The Stack, the Code, and the Yard – three signals converging in 2026
This month, in London, the IMO’s Maritime Safety Committee adopts the non-mandatory MASS Code, opening the regulatory pathway for commercial autonomous shipping and the three-year Experience-Building Phase that will inform the mandatory Code due in 2030. Later this year, in Austin, a startup will begin cutting steel for a shipyard designed from the ground up around robotics, software, and AI — with the stated goal of taking American shipbuilding output from five large ships a year to five hundred. Two events, opposite sides of the Atlantic. The same underlying shift.
Physical AI has reached maritime.
Physical AI is what artificial intelligence becomes when it leaves the screen and starts perceiving, deciding, and acting in the physical world. The technology stack — world foundation models, vision-language-action systems, large-scale simulation environments — has been deployed at commercial scale in autonomous vehicles, warehouse logistics, defense, and industrial robotics for several years. None of it was built for maritime. The cumulative R&D investment, measured in tens of billions of dollars, was financed by industries whose revenue scale and reinvestment cycles maritime cannot replicate.
The development cost of this technology has already been borne by other industries. Maritime inherits the result without funding its creation. That distinction defines the strategic window now opening for the sector.
Why maritime was late, and why that posture has been rational
The last innovation that fundamentally changed how the industry operates was the shipping container. Malcom McLean’s SS Ideal-X, 1956. In the seventy years since, maritime has optimized inside the container paradigm without breaking out of it. Vessels got bigger, terminals got more automated, and software got better. The operating logic of moving boxes from ship to shore to truck has remained intact.
This conservatism has been rational. A ship built today will operate for twenty-five years. Failures at sea carry physical consequences to hull, cargo, crew, and coastline that no software rollback can address. Regulation crosses every flag state and port jurisdiction. Insurance is priced on slow-moving actuarial tables. Capital is patient by necessity. The industry’s hesitation reflects the structural cost of being wrong, which exceeds nearly every other sector.
That calculus is now breaking. Necessity is the mother of invention, and the necessities have arrived simultaneously: capital that demands verifiable returns instead of sustainability narratives, climate volatility that has moved from forecast to operating condition, conflict that is weaponizing trade routes in real time, and constraints — regulatory, resource, workforce — that no longer permit incremental response. The Strait of Hormuz is closed as I write this. The Panama Canal is contested. Spending on port infrastructure will rise by more than a third to $90 billion annually by 2035. We now operate inside the decade of disruption described in earlier editions of my LinkedIn newsletter.
Physical AI arrives at this moment from the direction the industry needed: pre-built, pre-funded, and operationally validated in adjacent sectors. Maritime does not need to invent the technology. The remaining work is identifying where adoption creates competitive advantage and where it produces expensive duplication of capacity that will not earn its return.
What is new in 2026
Autonomous shipping research has been visible for more than a decade. Saildrone has operated unmanned ocean-going vessels since 2014. The Mayflower made an unmanned transatlantic crossing in 2022. Defense and oceanographic programs have run autonomous platforms at sea for longer than that. None of this is novel.
What is novel is that three things have converged in the same year for the first time.
The underlying Physical AI stack has matured to commercial grade. World foundation models such as NVIDIA’s Cosmos give machines a learned understanding of how objects move and how environments change. Simulation environments such as NVIDIA’s Isaac Sim let autonomous systems train against billions of synthetic scenarios before touching the physical world — a welding robot can practice ten thousand hull seams before it sees steel; an autonomous tug can encounter every weather pattern the North Atlantic produces without a real near-miss. Vision-language-action models translate human instructions into machine behavior. Earlier autonomous vessels could not use any of this, because none of it existed.
The regulatory pathway has opened. The non-mandatory MASS Code being adopted this month, followed by the Experience-Building Phase running through 2028, gives flag states, classification societies, and operators the structured environment they need to commit capital. Until now, the legal status of commercial autonomous operations remained unresolved. The MASS Code adoption converts that uncertainty into a structured operational pathway.
The production layer has reached comparable maturity. This dimension has received less attention than the regulatory and operational ones, but it carries equivalent weight.
Two layers, one stack: production and operation converge
Maritime’s production side — shipyards, fabrication, repair — and operation side — ports, terminals, vessels at sea — used to run on different technologies, different vendors, different talent pools. They were two industries connected by an asset they each handled. Physical AI is rewriting that distinction.
Valstad, the Austin-based shipbuilder cutting steel later this year, describes itself as “building the machine that builds the ships.” The system is best characterized as an end-to-end autonomous production architecture: feed in a 3D model, get back a complete build sequence — every cut, every weld, every robotic motion, every inspection point. Phased array laser welding runs five to six times faster than traditional submerged-arc or MIG welding for long seams, with consumable use reduced by up to 90%. Modular robotic cells handle stiffener attachment, micropanel fabrication, and structural welding. Precision-built kits are transported by truck and assembled at customer facilities, dispersing production capacity across the country instead of concentrating it in a few legacy yards.
Valstad’s founder, Dustin Walper, has made a sharper point worth integrating into how we understand this transformation. The assumption that an autonomous shipyard means welding automation is incorrect. The cycle-time gains come from automated handling of heavy components — cranes, conveyors, autonomous mobile robots, heavy-payload manipulators. Welding is the most visible element of the system. The coordinated handling of large physical objects across the yard is where productivity concentrates.
That handling runs on the same locomotion and perception stack that governs autonomous straddle carriers in the most advanced container terminals — the same stack that governs the next generation of commercial autonomous vessels. A shipyard, a port, and an autonomous ship are three deployments of one underlying technology with different mechanical interfaces. Skills transfer. Vendors compete across all three. The talent pool is shared.
