Drawing on an analysis of 16,000 German companies, Andreas Venus, Managing Partner for McKinsey in Germany and Austria, examines why productivity growth is concentrated among a remarkably small group of firms, and what their experience tells us about the respective roles of corporate leadership, technology and public policy.
More than the current debate suggests, but not everything. Let me take the second part first, because it is real. Growth-oriented regulation, a flexible labor market, competitive energy prices, working infrastructure and a skilled workforce are not nice-to-haves. They are fundamentals for productivity growth. Our research shows that Germany has considerable ground to make up on several of them compared to other countries. Depending on the industry, production costs are 35 to 50 percent higher than at the most competitive location abroad. These are factors that lie beyond the control of any single company.
But the data also shows how much sits with companies themselves. We analyzed some 16,000 German companies, together about a third of the country's economic output. The framework affects industries very differently, yet productivity champions emerged in many of them, while the broad middle barely moved. And when we asked board members what holds back transformation, they named their own company culture as often as regulation, 29 percent each. That is an honest answer, and it is why I say productivity is a decision, not a destiny. The framework sets the field. Who plays well on it is decided inside the company.
Both. Twenty-nine companies, 0.2 percent of the sample, generated 47 percent of the productivity growth between 2019 and 2023. Around six in ten companies in the sample contributed nothing to it. Read as a warning, that says growth rests on a very narrow base. I do not want to talk that away.
But I read it mainly as a lever. If so few companies can move a country's numbers this much, the country does not need everyone to change. It needs a few more to decide. If only a few dozen joined that group, Germany's growth problem would look very different. I call this the power of the few. And it is not a German peculiarity. Our McKinsey Global Institute has found the same concentration in other advanced economies. So, the agenda for any CEO or policymaker is concrete: understand what the leaders did and make it repeatable for the broad middle.
The encouraging part first: they are not one sector and not one type of company. Among them are manufacturers, media companies and pharma, large incumbents as well as Mittelstand firms and scale-ups. What they share are four patterns. They concentrated their resources on segments with growth potential. They kept improving their products, and in the decade before had spent two to three percentage points more of revenue on research and development than the laggards. They scaled in the lead markets of the United States and China. And they made processes, structures and supply chains more efficient with real cost discipline. None of this is secret, and none of it is tied to a German postcode. A family company in Lombardy or a manufacturer in Pune can make the same choices.
What is harder to copy is the world in which those patterns worked. Geopolitics is shifting the lead markets, capital moves faster, and AI is changing the rules of competition faster than planning cycles. So, the patterns remain the foundation, but they will not be enough. The next round is about occupying new growth fields, allocating capital far more dynamically, partnering across sectors and letting go of businesses without a future. That is replicable anywhere. What it takes is a clear strategic choice, and the rigor to carry it through.
Probably both are true at once. The direction is real. AI is moving robotics from fixed automation to machines that perceive, learn and adapt. It is happening in stages: from one robot for one task, to robots trained for whole families of tasks, which is where we are today, to truly general-purpose robots that can be deployed without training, which we do not expect before the second half of the 2030s. Our colleagues estimate that robotics and physical AI could create at least a trillion dollars of economic value by 2040, most of it in manufacturing and logistics. Investors are clearly betting on it: investment in general-purpose robots and the AI models that control them has multiplied several times since 2023, and it is flowing into hardware and models at the same time.
But expectations are running ahead of the physics. Today's pilots are mostly about learning. The returns follow once robots run reliably without human intervention, stop safely and pay for themselves, and in many industrial tasks 95 or even 99 percent reliability is not enough, you need 99.999 percent. That takes reliable AI models and reliable hardware at the same time. On the hardware side, the tightest constraint is the supply chain for actuators and for force and tactile sensing, the parts that turn a signal into a precise, safe movement. This is an industrialization challenge as much as an AI challenge. So yes, it is an inflection point, but one that will be decided on the factory floor, not in the demo video.
