Beckhoff posted 6% in a flat market. Rockwell laid off 900 people (and counting).

Beckhoff Automation closed 2025 with €1.24 billion in revenue, up 6% year over year, the first positive growth year since the 2023 downturn. Q1 2026 was described in the company’s careful German as “strong progression.” In the same twelve months, Rockwell Automation cut 900 people, softened guidance, and spent its earnings calls explaining that “AI tailwinds across the portfolio” should be visible any quarter now. The line is the financial equivalent of saying the weather has been weather. It is also, on it’s merits, bullshit.

I have made PLC platform decisions across multiple organizations for almost thirty years. The version of this story running through the trade press right now, Rockwell versus Siemens trading punches in North America and Europe, is not the story. The story is that three different forces are pulling new machine designs away from the dedicated-PLC franchise at the same time, and Rockwell’s response so far has been to ship a half-step product (FactoryTalk Optix), retitle a digital-twin demo as “design before build,” and ask analysts to be patient. That is not a strategy. That is a stall.

This piece is a long way of saying: the architecture argument is real, the demographic clock is what makes it terminal on a fifteen-year horizon, and the captive-spec lock-in tax is what finally got machine builders to do the math.

The lens the trade press keeps using is wrong, and tired

The vendor narrative for the last fifteen years has been bipolar. Rockwell owns North American discrete and large chunks of process. Siemens owns Europe and most of Asia. Everybody else, Mitsubishi, Omron, B&R before ABB ate it, Beckhoff, and the IEC 61131-3 open-tooling community, got grouped under “specialty” or “regional” and politely ignored. The polite ignoring is the part that aged badly.

Real competition has never been Rockwell-brand against Siemens-brand. It has been a specific architectural argument: should the controller be a purpose-built PLC running proprietary firmware on a vendor’s backplane, or should it be a general-purpose industrial PC running a real-time kernel and an IEC 61131-3 runtime alongside everything else the machine wants the same CPU to do. Rockwell and Siemens built their installed bases on the first answer. Beckhoff built its company on the second. CODESYS turned the second into a vendor-neutral runtime that ships on 350-plus OEM device families and now adds AI to its 2026 workshop curriculum without asking permission from the Big Two.

Rockwell saw the PC-based question land on its doorstep fifteen years ago and chose to ride ControlLogix margin while shipping FactoryTalk Optix as a defensive software layer. FactoryTalk Optix is not a runtime. It is a visualization and connectivity wrapper sitting next to a runtime that still lives in proprietary firmware on a proprietary backplane. Calling that an architectural response is a sales-deck move, not an engineering one. The bill on that decision is what 2025 looked like.

Problem 1: The Architecture Argument

Beckhoff’s growth in 2025 came almost entirely from machine builders specifying TwinCAT and EtherCAT on new designs they had every option to spec on Rockwell or Siemens hardware. They chose otherwise. Five reasons keep showing up in the spec conversations I sit through:

• One CPU runs the PLC runtime, the vision pipeline, the motion kernel, the OPC UA stack, and the digital-twin endpoint. No second PC. No second vendor relationship. No integration tax.
• EtherCAT timing on high-axis-count motion is a structural advantage that CIP Motion does not match at the same price-per-axis envelope.
• The IDE looks closer to a modern software shop than to a 1995 ladder editor, and the engineers being hired into machine-building roles today expect that.
• AI-assistance in the IDE is a feature the runtime can host natively, with no separate appliance, on a CPU already running the rest of the stack.
• Global SKU rationalization on a single runtime is a real cost reduction for machine builders selling into Europe and Asia from a North American base.

The fifth reason, by itself, would have moved share over the last decade. The first four are how Beckhoff posts 6% in a flat capex year while the public franchise vendors post headcount cuts.

Problem 2: The Demographic Clock

This is the engine the trade press underweights. Rockwell’s installed base is real. Rockwell’s installed base is also, on average, getting older every year. The plant engineers who specified Allen-Bradley in 2002, 2005, 2010 are retiring. The machine builders specifying new lines in 2026 are run by people in their late twenties and thirties whose first encounter with controls programming was an open-source PLC simulator on a laptop, not a panel-shop ride-along at a chemical plant.

There is no marketing budget that beats a generational handoff. Rockwell has very loyal customers and a very small share of the engineers who will specify the lines that ship in 2030. Beckhoff has the inverse profile. CODESYS has the inverse profile by a wider margin. The architecture war is the surface. The clock running is what determines who is still the default in fifteen years.

This is not a “Rockwell collapses by 2028” call. The installed base is sticky enough that the retrofit and parts business will hold for another decade plus with modest erosion. It is a call that new-machine share for any greenfield line specified after 2027 looks structurally different from new-machine share specified in 2017, and Rockwell’s product response to date does not change that trajectory.

