Blue Food Systems

Established In 2022 farmed aquatic animals outproduced wild capture for the first time, and that single crossing is the whole subject. The Food and Agriculture Organization’s 2024 assessment put aquaculture of aquatic animals at about 94 million tonnes against roughly 91 million tonnes landed by capture fisheries, inside a total of about 223 million tonnes once algae are counted. The aquatic food system is therefore now a farming system. It has a farming system’s problems — feed, disease, genetics, effluent, welfare, capital — and it is still supervised, almost everywhere, by agencies built to manage hunting.

Frontier The brief’s claim is that tonnage is not the constraint and has not been for a decade. Two things bind: the long-chain omega-3 lipids carnivorous species need, which come overwhelmingly from a small set of reduction fisheries, and disease and parasite load in open water, which sets mortality, drug use and social licence. The intensively engineered alternative — land-based recirculating aquaculture — has now produced enough commercial failures to be evidence rather than promise, and reading that failure record honestly is the most useful thing this brief can do.

Established Scope boundary. Future Agriculture owns the terrestrial food system, including the controlled-environment sector whose bankruptcies this brief borrows as an analogy. Ocean Engineering owns structures, moorings and the seabed regime that offshore farms depend on. Synthetic Ecosystems owns closure — what it costs to run a nutrient loop with no outside — which is the theoretical limit case of a recirculating system. Planetary Stewardship owns the boundary framing for nutrient and biosphere pressure. What is left, and what this brief owns, is the farm: the animal, its feed, its pathogens, its capital cost and the regulator that cannot decide whether it is livestock or wildlife.

Frontier Precision aquaculture is real as sensing and unaudited as economics. Machine vision counts sea lice, estimates biomass and controls feeding on commercial farms today, and operators report improvements in feed conversion and treatment timing. Almost none of those figures come from an independent trial with a pre-specified endpoint. The gap between a working sensor and a demonstrated production gain is the one that has swallowed precision agriculture for thirty years.

Established A note on sourcing. This brief was commissioned in September 2026 from the Institute’s research base. Reading-list entries without links are cited from the bibliographic record rather than re-fetched, and claims are dated no later than early 2026 unless carried by a linked source.

Established The production record is concentrated, not global. Growth over three decades has been overwhelmingly Asian, with China alone around half of world output, while Africa — where the nutritional case is strongest — contributes a few per cent. Apparent consumption of aquatic animal foods runs around 20 to 21 kg per person per year and is rising faster than population. Any sentence about “the aquaculture sector” that does not name a country and a species is describing an average of things with nothing in common.

Established Wild stock status has a bimodal answer that the headline figure hides. The intergovernmental assessment puts somewhere near 38 per cent of assessed stocks fished beyond biologically sustainable levels, a share that has worsened slowly for four decades. The informative result is the divergence underneath it: where stock assessment, enforced quotas and a science-to-management pathway exist, biomass has frequently rebuilt; where they do not, the trend runs the other way. Overfishing is a governance variable with a fisheries covering, which is why this brief treats capture as an institutional subject and farming as a technical one.

Established The environmental-performance comparison across blue foods was done properly and its ordering is stable. The multi-paper Blue Food Assessment work in 2021 put farmed bivalves and seaweed at the bottom of the stressor ranking for greenhouse gases, nitrogen, phosphorus, land and water, small pelagic capture fisheries close behind, and farmed salmonids and carp broadly comparable to industrial chicken. The dominant term for every fed species is the feed, not the farm. That single result reorganises the subject: improving aquaculture mostly means improving what goes into the bag.

Established Fishmeal has been substituted and fish oil has not, and the difference is chemistry rather than effort. Protein is fungible: soy protein concentrate, wheat gluten, rapeseed and poultry by-product have driven fishmeal from around forty-five per cent of a salmon diet in the early 1990s to roughly a tenth today. The long-chain omega-3 fatty acids EPA and DHA are not fungible; the fish do not synthesise them efficiently, consumers buy the fillet partly for them, and the world supply comes from a few reduction fisheries whose catch has been flat or falling for decades. The binding constraint on carnivorous aquaculture is a lipid, not a protein.

