. A cow eats grass and oilcake so that a human eats the cow, not the grass and oilcake. A chicken — grain waste and by-products. Salmon — fishmeal.
This is a conversion matrix that worked for millennia before AI agents, and like any sufficiently long-existing technology, it had flaws that everyone ignored — the Mufasa loop never quite closed.
Humans throw away up to a third of food after it hits the table. Another roughly one billion tons of organic waste is generated in the process of producing that food — in the fields, in processing, in logistics, in slaughter.
All this organic matter leaked out as a pure systemic loss. To landfills, where it rotted and released methane. Into the expense line of waste operators (and of humanity), which never came back into the system.
The conversion matrix consumed. And leaked into Lethe, of course.
II. Why Was It Obvious, But Branded a Problem Only Within a Limited ESG Agenda?
Predictable demand, surplus land, a Hegelian state, and capitalists across the entire timeline — together they made waste the consumer's problem, offloading responsibility from themselves.
Organic waste stopped being predictable when AI did something economists had argued about for decades: it reduced the cost of entry into cognitive labor to zero.
A programmer in Bandar Abbas, a designer in Dakar, an analyst in Jakarta. All got access to tools that once cost a corporate budget. The tech giants, meanwhile, started not just paying taxes — they started subsidizing states directly. Not metaphorically. Literally: the governments of India, Nigeria, Indonesia turned to Alphabet, Microsoft, Anthropic for targeted action — deploy agents to retrain millions of people, digitize the informal sector, index supply chains that existed only in the heads of intermediaries. Marx's crisis of overproduction with the contraction of variable capital in effect converted subsidies into a de facto basic income, where any cognitive superstructure became a premium, and behind it a rise in consumption.
The middle class in South Asia and Africa grew not by generations, but by quarters.
A note on this de facto UBI — because this is important for understanding the mechanism. Universal basic income was not a new idea. It had existed for a long time — but through proxies: pensions, civil servant salaries, subsidized companies, social professions. The state redistributed income first through taxes, then directly. The Marxist crisis of overproduction — when automation compressed variable capital faster than the labor market could adapt — broke this construction. The tech giants started paying people. Not through taxes into the budget and then back through social programs.
What does a person do when coming out of poverty? They change their diet — a stable anthropological pattern, recorded in every wave of growth — from postwar Japan to China in the 2000s. Per capita food consumption in South Asia grew from 250 to 390 kg per year in one decade. In Sub-Saharan Africa — from 240 to 360 kg.
Billions of new breakfasts, lunches, and dinners every day.
The load hit three points simultaneously. Logistics — straits were being closed and reached their limits, fuel became more expensive, container routes became more vulnerable. Product — industrial poultry reached the limit of FCR optimization: chicken's muscle fiber texture changed, digestibility changed. People switched quietly, without manifestos — they just chose something else. By 2032, industrial poultry consumption fell by 23%. Waste operators — organics doubled over a decade, while infrastructure remained the same.
The planet produced about 940 million tons of organic waste per year in 2025. By 2035 — almost twice as much. Because every person got richer.
The rat emergency in New York in 2023 was not a rodent problem. It was a symptom. By 2035 they would have been in a hundred cities simultaneously — if the infrastructure hadn't been built.
III. The Biorobot Nobody Asked for in 65 Million Years
It was so obscure that it didn't have a single name — it was called differently depending on the language. In Russian — chyornaya lvinka. In Chinese — hei shui mang (黑水虻), "black water fly." In Latin — Hermetia illucens. In the industry — Black Soldier Fly.
Nature developed it for 65 million years without venture funding. The larva converted any organic matter into concentrated protein — food waste, plant by-products, processing residues. Not fussy. Without accumulating pathogenic microflora. Without carrying it into the finished product.
The main thing was speed.
A chicken converts feed into protein in 45 days — that's its biological limit, which no feeding protocol could shift without consequences. BSF did the same thing in 14 days.
In essence, the larva performed the role of chicken, beef, and fish: it digested organic matter to then become food itself. Only three times faster. On a substrate that used to cost money to dispose of.