(Side story: At RND we once were required to have a Rockwell PLC in the system. We used the Beckhoff controller to control the track, robots and 95% of the machine while the Rockwell just powered the HMI. The customer knew it was a farce but they got around their company mandated rules by doing so.)

Problem 3: The Captive-spec Tax

Three things changed in the last two years. Machine builders started getting end-customer pushback on multi-vendor specs from globalized customers who do not want a Rockwell line in Wisconsin and a Siemens line in Stuttgart running incompatible HMI standards. AI-in-IDE made it visible that the proprietary-runtime vendors will ship that feature on a slower release cadence, on a more locked stack, than the PC-based camp will. Motion-heavy and vision-heavy designs hit the wall on what CIP-only architectures can do at a competitive price point.

When the captive spec stops being treated as a fixed cost and starts being treated as a tax, machine builders run the math. The math is the part of this story that does not show up in a Hannover photo op, and it is the part that moves a 6% growth number in a flat market.

The Hannover demo

The theater part is the part to discount. AI-generated IEC 61131-3 logic that you would actually deploy in a safety-rated machine, with a real integrator handoff, a real change-control workflow, real version diffs that a safety engineer can sign off on, is still between 3-7 years away from being a default. Anyone selling otherwise in 2026 is selling sizzle. I do not care how many heads of state show up at the demo.

The signal part is the part to take seriously. The fact that the AI-assistance is running inside the TwinCAT runtime, on the same CPU as the motion kernel and the EtherCAT master, on the same release train as the rest of the stack, is the architectural pitch landing. Rockwell cannot ship that capability on the same integration profile, on the same release cadence, on the same hardware. Siemens can get closer on the IDE side with TIA Portal but still it’s less polished. The Chancellor’s twenty seconds at the booth was theater. The reason the booth was the one he stopped at was not.

Where PC-based is not winning (and probably will not for a while)

PC-based control on greenfield North American installs is winning maybe 2 out of 10 clean-sheet new lines today, not 6 or 7. The moat is leaking, not breached.

The places Rockwell still wins on greenfield, today and for the foreseeable future:

• End-customer Rockwell-mandate specs on the floor at large North American manufacturers. The OEM does not get a vote.
• Brownfield-adjacent lines in plants with twenty years of ControlLogix, panel-shop standards, and an integrator bench that codes ladder cold. The switching cost is real.
• Process control at scale (oil and gas, power, chemical, food) with validated safety stacks, decades of installed base and regulatory inertia. PC-based is not credible there yet (but “yet” is doing fifteen years of work in that sentence.)
• Discrete lines where the panel-shop and integrator relationships are doing more than half the selling. PC-based vendors have not built that distribution depth in North America at comparable density.

Rockwell’s franchise on those four buckets is durable for another decade plus. The franchise on greenfield, motion-heavy, vision-integrated, globally-sold new machines are not.

The scorecard

Beckhoff posted 6% growth in 2025 in a market where Rockwell contracted. Beckhoff secured a head-of-state visit at Hannover with AI integrated into its PC-based runtime. CODESYS is shipping AI-in-IDE training on a vendor-neutral runtime that ignores the Big Two entirely.

That is one story told three ways. The architecture is winning on clean-sheet new designs. The demographics are running the clock down on the loser of that argument. The captive-spec tax is the line item on the bill of materials that finally got machine builders to do the arithmetic. The trade press wants to frame it as a vendor scrap. The earnings calls want to frame it as a temporary AI-tailwind delay. Both framings are wrong, and the second one is an excuse.

If you are specifying a new machine in the next twelve months, the question is not which PLC brand you trust. The question is whether the controller you commit to today will still be the right architecture in 2030, on the right release train, with the right integration profile, run by engineers your hiring pipeline will actually produce. Make that call deliberately.

The shelf life of “we have always run Allen-Bradley” is shorter than the lifecycle of the machine you are about to build.

Thoughts

1. As Sean shares, change in the automation space, especially in the US, will be slow. Perhaps a decade, maybe longer, but change will be slowly occurring which infers that Rockwell’s market share may slowly erode. Could this be based upon geographic area? Some vertical markets? Or perhaps tied to generational change?
2. Perhaps part of the future strategy for automation companies should be to take a page from Klein and invest heavily in the next generation. Klein, and subsequently other tool companies, entered vocational schools and flooded them, and the students, with samples and discounted product. They learned on the tools and continued buying them. Could this happen with engineers? Other specialists? The key is that this is playing for the long haul.
3. Long-term acquisition strategy. If you are an independent Rockwell distributor planning to be in business 10+ years, recognize that your industrial business in your Rockwell APR may be stagnant, or decline. If you want industrial in current non-Rockwell APRs, there are slim pickings. But, if you want business diversification, have your specialists evaluate the different automation companies and their roadmaps and then consider the automation distributor(s) you may want to acquire. There are probably three criteria – your preferred line, relationships, and age, of key salespeople, and the same with automation specialists / technical staff.
   • And expect to pay a premium.

Do you agree with Sean?