Frontier Two routes out of the lipid constraint exist at commercial scale and both are contested on cost rather than on function. Heterotrophic marine algae grown in fermenters produce an oil rich in EPA and DHA and have been sold into salmon feed since the late 2010s; transgenic oilseeds engineered with algal desaturase and elongase genes — camelina in the research line, canola in the commercial one — produce a long-chain omega-3 oil approved in several jurisdictions. Neither has displaced fish oil at the margin, and the reason is price per kilogram of EPA and DHA delivered. The adversarial cost analysis available for bioreactor-grown protein is the right template for reading any claim here: published scale-up economics for cultured animal cells land at tens of dollars per kilogram of wet cell mass, driven by amino-acid and media costs that a fermenter cannot design away.

Frontier The insect and single-cell protein ingredients have announced far more capacity than they have delivered. Black soldier fly meal and methanotroph-derived single-cell protein are technically sound, approved for aquafeed in major markets, and operating at volumes that are a rounding error against the tens of millions of tonnes of aquafeed produced annually. The sector’s largest European insect-protein company entered insolvency proceedings in 2025, the kind of fact a technology assessment should carry and a market forecast usually does not. Neither delivers EPA or DHA, so neither answers the lipid problem.

Established The single largest measured win in the history of aquaculture is vaccination, and it is rarely quoted. Norwegian salmon farming reduced therapeutic antibiotic use by well over ninety-nine per cent from its late-1980s peak while production grew by an order of magnitude, on the back of effective injectable vaccines against the dominant bacterial diseases. Chile, facing a different pathogen mix dominated by piscirickettsiosis, has never achieved the same reduction and remains a heavy user by comparison. The contrast is the cleanest natural experiment in the field: the same species, the same technology base, two pathogen environments, two completely different drug-use outcomes. Post-Antibiotic Medicine owns the resistance consequences.

Established Sea lice and mortality are the industry’s own published indictment. Norway runs a production-area traffic-light system that regulates farm capacity on measured lice-driven mortality in wild salmonids, with a limit around half an adult female louse per farmed fish in the spring window. The management cost is large — the standard estimate puts sea lice at several per cent of production value — and the control measures have their own welfare costs: thermal and mechanical delousing injure fish, and cleaner-fish stocking has mortality high enough to have become an ethical issue in its own right. Sea-phase mortality of farmed Atlantic salmon in Norway has run in the region of fifteen per cent a year, which at national scale is tens of millions of animals. No one disputes these figures, because the industry reports them.

Established The land-based alternative has a failure record, and the failures are informative. Recirculating aquaculture removes lice, escapes and benthic loading by construction, and substitutes a capital-intensive, energy-intensive, biologically brittle plant. The largest land-based Atlantic salmon venture in North America has repeatedly lost standing biomass to system events while accumulating heavy losses; two flagship North American projects were cancelled outright in 2024 and 2025 after permitting was substantially complete; the company that owned the first genetically engineered food animal halted construction and wound down. Controlled-environment agriculture ran this experiment first and more publicly: one leading indoor-farming company raised roughly half a billion dollars and shut down in November 2024 with plant disease among the causes, another went through Chapter 11 in July 2023 after a $166 million loss, and a third survived bankruptcy only by abandoning every site but one. Closed systems concentrate pathogens as efficiently as they exclude them.

Established Genetics has delivered quietly and the headline case failed commercially. Family-based and genomic selection has produced compounding gains in growth rate and feed conversion in Atlantic salmon, tilapia and shrimp, and marker-assisted selection against infectious pancreatic necrosis is a textbook case of a disease effectively removed from a farmed population by breeding. Japan approved genome-edited red sea bream and tiger puffer for sale in 2021 under a notification rather than approval pathway. Against that, the first genetically engineered food animal cleared for human consumption — a fast-growing Atlantic salmon authorised in the United States in 2015 and finally sold in 2021 — reached market and then lost its producer to economics, not to regulation. Approval was never the binding constraint; cost of production was.

Frontier Seaweed and bivalves are the lowest-impact blue foods and the hardest to make pay outside Asia. Farmed algae is a tens-of-millions-of-tonnes industry almost entirely concentrated in East and Southeast Asia, and Western attempts have repeatedly failed on harvest, processing and market price rather than on biology; a dedicated United States macroalgae programme in the late 2010s funded a portfolio of designs and did not produce a commercially self-sustaining cost. The biological case for seaweed as a resilient calorie and nutrient source is real and is usually stated without any supply-chain claim attached, which is precisely where the difficulty lives. Bivalves are the strongest environmental story in the entire food system and depend on hatchery and wild seed supply that ocean acidification has already disrupted in at least one documented regional case.