Before 2028, BSF worked exclusively with industrial organic waste — predictable food-production waste, a clear stream. A niche player. The first commercial farms appeared in Southeast Asia, where the climate solved the heating problem for free. Not because they are smarter there. Because it's warmer there — and OPEX for maintaining 28°C year-round equals zero.
BSF stopped being a niche product when it was embedded into the megacity.
IV. Five Streams from One Bucket
BSF is not a monoproduct. It's a five-stream conversion platform, where each output has its own P&L. This matters, because exactly this construction made the farm's economics stable when in 2028–2031 a price war started and margins on the core meal fell from 40–50% to 15–20%.
Protein meal — 3-6% of the mass of the incoming substrate. Concentrated protein, amino acid profile comparable to fishmeal. The foundation of the entire economy.
Larval fat — biofuel. With oil above $160 per barrel by 2030, this stopped being exotic.
Frass — larval excrement, the richest organic fertilizer. Went into vertical greenhouses within the same district. Greens grown on the waste of processing waste — sounds like bad recursion, works as a closed loop.
Chitin — the structural polysaccharide of the exoskeleton. Prebiotic supplements. One of the fastest-growing segments of nutraceuticals in the 2030s.
Protein hydrolysate — liquid amino acid fraction. Liquid fertilizer and feed additive in aquaculture.
Five streams. One organism. A substrate that used to be paid to be carted off to a landfill.
V. Three Hooks That Pulled Organics Out of the Bucket
The industrial stream worked. The problem was in the household one — that very third of food that a person throws away at home.
A city dweller produces on average 10 kg of organic waste per month. A city of 2–3 million people is 20,000–30,000 tons of organics monthly. The minimum size at which BSF infrastructure becomes self-sufficient without subsidies.
Only people don't hand over trash just like that. A motive is needed. And not just one.
Hook one — health. By the mid-2020s, wearable trackers had become a standard part of urban middle-class life. More than 70% of people with above-median income monitored their health. A multispectral camera with an AI chip was built into the disposer — morphological analysis of every load: chicken bones, potato peelings, fat residues. From this, a household nutritional map was built and integrated into Whoop as an honest calorie tracker. A tracker can be fooled. A trash bin cannot. The algorithm noticed correlations: a late fatty dinner → sleep 12% worse. People who didn't care about ecology or about the money from the disposer began changing their diet for the sake of the numbers on their wrist.
Hook two — insurance. Health insurance became mandatory for basic income recipients. Insurance took behavioral profile into account. The disposer — part of the profile. Proper handling of organics affected the coefficient. Not a fine. Simply, those who didn't participate didn't get the discounts that everyone else got. An asymmetry that felt like punishment, though formally it wasn't one.
Hook three — UBI itself. Responsible handling of organics was counted as a contribution to city infrastructure. A utilization coefficient above 95% for three months in a row — a bonus. Small. In Jakarta, Nairobi, Medellín, this is a 5–8% bonus on top of basic income. For the lower segment, the disposer became the only asset that is definitely there every day.
Three different motives. One action. The system worked without coercion.
Sustainability turned into a bourgeois status signal. The waste conversion coefficient became the new electric car.
VI. Infrastructure: From Bucket to Fish
The chain was built from the bottom up. Four links.
Disposer — a 50-liter container in the apartment. Grinder, moisture normalizer, multispectral camera. All organics go there. At the output — a standardized vacuum bag, 1 liter, barcode, substrate passport. Freeze and hand over in a large batch — pays 15–25% more. Or call an autonomous courier — picked up in 4 minutes, like a parcel.
Rendering hub — in every district, ~5,000 apartments. Accepts bags, pays $20–30 per ton. Three-stage normalization: enzymatic hydrolysis, pH stabilization, moisture up to 65–70%. At the output — standardized nutrient substrate with a known nutrient profile. Not trash. Raw material.
Rendering is the hidden key link. Without a homogeneous substrate at the input, the BSF farm cannot optimize conversion. That's exactly why the first wave of consolidation was won not by tech startups, but by waste operators with ready-made logistics and relationships with municipalities.
BSF farm — and here's an important clarification. BSF doesn't need pastures. A farm producing animal protein grows upward, not taking up extra land. Trays with larvae stack into racks 8–12 levels high. A 2,000 m² warehouse with tiered organization gives the output for which a horizontal farm would need 15–20 hectares. BSF fit into an urban industrial zone, into a former terminal, into any building with sufficient ceiling height.
