Cohere and Hanwha Bring Generative AI Into Shipbuilding

Summary: At a Toronto forum on January 26, 2026, Hanwha Ocean, Hanwha Systems, and Cohere signed an MOU to develop enterprise LLM and multimodal AI for submarine design, production planning, and manufacturing. The work supports Hanwha’s bid for Canada’s Patrol Submarine Project and fits Canada’s Buy Canadian procurement climate. The article says success depends on the bid outcome and whether the engineering-specification pilot actually runs.

January 26, 2026, local time: at a Korea-Canada industrial cooperation forum staged in Toronto, South Korea’s Hanwha group put signatures on memoranda of understanding with five Canadian companies. The agreement with the greatest bearing on the AI industry is the three-way deal among Hanwha Ocean, Hanwha Systems, and Canadian AI company Cohere: together they will develop AI built on Cohere’s enterprise large language models and multimodal models, aimed at submarine design, production planning, and manufacturing.

A concept-demo press release this is not. Holding it up is Canada’s Patrol Submarine Project (CPSP), a procurement contest in which the Korean and German shipbuilding camps go head to head — and “AI capability” has, for the first time, been written into a shipbuilding coalition’s bid. For Cohere, the deal is a landmark instance of its “secure enterprise AI” route finding a way into a defense supply chain.

Five MOUs in a Single Day: The Toronto Scene

A South Korean government and industry delegation was on hand, and Hanwha came away with five MOUs signed:

Hanwha Ocean and Algoma Steel: a new steel mill in Canada, steel supply for next-generation submarines, and additional maintenance, repair, and overhaul capacity — with a plan for Hanwha Ocean to invest up to $250 million
Hanwha Ocean, Hanwha Systems, and Cohere: AI models for submarine production planning, design, and manufacturing, to be developed as a joint effort
Hanwha Systems and Telesat: collaboration on low earth orbit satellite networks; Telesat’s goal is to launch 198 LEO satellites by year’s end, and the two will develop a new military LEO communications system for Korea
Hanwha Systems and MDA Space: combined strengths around MDA’s Aurora software-defined satellite platform
Hanwha Systems and PV Labs: pushing electro-optical and infrared sensor technologies forward for future defense programs

What all five deals share is Canada’s “Buy Canadian” procurement climate: technology, investment, and jobs bundled into the bid.

Where Cohere Fits: Models, Multimodal, and an Agent Platform

The split of duties is hard to miss: shipbuilding domain expertise is Hanwha Ocean’s contribution, defense systems integration falls to Hanwha Systems, and secure enterprise AI is what Cohere brings.

The technical reach is more concrete than you might expect: the models developed jointly cover large language models and large multimodal models alike, serving submarine operations and shipyard efficiency at once. Hanwha Systems CEO Jae-il Son described the work as reaching across the shipbuilding industry’s “design, production, and operations.” Andrew Chang, Cohere’s APAC vice president, said the partners will combine an “agentic AI platform and enterprise-grade AI models,” customize them for shipbuilding and defense, and deploy them into mission-critical applications that include the CPSP — pointing to real-time analysis of operational data and equipment diagnostics in deep-sea operations as examples. Canadian defence outlet Vanguard Defence reports that the pilot will apply enterprise-grade generative AI to produce engineering specifications from existing vessel data and new design inputs, so that specification documents are generated and validated automatically.

Cohere’s own profile explains why this pairing works. Headquartered jointly in Toronto and San Francisco, valued above $7 billion by The Korea Times’ reporting, and carrying Nvidia and Oracle among its investors, the company has spent years making its case with security-controlled, privately deployable enterprise models — the only kind of AI an industry this sensitive and this close to defense will seriously weigh.

The Actual Prize at Stake: Canada’s Patrol Submarine Project

The CPSP is there to hand the Royal Canadian Navy a successor to its aging Victoria-class fleet. Hanwha has brought its KSS-III boat into the contest, staking the bid on Arctic deployability and endurance, and has attached a delivery promise: sign in 2026 and the first four submarines arrive before 2035, with the whole twelve-boat fleet in place by 2043. Standing opposite is the German camp. Sung-chul Eoh, president of Hanwha Ocean’s Naval Ship Division, described the agreements as “a declaration of Hanwha’s commitment to becoming deeply integrated into Canada’s technology ecosystem.”

As bid strategy, the AI partnership is the latest chip pushed onto the table: the steel mill takes care of the supply chain, the satellites take care of communications, the sensors take care of mission systems — and the politically useful slot of “Canada’s own AI company” is where Cohere lands.

Where Vertical AI Lands Next: High-Value Hardware

For anyone following AI, the signal outweighs the contract. First, enterprise AI competition is drifting from generic assistants toward verticals, and shipbuilding — dense with documents, costly in labor, slow to validate — is exactly the ground where LLM ROI is easiest to demonstrate; automating and validating engineering specifications is the textbook high-value document workflow. Second, security and sovereignty claims are the ticket into regulated industries: through its private-deployment route, Cohere sets itself apart from cloud-first rivals in defense procurement. Third, a path is taking shape for agent platforms to enter physical industries: design documents, production planning, equipment diagnostics — shipbuilding offers far more measurable deployments than consumer apps.

Two outcomes will show whether it works: who takes the CPSP in 2026, and whether the pilot gets the specification workflow genuinely running.