Frontier Can EPA and DHA be delivered at feed-grade cost without a marine input? This is the field’s central open question and it is an economic one with a biological floor. Fermented algae and transgenic oilseeds both work; whether either can reach the cost per kilogram of omega-3 that a commodity feed formulator will pay, at the volumes a growing carnivorous aquaculture requires, has not been demonstrated and the published cost data are mostly from interested parties.

Frontier Is there a vaccine or a resistant line for sea lice? Decades of effort have produced no commercially effective sea lice vaccine, unusual for a parasite this economically important, and the reason — an ectoparasite exposed to a weak mucosal immune interface — is understood well enough to be discouraging. Genetic resistance appears heritable but modestly so. If either worked, most of the argument about open net pens would change shape.

Frontier Does sensing translate into production? The measurement case is settled: cameras count lice and estimate biomass better than manual sampling, and acoustic and environmental sensing detect feeding behaviour and oxygen events. Whether closed-loop control on those signals improves feed conversion, survival or welfare by an amount that pays for the system is claimed by vendors and has not been established by independent trial.

Frontier How much offshore capacity is real? Exposed-site farming has demonstration structures in Norway and China and a plausible physical rationale — deeper water, stronger flushing, lower lice pressure — against higher capital cost, harder service access and unresolved insurance. Ocean Engineering owns the structural question; what is unresolved here is whether the biological benefit survives contact with the operating cost.

Established The first bottleneck is the lipid, and it is a hard supply limit rather than a cost curve. Global fish oil output is bounded by reduction fisheries already fully exploited and strongly affected by El Niño events in the Peruvian anchoveta fishery, which alone can swing world supply by a large fraction in a single year. No amount of demand converts into more fish oil. Substitution is the only route, and substitution means paying for synthesised or engineered lipid.

Established The second is that closed systems concentrate biological risk as efficiently as they exclude it. A recirculating plant depends continuously on biofiltration, oxygenation, degassing and water treatment; hydrogen sulfide events, off-flavour compounds from geosmin-producing bacteria, and early sexual maturation in fresh water are the recurring technical failure modes, and each has taken out commercial-scale standing biomass. Redundancy costs capital, and capital is what the business case has least of.

Established The third is energy, and the analogy from land is exact. A recirculating plant pays continuously for pumping, oxygen, temperature control and water treatment, where a net pen pays for none of them. The published life-cycle comparison between vertical farming and greenhouse lettuce — roughly 10 to 95 megajoules per kilogram against 120 to 133 — is instructive not for its numbers but for its structure: the controlled system wins or loses on where its energy comes from and whether waste heat is available, and nobody sites salmon plants on that basis.

Frontier The fourth is seed supply. Several of the highest-volume farmed species still depend partly or wholly on wild-caught juveniles or broodstock — most acutely in shrimp, eel and several marine finfish — which converts a farming industry into a harvesting one at its most vulnerable point. Closing the life cycle in the hatchery is the single highest-value unglamorous research programme in the sector, and it is done species by species with no general method.

Established This topic depends on terrestrial agriculture for most of its inputs. Soy, wheat, rapeseed and their protein concentrates are aquafeed. Every constraint Future Agriculture documents on yield, nitrogen and land is transmitted directly into the cost of a farmed fish, and the transgenic omega-3 oilseed route makes the dependency explicit: the lipid problem of the ocean is being solved on arable land.

Established It depends on veterinary biologics and on the regulatory pathway for them. The Norwegian vaccination result was an animal-health achievement delivered through a licensing system that could approve fish vaccines quickly. Jurisdictions without that pathway cannot reproduce the result regardless of the science, which is why the antibiotic-use gap between producing countries persists.

Established It depends on ocean condition, which is changing faster than farm siting decisions. Marine heatwaves, harmful algal blooms, jellyfish incursions and acidification all act on fixed installations with multi-year investment horizons. Compound Climate Hazards owns the joint distribution; a farm is a single-site bet against it.