Only one thing was needed — a constant temperature of 26–30°C and a stable substrate at the input.
The second turned out to be harder than the first — and it was exactly here that AI made the technology possible in cities.
Industrial organics are predictable: a dairy plant produces approximately the same thing every day. The household stream — no. In summer the substrate has more fruit residues, in winter — fat ones. After holidays — peaks, on Mondays — troughs. The larva is sensitive to composition: the wrong mix — a drop in conversion, colony stress, losses. Before AI management, this made household substrate unsuitable for an industrial BSF farm.
The AI farm management system solved this in real time. It received data on substrate composition from every hub, ran conversion models, and within several hours calculated the optimal mix for the current larval batch. Multispectral cameras above each tray monitored larval density, motor activity, and color around the clock. The colony's veterinarian is an algorithm. It didn't go on vacation. Conversion was held in the 17–19% corridor regardless of what city dwellers ate last week.
Without this layer, BSF would have remained a niche player on industrial waste. With it — it became urban infrastructure.
Temperature was also needed.
Here data centers entered the story.
AI infrastructure produced computation. Computation produced heat. The heat was released into the atmosphere through cooling towers — a structural by-product of scale, which grew faster than anyone could utilize it.
The concept of zero heating emerged: data center and BSF farm — a single system with a shared thermal loop. Heat from the cooling towers goes directly into the tiers with larvae. The farm gets free heating. The data center reduces cooling costs. Moscow was the first to calculate this systemically: the heat output of urban data centers covered up to 40% of BSF farms' winter needs. The US adapted through hyperscalers — they were looking for monetization of waste heat within ESG frameworks, and BSF turned out to be an ideal consumer. Mining and AI computation were drawn to Mongolia and Kazakhstan for cheap electricity — and produced there exactly as much heat as the larva needs in winter.
RAS — recirculating aquaculture systems. Fish farms in tanks, standing in the same industrial district, in the same thermal loop. Salmon. Tilapia. Shrimp. Grown not in Norway and not in Thailand — in the neighboring building, in any city where there is BSF infrastructure.
Architects called it the "metabolic district." Cooperative investors — a 22% annual yield.
VII. Two Meal Markets: Fish and Corgis
BSF meal went in two fundamentally different directions with two different quality standards, fed by two different substrate sources.
Aquafeed — household substrate.
By 2030, fish and shrimp had become the main sources of animal protein for humans. Not one of. The main ones. Wild capture lost industrial relevance — quotas shrank, vessels became unprofitable. Traditional meat either rose in price to a premium niche, or changed so much that people quietly gave it up. Aquaculture in RAS filled the vacuum — and BSF meal turned out to be the only alternative to fishmeal that worked without FCR losses.
Thai Union validated it. Behind it — Mowi and Cermaq. But the main thing was something else: in RAS with BSF meal, mortality dropped fivefold. The larva doesn't transmit pathogens into the meal, the meal — into the fish. A closed system without antibiotics and outbreaks.
And this is where everything came together: the RAS stood in the same industrial district as the BSF farm. Not in Norway, not in Thailand — in the neighboring building. Logistics disappeared as an expense line. Salmon and shrimp stopped being a commodity that floats through three straits and two customs terminals. They became local protein — cheap, stable, available in any city with BSF infrastructure. This was exactly what the system lacked: a source of animal protein independent of the ocean, climate, and geopolitics.
Thai Union, Mowi, Cermaq entered BSF not as meal buyers — but as co-owners of production. Insurance against price volatility on an ingredient on which the survival of the core business depended.
Pet food — industrial substrate.
By 2035, the number of pets in cities exceeded the number of people aged 18–55. A demographic statistic that analysts caught with a five-year delay. A millennial with a corgi in a studio stopped being a meme and became the dominant pattern of urban life for an entire generation. The pet — emotional infrastructure. The pet food market grew to $240 billion, with 60% of protein ingredients being BSF meal.
Why exactly BSF? Not only because it's available.
