A3: Q2 Robot Orders Reach $622 Million as 56% Go to Nonautomotive Customers

The useful news in the Association for Advancing Automation’s Q2 data is the mix: 56% of robot units went to nonautomotive customers as automotive manufacturing softened. That’s a broader automation market, but it still falls well short of proving that welding-cell demand—or a 14-person fab shop’s payback case—is accelerating.

North American companies ordered 8,940 robots worth $622 million in the second quarter of 2026, A3 reported. Unit volume rose 4.3% from Q2 2025, while order value climbed 21.3%. The figures cover semiconductors and electronics, automotive components, food and consumer goods, metals, and life sciences, among other industries.

The growth is spreading across the customer base

The headline total matters less to a job shop than where those orders landed. A3’s figures show demand broadening into electronics and photonics, life sciences and pharmaceuticals, automotive suppliers, food and consumer goods, plastics and rubber, metals, and other industries. The Robot Report described that expansion as offsetting weaker investment from automotive OEMs, whose orders reportedly declined about 25% in the first half of 2026.

For integrators, that wider customer base could support steadier backlogs than an automotive-heavy market. It may also mean more deployment experience with mixed production requirements. The public figures don’t show whether those projects are repeatable short-run work or high-volume lines with a completely different fixture and programming burden.

Cobots have a footprint, but welding is still an open question

A3 counted 1,137 cobots in Q2, or 12.7% of all robot units. They represented $44 million, or 7.1% of the quarter’s revenue. That makes cobots a meaningful part of the market, while leaving conventional industrial robots in the clear majority.

The cobot number is a signal that lower-scale or more flexible automation is finding customers. It isn’t a welding number. The public data don’t identify how many cobots went into arc-welding cells, how much metals demand involved welding rather than machine tending or inspection, or what share of deployments used parts and fixtures resembling a short-run fab shop’s mix.

That distinction matters because robot demand can rise without improving the economics of a welder on a cart. A cell running repeatable brackets for one customer has a different utilization profile from a shop switching among 20- to 200-piece jobs, qualifying D1.1 work, and rebuilding fixtures between parts. For a useful look at why headline throughput claims need their baseline, see our earlier examination of a cobot throughput claim.

The ROI question is still weld-inches per shift

For a small fab shop, the Q2 data supports a market question rather than an investment decision: are integrators seeing enough welding work outside automotive to build and support cells at a sensible price? A3’s release doesn’t provide the numbers needed to answer it. There are no welding-cell payback periods, fixture costs, total cost of ownership, parts volumes, cycle times, rework rates, or arc-on-time benchmarks.

The practical calculation remains straightforward. Start with the fully loaded cell cost—including fixturing, positioners, safety equipment, programming, installation, training, maintenance, and warranty—and compare it with the annual contribution from incremental weld-inches. Count the labor actually displaced or redeployed, the setup time for each new part, and the jobs the cell can run when the best welders are already occupied. A $622 million market total cannot substitute for those inputs.

Questions to put to the integrator

The next useful vendor conversation should turn the broad market signal into shop-level evidence:

How many welding cells—not general robot projects—did you commission in the past year, and how many were in shops with short runs and varied fixtures?
What is the delivered price and lead time for the complete cell, including fixtures, positioners, safety hardware, programming, training, and spare parts?
What parts-per-batch, arc-on time, and weld-inches-per-shift assumptions support the payback model?
Who fixtures each new job, how long does programming and prove-out take, and what happens when the one programmer is unavailable?
Which D1.1 procedures and documentation have comparable customers qualified, and who owns the WPS and production records?
What warranty, service response, and resale data do you have for cells in this application?

A broader market, with the hard evidence still to come

Q2 shows automation demand expanding beyond automotive and gives cobots a tangible foothold. That could improve the ecosystem around smaller, more flexible cells if the growth reaches metals and welding applications. The data do not yet establish that it has, or that a shop’s particular mix can generate a sub-24-month return.

Until A3 or integrators publish welding-specific deployments and economics, the 56% nonautomotive share is a reason to ask sharper questions—not a reason to pencil a cell into the budget.