Frontier The decisive result is an audited, multi-cohort, full-cycle cost of production from a commercial-scale land-based grow-out facility. Every important claim about the future shape of this industry — whether open net pens can be replaced, whether the lice argument can be engineered around, whether the capital intensity is transitional or structural — turns on a number that no operator is required to publish and none has published in auditable form across several complete cohorts. Investor presentations give target costs and bankruptcies give the other half of the distribution; nothing gives the mean with an error bar. Until that exists the land-based case is argued between brochures and obituaries.

Established Nothing new needs to be built to produce it. Facilities exist and have been operating for years. What is missing is a disclosure requirement, or a public financier willing to make audited cost reporting a condition of support, which is an institutional act rather than a scientific one.

Frontier The second experiment is a policy natural experiment already scheduled. Canada announced in 2024 a transition away from open net-pen salmon farming in British Columbia coastal waters with a stated end date later this decade and a draft transition plan. If it proceeds, it is the first time a significant producing jurisdiction has been required to move an established industry out of open water on a deadline, and it will generate the cost, employment, First Nations governance and production-displacement data that the entire argument currently lacks. If it is delayed or reversed, that is also a result, and a measurable one.

Frontier The third is an independent, pre-registered trial of closed-loop precision feeding. Two matched commercial sites, one running vendor control and one running standard practice, with feed conversion, survival and welfare scored by a third party over a full production cycle. The technology is deployed widely enough that the trial is a scheduling problem rather than a research problem, and its absence is the reason the sector’s productivity claims cannot currently be believed or dismissed.

Established The net pen is an extraordinarily cheap structure and that is the entire economic problem. It borrows buoyancy, water exchange, oxygenation and waste removal from the ocean for nothing. Every alternative architecture — land-based recirculation, closed floating bags, offshore submersibles — competes against a zero-cost utility, which is why engineering improvements keep failing to close a gap that is not primarily about engineering.

Established Recirculating plants are water-treatment plants with fish in them. The engineering content is biofiltration, solids removal, degassing, oxygen injection, ozonation or ultraviolet treatment, and instrumentation with alarm redundancy. The dominant design error in the failure record is treating fish husbandry as the hard part; the hard part is process control on a plant that cannot be shut down and restarted.

Frontier Offshore structures import a different discipline. Exposed-site farms are moored marine structures with fatigue, hydroelastic response and service-access problems that belong to offshore engineering rather than to fish farming, which is the subject Ocean Engineering treats directly. The operational question that decides viability is not survival in a storm; it is whether you can feed, inspect and harvest on a site that is inaccessible for days at a time.

Established Waste is a resource in principle and a permit problem in practice. Recirculating systems concentrate nitrogen and phosphorus into a stream that can be recovered, and the recovery chemistry is established — nitrogen and phosphorus recovery from organic waste is settled work, with the residual difficulty in the other fractions. Integrated multi-trophic aquaculture — seaweed and bivalves beside fed finfish — has decades of promising trials and almost no commercial adoption, because the co-products are worth less than the compliance cost of the permit change. Industrial Ecology owns the general form of this failure.

Established Terrestrial agriculture is the upstream system, not a parallel one. Future Agriculture carries the feed inputs, the transgenic oilseed route and the cautionary record of capital-intensive controlled-environment production that this brief reads across into land-based fish farming.

Established Ocean structures and the seabed regime are a separate discipline. Ocean Engineering owns moorings, hydroelastic response and the law of large non-ship structures, all of which bound offshore aquaculture before any biological argument is reached.

Frontier Closure is the theoretical limit of a recirculating system. Synthetic Ecosystems records what it actually costs to run a loop with no outside — complete air and water regeneration, food regeneration falling short, and an inedible-biomass deadlock that never closed. A recirculating fish plant is a partially closed loop bought at a known price, and the closed-system literature is the best available prior for what the last increments of closure cost.

Established Two neighbours own the consequences. Post-Antibiotic Medicine owns antimicrobial resistance, of which aquaculture is a significant and geographically concentrated contributor. Biodiversity Restoration owns the wild populations that escapes, introgression and lice pressure act on.

Established The defining institutional defect is that farmed aquatic animals are regulated as a fishery in most jurisdictions. Siting, effluent, animal health, food safety and welfare are typically split across a fisheries department, an environment department, a food agency and a veterinary authority, with no single competent regulator and, in several producing countries, no aquaculture statute at all. Canada is the clearest case: decades of expert recommendation for a federal aquaculture act, and no act.