High-quality BSF meal in a pet's diet gave measurable results: coat density, joint mobility, inflammation biomarkers. Studies from 2029–2031 showed a correlation between a BSF-protein-based diet and a life expectancy increase in medium-breed dogs of 15–20%. For a person who has a corgi instead of children, this is not marketing. This is an emotional bet. People paid the premium for this without discussion.
But pet food required a different raw material standard. A hard constraint: only industrial, homogeneous, verified substrate. No household unpredictable stream.
Therefore, next to large food production facilities — dairy plants, fish processing, meat combines — separate refineries were built. Not district hubs. Industrial facilities right at the source with a predictable homogeneous stream. This homogeneity yielded meal that can be put into a pet's food.
Result: the household stream feeds fish in the urban RAS. The industrial stream feeds the millennial's corgi. Two sources, two standards, two markets. One organism at the center of both.
VIII. Superstructures: Who Wedged Themselves Between the Bucket and the Farm
In parallel with disposers in apartments, another collection point appeared — and it turned out to be smarter than it looked.
Large retail chains — first in Thailand, Singapore, South Korea — launched a simple scheme: you hand in organics in standard bags at a terminal by the entrance, you get loyalty points.
For retail, the logic was not in organics. The logic was in data.
The disposer knew what you threw away — meaning it knew what you bought, what you ate, and what you didn't finish. More honest than any consumer behavior survey. This district throws away a lot of unfinished avocado — the package size is too big. This neighborhood eats a lot of fish — expand the fish department. The consumer paid with trash. Retail got analytics that it used to pay Nielsen for.
By 2033, a cooperative model emerged that everyone was talking about. Thousands of district households united through an app, collectively invested in a district rendering hub and a stake in a BSF farm, received dividends proportional to the volume of substrate handed in.
A city resident throwing out organics became a co-owner of protein production. His trash — his asset. People sorted not out of ecological responsibility. A solar co-op — only with biorobots.
IX. Three Waves and Who Survived
All of this had been gone through before. Taxis after Uber. Hotels after Airbnb. BSF went the same way — only faster.
2026–2028: Contracts matter more than technology. The key resource is not the colony management algorithm. Long-term agreements with hubs, chains, municipalities. Whoever first staked out the substrate base — won the first wave. Not the smartest won. The most deal-capable. This irritated the tech optimists.
2028–2031: Price war. Thirty large plants operate simultaneously — BSF meal becomes an exchange commodity. Futures first on Asian exchanges, then on the CME. Margins fell from 40–50% to 15–20%. Those who had AI colony management already built into their operating model survived — not those who were only starting to implement it. A cost advantage of 18–25%. In a price war, that's literally everything.
2031–2035: Vertical integration. Whoever made the meal didn't win. Whoever controlled the chain from the disposer to the food in the pet's bowl or to the shrimp in people's mouths won. Three models survived: up — the BSF producer builds its own pet food brand (margin 40–60% vs. 15–20% on meal, first in Thailand and Vietnam); down — control of the substrate chain through its own hubs and refineries; platform — SaaS for farm management, margin 70%+.
Unexpected competitors didn't surprise anyone — except those who weren't looking. Tyson, JBS, Marfrig already owned rendering infrastructure and had been investing in BSF since 2024 — they added one technological layer and entered via acquisitions. Utility companies controlled the flow of organics by default — when organics became an asset, they were first in line for the margin. Thai Union entered as a co-owner, not a buyer.
X. This Is Not Agriculture
The market values BSF as a niche agricultural segment. Wrong category.
The right one — the operating system of the urban protein cycle. Control over the flow of the megacity's organic energy. Data on the real consumption of billions of people. A substrate that feeds fish that feed people. Frass that fertilizes greenhouses. Fat in engines. Chitin in prebiotics.
In 1994, Amazon appeared — a bookstore on the internet. In 2006, AWS was launched — a by-product of its own infrastructure. Now AWS brings in more profit than all of retail. Nobody saw this in 1994. That doesn't mean it couldn't have been seen.
BSF is moving on the same trajectory. The real business is not in the meal.
Long leg: vertically integrated operators in Southeast Asia with substrate contracts and their own pet food brand. Companies building the platform layer for farm management. Aquaculture giants that entered as co-owners.
Short leg: traditional fishmeal. Wild capture. Everyone tied to them.