Industries that were once defined by which technologies they could afford to develop are increasingly defined by which technologies they can afford to integrate. Maritime has paid for almost none of the underlying R&D and is now positioned to integrate at speed. The decisive capability is integration across both production and operation — shipyards on one side, terminals and vessels on the other — within the same investment cycle. Legacy scale does not necessarily predict that capability.
The four forces converging in 2026
The technology has been viable for two to three years, with both capital and talent available throughout that period. Adoption has accelerated in 2026, not earlier, and the reasons are worth setting out directly.
Four forces have converged that did not previously exist together.
Capital is no longer abundant. Infrastructure investors are demanding verifiable returns; sustainability narratives without P&L impact no longer attract funding. The patient capital that financed maritime decarbonization on faith has been replaced by capital that requires a defensible thesis on how a port, a shipyard, or a fleet generates differentiated returns. Physical AI provides such a thesis: automation produces operating cost reductions that finance teams can verify in quarterly reporting.
Climate has shifted from forecast to operating condition. The IMO 2030 trajectory is binding, insurance is repricing routes, and cargo owners are demanding emissions transparency at the manifest level. Every ton of fuel saved by autonomous routing, and every hour of dwell time eliminated at AI-driven terminals, carries measurable financial value.
Conflict is rerouting trade. The Strait of Hormuz is closed. The Panama Canal is contested between American and Chinese commercial interests. Chinese firms now operate or hold stakes in at least 129 ports outside China and have spent more than $80 billion on port construction abroad. Total non-Chinese spending on port infrastructure will rise by more than a third to $90 billion annually by 2035 — a figure that exists primarily because Western governments, infrastructure funds, and shipping lines have decided that geographic redundancy has become a strategic necessity.
Constraints have moved from compliance burden to operating condition. Regulatory, workforce, and resource constraints that previously functioned as background concerns now operate as foreground variables. The largest cost in maritime operations today is the inability to act quickly.
Each of these forces would individually push the industry toward Physical AI. In combination, they make non-adoption the higher-risk position.
This completes the investor shift that started two years ago
Some months ago, in this newsletter, I described why maritime investors had stopped funding dashboards and started funding hardware. The shift was visible: pure software ventures were being declined while a hull-cleaning robotics startup closed a $52 million round. Wind-assist hardware, hydrogen fuel cell consortia, port automation — capital had moved to hardware-first solutions because these address regulatory compliance and P&L impact simultaneously, with results that survive financial scrutiny.
Physical AI is the next layer of that shift. It embeds adaptive intelligence into hardware that until recently operated on static parameters. Wind-assist systems now adjust trim continuously against real-time weather data. Robotic welders accumulate learning across thousands of hull seams. Autonomous tugs reroute around dynamic obstacles based on conditions encountered in operation. The economic case for embedded intelligence has moved from speculative to verifiable in the same financial register that drove the original shift from software to hardware.
The investor logic that drove the dashboard-to-hardware shift now drives the hardware-to-Physical-AI shift. Compute and model costs have fallen by an order of magnitude in the last two years. The maintenance argument for adaptive hardware over static hardware is in the same financial register as the original compliance argument for hardware over software. CFOs can verify it. Insurance can price it. Cargo owners can demand it.
For portfolio construction, this implies that any maritime asset deployed today without a Physical AI integration roadmap is depreciating relative to assets that have one. The relevant time horizon is five years.
The strategic stakes for the $90 billion build-out
The global port build-out is now visible at scale. BlackRock and MSC are consolidating Western terminal operations. Stonepeak and CMA-CGM have formed a $10 billion joint venture, United Ports. India is in the middle of a port-building program scheduled to run until 2047. Saudi Arabia has signed a $450 million deal for Jeddah Islamic Port. Singapore is building a $20 billion automated port and shipping hub. Capital is being deployed at an unprecedented scale.
This capital is also being deployed inefficiently. Recent analysis warns that the rush to build port infrastructure will result in significant duplication and disappointing returns. Tanger Med has already taken volume from Algeciras. India’s ports risk cannibalizing Singapore and Salalah — and each other. The duplication that geopolitical anxiety is creating cannot all be commercially justified.
Physical AI matters most at this point in the build-out cycle. The next decade of port competition will be decided by differentiated operating capability — vessel turnaround time, energy efficiency per ton handled, weather-adjusted reliability, and integration with adjacent infrastructure. Physical capacity alone no longer determines competitive position.
Western ports cannot match Chinese capital deployment on physical infrastructure. That competition has already been decided. The 129 ports outside China where Chinese firms operate or hold stakes represent a buildout that took two decades and that no Western infrastructure program will replicate at the same cost basis. What Western ports and their infrastructure investors can do is compete on intelligence. Physical AI provides the substrate on which intelligence-based competition is built.
A port that integrates Physical AI early differentiates itself as a non-commodity asset. Ports that do not find themselves in direct price competition with structurally cheaper alternatives.
From inheritance to advantage
The technology stack is available, the capital is moving, and the strategic case is clear.
Adoption nevertheless presents structural challenges. Inheriting Physical AI requires aligning port authorities, vessel operators, regulators, insurers, classification societies, and labor — none of which any single entity controls. The technology has been developed elsewhere. The integration work belongs to maritime alone.
The integration challenge does not alter the underlying point. Maritime has access to a technology stack representing tens of billions of dollars in cumulative R&D, funded by adjacent industries. The decision is whether to lead the integration or be repositioned by it.
Maritime has spent seven decades adopting innovation last. That posture will not be rewarded over the next decade. Physical AI marks the inflection point at which industry conservatism becomes a net cost.
The technology exists. The remaining question is who inherits it first.
— Beatriz
Dr. Beatriz Canamary
Founder & CEO, SuRe Strategy Group | Next Wave Systems
Strategic Advisory for Infrastructure Investors and Operators — Maritime, Ports & Energy
Comments are closed