Frontier The British Columbia transition is the most consequential aquaculture policy decision taken in a Western jurisdiction this decade, and its status is genuinely uncertain in 2026. A stated intention to end open net-pen salmon farming in British Columbia waters by the end of the decade, with a draft transition plan, sits against licence litigation, provincial and First Nations positions that do not all align, and a production base that cannot be relocated cheaply. This brief will not predict the outcome. What would confirm either reading is the regulatory record: licence conditions actually issued, and whether any replacement capacity is financed.

Established Welfare law for fish lags terrestrial livestock law by about thirty years. Slaughter regulation covering farmed fish is recent, partial and unevenly enforced; stocking density, handling and delousing procedures are mostly governed by industry codes; and the mortality figures the sector publishes would be a scandal in any terrestrial species. The direction of travel in European regulation is toward species-specific requirements, and the binding question is whether welfare obligations arrive before or after the industry has committed capital to particular production systems.

Established The welfare arithmetic is very large and rarely stated. Farmed finfish are counted in the tens of billions of individuals a year, more than all farmed terrestrial mammals and birds combined by headcount, and the sea-phase mortality rates the industry publishes imply animal deaths in the tens of millions annually in a single producing country. Whether fish sentience justifies weighting those lives is contested; that the numbers are this large is not.

Established Escapes and introgression are a documented harm with an unresolved remedy. Genetic material from farmed Atlantic salmon has been detected in a substantial share of surveyed wild Norwegian river populations. Sterility by triploidy or by genetic means would remove the mechanism and both are technically available; neither is mandated, because sterile fish carry production penalties no producer will accept unilaterally. This is a collective-action problem with a known technical solution, and the most tractable ethical problem in the brief.

Frontier Coastal communities and Indigenous governance are not a side constraint. In British Columbia, in Norway, in Chile and in Scotland, siting decisions are simultaneously employment decisions, title and rights decisions, and wild-fishery decisions, and the parties do not agree with each other within any single category. Framing the argument as environment against jobs misdescribes it; the actual disagreements run through who holds the decision right.

Established The nutritional stake is micronutrients, not calories. Aquatic foods supply a modest share of global protein and a disproportionate share of bioavailable iron, zinc, iodine, vitamin B12 and long-chain omega-3, in populations where those deficiencies are common and substitution is difficult. An increase in blue food availability in low-income coastal countries has a nutritional value its tonnage understates.

Frontier Blue foods are the most plausible route to lower-stressor animal protein at scale. If the ranking that puts bivalves, seaweed and small pelagics at the bottom of the stressor distribution holds up, then shifting a few percentage points of animal-protein demand toward them does more per unit of effort than most agricultural interventions. The obstacle is demand rather than supply: consumers in wealthy markets buy salmon and shrimp, which are the highest-stressor blue foods.

Speculative The strong version of the argument is that the ocean becomes a managed agricultural surface. Zoned, leased, monitored, insured and genetically improved, with wild capture reduced to a regulated remnant and most aquatic protein farmed. Every component of that already exists somewhere. What has never been tested is whether coastal societies will accept the property regime it implies, and there is no historical case of a maritime commons being converted at that scale without sustained conflict.

Handwave The claim that aquaculture will feed a growing world population is where the argument works by assertion. Production growth is real, but it is concentrated in countries and species that do not map onto the populations with the largest nutritional gaps, most high-value output moves through international trade toward wealthy consumers, and the feed inputs compete with human food. State the mechanism by which a farmed tonne reaches an undernourished person, or drop the claim.

These horizons track the feed constraint, the disease constraint and the capital question, which are the three variables that decide everything else.

• 10 yr: Frontier Non-marine omega-3 supply is a material fraction of feed lipid, or it is clear it will not be; the British Columbia transition has either happened or visibly failed; audited land-based production costs exist for at least one operator; genome-edited disease-resistant lines are in commercial production in at least one jurisdiction.
• 25 yr: Speculative Offshore and exposed-site farming is a routine production mode rather than a demonstration; a sea lice vaccine or a resistant line exists, or open net-pen farming has been substantially restricted in high-income jurisdictions; seaweed has a Western industry or has been conceded to Asia.
• 50 yr: Speculative Most aquatic animal protein is farmed from closed life cycles on engineered feeds, wild capture is a managed minority supply, and the coastal property regime has been renegotiated in the major producing regions.
• 100 / 250+ yr: Handwave Claims at this range depend on ocean state, on protein demand paths and on whether cellular or fermentation-derived seafood displaces farming entirely. Anyone giving a date is describing a preference.

