■ MODERATE RISK ■ Manufacturing & Production
Robots are already laying the long, clean welds on flat hull sections, so that portion of the trade is going. The ship welder crawling into a double-bottom tank to fix what the robot can't reach will be employed for a long time yet.
“Automated hull welding is stronger and doesn't require underwater certification.”
Our AI replacement risk score — how we score jobs
Shipyard welding is not one job but a spread of them. New-construction work means joining massive steel plates and stiffeners into hull blocks — miles of weld per vessel, much of it repetitive fillet and butt welds in predictable positions. Repair work is the opposite: cutting out corroded plate on a 30-year-old bulk carrier, welding overhead in a ballast tank with condensation dripping on your arc, or going underwater in a drysuit to patch a hull without drydocking. Add fitting, gouging, inspection prep, and reading weld symbols off drawings that don't quite match the corroded reality in front of you.
The new-construction side is automating fast. Panel lines in modern yards run gantry welders and robotic arms that produce consistent, x-ray-clean seams on flat and gently curved sections, and offline programming keeps getting better at handling the curvier blocks. That's where our risk score of 40 comes from: the high-volume, repeatable welds — the bread and butter of production yards — are increasingly machine work. Korean and Japanese yards have been pushing this for years, and it spreads with every new panel line installed.
What resists is geometry and grime. Robots need access, fixturing, and predictability; a ship offers confined tanks, out-of-position joints, distorted repair surfaces, and one-off fixes where programming a robot takes longer than just welding it. Underwater wet welding and drydock repair are essentially robot-proof for the foreseeable future. The trade is shifting from 'person who welds seams' to 'person who handles everything the machines can't, and sometimes runs the machines.' Certified welders with repair, tank, and underwater tickets will stay busy; production welders doing flat work on a panel line are the ones being designed out.
Automatability: our editorial assessment of current and near-term AI capability
Production welding in large new-build yards is automating now and through the 2020s — that's where headcount shrinks first. Repair yards, drydocks, and underwater work face little near-term automation. By the mid-2030s expect the trade to be visibly split: fewer manual production welders, steady demand for repair specialists, and a growing tier of welder-operators programming and supervising the machines.
No, but it's narrowing. Automated panel lines are absorbing the repetitive new-construction welds, so production-only welders face shrinking demand. Repair welding, confined-space work, and underwater welding remain thoroughly human and are, if anything, short of qualified people. The trade rewards specialization now more than volume.
Not practically. Robotic welders need access, stable fixturing, and predictable geometry. A ballast tank offers a manhole-sized entry, awkward angles, corrosion, and surfaces that never match the drawing. Crawler and orbital systems handle some pipe and panel work, but the cramped, improvised repairs stay manual for the foreseeable future.
Three things pay off: repair-oriented skills (out-of-position, confined-space, underwater certs), robot operation and offline programming for mechanized systems, and inspection qualifications. A welder who can run a gantry line, fix what it botches, and sign off on quality is more valuable than one who can only lay wire — and much harder to replace.
On flat, well-fitted production joints, yes — machines hold travel speed and arc parameters more consistently than any human shift. On distorted repair surfaces, bad fit-ups, and awkward positions, an experienced welder still beats the machine, because half the job there is adapting to conditions the robot was never programmed for.