AI Aquaculture Tech in 2026: ReelData, AKVA, Optoscale and the Norwegian Salmon Stack

Aquaculture is the rarest type of AI vertical — one where the production deployments are deeply mature in a small geography (the Norwegian and Chilean salmon belts) and almost non-existent elsewhere. The economics of intensive salmon farming, the regulatory pressure from Norwegian authorities on welfare and environmental indicators, and the unusual willingness of an industry dominated by a handful of vertically integrated players to fund applied research have produced a stack that genuinely runs in production. This post walks through what that stack looks like and where the technology is heading.

The Norwegian salmon stack#

Norway produces over half of the world’s farmed Atlantic salmon, with Mowi, SalMar, Leroy Seafood and Cermaq dominating the producer side. The four of them have collectively funded much of the applied research that produced the modern aquaculture AI vendor base, and they remain the most demanding customers. The standard production stack in 2026 includes underwater cameras with edge inference for biomass estimation and behaviour monitoring, hydroacoustic sensors for fish position and feed-pellet tracking, environmental sensors for oxygen, temperature, salinity and current, and increasingly sea-lice counting cameras that have replaced the manual counting protocols Norwegian authorities historically required.

Optoscale, founded out of NTNU in Trondheim, is the dominant biomass-estimation vendor — stereoscopic underwater cameras with computer vision that estimate individual fish weights to within a few percent of physical sampling. The system is deployed at most of Mowi and SalMar’s Norwegian sites by 2026 and an expanding footprint in Chile, Scotland and the Faroe Islands. The economic case is compelling because biomass estimation drives feed decisions, harvest timing and stocking density, and the historical alternative was infrequent physical sampling that introduced both stress and error.

ReelData and the land-based pivot#

ReelData, the Canadian aquaculture AI company, has positioned around the land-based recirculating-aquaculture-system (RAS) growth that accelerated through 2023-2025 as several large land-based producers came online in North America, the Middle East and Asia. Its ReelAppetite, ReelBiomass and ReelWelfare products target the specific operational questions that RAS sites face — feed control in tanks where ad-libitum feeding generates rapid water-quality problems, biomass estimation in turbid recirculating water, and welfare monitoring at higher stocking densities. The land-based salmon story has had a rocky 2023-2026 with several high-profile project failures, but the operators who survived have increasingly standardised on ReelData and a small number of competitors.

Stingray Marine Solutions, the Norwegian sea-lice laser-treatment company, sits in the unusual position of being both a hardware vendor (the underwater laser nodes that target and neutralise sea lice on individual salmon) and a computer-vision vendor (the detection model that identifies lice on fish swimming past). The combined system has displaced a meaningful share of chemical and bath-treatment sea-lice protocols at Norwegian sites and is now expanding in Chile and Canada.

AKVA Group and the platform layer#

AKVA Group remains the largest aquaculture equipment supplier and the dominant platform vendor through its Fishtalk and AKVAconnect products. The platform sits underneath most of the third-party AI vendors at Norwegian sites — feed barge integration, environmental sensor aggregation, regulatory reporting to the Mattilsynet (the Norwegian Food Safety Authority) and operational dashboards. The company’s transition from a pure-hardware vendor to a hardware-plus-software platform vendor has played out gradually since the early 2020s and is reasonably complete by 2026.

Innovaqua Salmon, the Spanish-Norwegian aquaculture technology consortium, has produced research outputs that several commercial products are built on, particularly around acoustic-telemetry-driven behaviour analytics and oxygen-management AI. The earlier Alphabet X Tidal project, which targeted aquaculture monitoring with underwater computer vision, was wound down as an internal X project but the underlying technology has continued through partnerships and licensing.

Environmental monitoring and the regulatory perimeter#

Norwegian and Chilean salmon farming both operate under regulatory regimes that constrain stocking density, sea-lice levels, escape events and benthic environmental impact. The trafikklys (traffic-light) system that governs Norwegian production capacity is based on regional sea-lice indicators, and the data feeding the trafikklys assessments comes increasingly from automated counting rather than manual sampling. Chilean Sernapesca regulation operates on a different but conceptually similar basis, with mortality reporting, antibiotic-use disclosure and environmental impact assessments increasingly tied to sensor-based data.

The environmental AI category — predicting harmful algal blooms, low-oxygen events, jellyfish incursions and current-driven escape risks — has accelerated through 2023-2026. Several Norwegian producers run combined oceanographic and ML systems that integrate satellite ocean-colour data, in-situ sensor networks, regional ocean models from MET Norway and the Norwegian Institute of Marine Research, and site-specific historical records. The harmful algal bloom predictions in particular have measurable economic value because a single bloom event can wipe out a generation at a site.

Why aquaculture AI works where other ag verticals stall#

The recurring question is why aquaculture has produced such a complete production AI stack when terrestrial livestock and crop verticals have moved more slowly. Three factors recur in the answer. First, the economic concentration — a few large vertically integrated producers can fund applied research and adopt new technology without the coordination problems that fragmented terrestrial agriculture faces. Second, the regulatory pressure — Norwegian and Chilean authorities have actively required documentable environmental and welfare monitoring, which created demand. Third, the physical environment — salmon in net pens are easier to monitor with fixed cameras and sensors than range cattle or open-field crops, and the value per unit is high enough to justify the sensor investment.

Tilapia, shrimp and other species lag significantly. Shrimp farming in Southeast Asia has begun to adopt water-quality sensing and pond-management AI from companies like XpertSea and JALA but the per-pond economics are different and the deployments are less integrated. Tilapia is similar. The salmon model is unlikely to transfer directly.

Where the technology still struggles#

Welfare scoring beyond sea lice and gross morbidity remains an open area. The industry has invested in computer-vision-driven welfare indicators — gill condition, body integrity, swimming behaviour — but the science of what counts as a reliable welfare indicator is less settled than producers and NGOs would prefer. Escape prediction — anticipating net failures before they happen — is a similarly open area where ML approaches show promise but production deployments are limited.

Where pdpspectra fits#

Our data engineering practice helps aquaculture operators build the multi-vendor sensor, camera and regulatory reporting platform that determines whether site-level AI investments scale across a portfolio. We work alongside AKVA, Optoscale and ReelData deployments.

Related reading: AI in agriculture and precision farming, maritime shipping data platforms, and maritime port operations in 2026.

---

Aquaculture AI in 2026 is the most production-mature applied AI in any agricultural vertical. Talk to our team about your aquaculture data platform.