• Depends on Three results on this map. From Future Agriculture: the arable inputs that are aquafeed, and the transgenic oilseed route that is the most advanced answer to the lipid constraint. From Ocean Engineering: the mooring, hydroelastic and service-access results that decide whether exposed-site farming is a production mode or a demonstration. From Synthetic Ecosystems: the measured cost of closure, which is the best available prior for what the last increments of recirculation cost.
• Requires (not on this map) Six things nobody on this map produces. An affordable non-marine source of EPA and DHA at feed-grade cost, because the lipid, not the protein, is what bounds carnivorous aquaculture. A field-effective sea lice vaccine or resistant line, because its absence is what forces every expensive alternative architecture. Audited multi-cohort cost disclosure from land-based grow-out, because the central economic claim of the sector is currently untestable. A single competent regulator for farmed aquatic animal welfare and siting, because a split mandate is why fish welfare law is thirty years behind. Closed-life-cycle hatchery seed for the species still taking wild juveniles, because a farm that harvests its own inputs is not yet a farm. And offshore mooring and service-vessel capacity for exposed sites, because the operating access problem, not storm survival, decides offshore viability.
• Enables A lower-stressor animal protein supply that Planetary Stewardship can actually count against a nutrient boundary; a nutrient-recovery stream that Industrial Ecology can close; and a reduction in fishing pressure that Biodiversity Restoration needs and cannot produce by protection alone.
• Adjacent Genetic Engineering owns the editing pathway whose first approved food animal was a fish; Post-Antibiotic Medicine owns the resistance consequences of the drug use this sector has partly eliminated; Compound Climate Hazards owns the marine heatwave and bloom correlations that fixed installations are exposed to.

Established “Farming fish just moves the problem, because you feed them wild fish.” This was true and is no longer the general case. Fishmeal inclusion in salmon diets has fallen by roughly three-quarters since the early 1990s, and the wild-fish input per unit of farmed output is below one on most published accountings for salmonids. The residual and legitimate version of the objection concerns fish oil specifically, and the equity of landing food-grade small pelagics for feed in regions where people eat them.

Established “Land-based farming solves the environmental problems of net pens.” It removes escapes, lice transfer and benthic loading, and it adds a large continuous energy demand, a concentrated waste stream requiring disposal or recovery, and a catastrophic-failure mode that has repeatedly destroyed entire standing biomasses. Whether the trade is favourable depends on the electricity supply and on a cost of production nobody has published in auditable form.

Frontier “Recirculating systems are proven; only financing is missing.” The technology works at hatchery and smolt scale, which is genuinely proven and commercially universal. Full grow-out to market size is a different engineering problem — larger biomass, longer residence, higher waste loading — and the commercial record at that scale includes several of the most expensive failures in modern aquaculture. Conflating the two scales is the sector’s most common error.

Established “Regulatory hostility killed genetically engineered fish.” The record says otherwise. A fast-growing engineered Atlantic salmon was authorised in the United States in 2015 and sold from 2021; Japan cleared genome-edited sea bream and puffer in 2021. What failed was the business, on the cost of land-based production and the scale of the market it could reach. Approval is a necessary condition that has now been met more than once, and it has not been sufficient.

Speculative “Seaweed will decarbonise agriculture, feed billions and absorb carbon.” Seaweed is a genuinely low-stressor crop with a real nutritional and material case, and the biological arguments for it are usually sound. The claims that do not survive contact with evidence are the supply-chain ones: harvest and processing cost outside Asia, the durability of any carbon stored, and market volumes that remain small in every Western jurisdiction that has tried. The biology is not where this fails.

Handwave “Precision aquaculture will raise productivity by a stated percentage.” Every figure of this kind currently in circulation comes from a vendor, a pilot without a control, or a modelling study parameterised on both. The sensing is real; the production claim is an assertion until somebody runs matched sites with a pre-specified endpoint and a third-party analyst. That trial would cost a fraction of what the sector spends annually on the equipment.