Let me be direct about scale first. China accounted for 54 percent of global industrial robot installations in 2024, files several times as many humanoid patents as the United States and backs this with a state guidance fund of around 138 billion dollars for AI, robotics and other high-tech sectors. A large part of its edge comes from the overlap with its electric vehicle industry: motors, power electronics, batteries and, above all, magnets, where China processes around 90 percent of the world's supply.
Europe will not win on assembly volume. But much of a humanoid's value sits in its components: actuators alone account for 40 to 60 percent of the bill of materials. So, Europe does not have to build the most robots to capture a meaningful share of the value. It has to supply the most demanding parts of them, wherever they are assembled, and set the standard for certified safety.
Europe's opportunity lies where that EV overlap helps least and precision matters most. I would prioritize three capabilities. First, the high-precision core of the robot: precision drives and actuators, roller screws, force and tactile sensors. These are exactly the components our colleagues identify as the likeliest bottlenecks, and they reward the kind of engineering depth Europe's suppliers have. Automotive and precision-motion suppliers are already claiming the role of the "engine maker" of the humanoid economy, a role that does not formally exist yet.
Second, certified safety. The new EU Machinery Regulation applies from 2027, and the AI requirements for machinery and a dedicated international safety standard for humanoids are due by 2028. Once those are in place, Europe has a certifiable path for robots that work next to people without fences. If we industrialize that first, safety becomes an export, not a compliance cost.
Third, deployment in complex, brownfield production, where robots have to fit into existing plants. Germany has one of the highest robot densities in the world, and those plants generate the real-world data physical AI learns from.
Having worked with this industry for more than fifteen years, I am confident the capability is there. The opportunity now is to claim that role deliberately and to back it with capital at scale.
In ageing societies this is not an abstract question. In Germany, 51 million people are of working age today. By 2035, one in four people will be 67 or older, and by 2038 the number of people of pensionable age will have grown by at least 3.8 million, while much smaller cohorts follow. Even high immigration cannot close that gap. And this is not only a European story.
China's population fell for the fourth year in a row in 2025, and 23 percent of its people are now 60 or older. China is automating for the same demographic reason, at far greater scale. So, the honest framing is not "robots or people." The question is increasingly who will do the work at all. But timing matters: humanoids are a 2030s answer to a 2020s problem. In the next few years, the relief will come from automation we already have, such as mobile robots, collaborative robots and mobile manipulators.
That said, I would caution leaders against hiding behind the demographic argument. Employees notice when efficiency is the real story. And our experience with AI shows how easily value gets lost on the way. In our latest global survey, eight in ten respondents say AI has made them personally more productive. But only about 37 percent say it has added to their company's earnings, practically unchanged from the year before. Individual productivity is not yet a company result. The leap between the two happens in the process, not in the tool.
The same will hold for physical AI. The companies that get this right treat robots as a change in what and how they produce, rebuild the process and the roles around the machine, and invest in the skills that go with it. Where it is treated only as a headcount exercise, it tends to meet resistance and usually does not deliver the desired productivity either.
Execution is where much of it is won or lost. Our transformation research shows that even companies that consider their transformation a success capture only about two-thirds of its potential value, and more than half of the value that is lost slips away during and after implementation, not in the plan. Germany has outstanding crisis managers. We proved it through the pandemic, the chip shortage and the energy shock that followed Russia's invasion of Ukraine. But crisis management is reactive. Deliberately rebuilding a company is a different craft, and few boards have led a real restructuring themselves, partly because tenures have become shorter. At the same time, according to our analysis, almost every second industrial company in Europe shows signs of restructuring need.
In practice, being a chief capital reallocator means three things. Replace budget-driven continuity with regular portfolio reviews and clear criteria for where to invest and where to exit. Be willing to let go of businesses that are still profitable but structurally shrinking, because they tie up capital and talent needed for new growth opportunities. Today German companies generate only about 6 percent of their revenue in such fields, less than a third of the share of their Chinese competitors. We are world champions at perfecting. Occupying new fields takes something else: rebuilding the old decisively and funding the new at the same time, before the wall is there, not after. It is uncomfortable, because the old business still pays the dividends and the new one does not yet. But waiting is the most expensive option, and in an upswing it is the most tempting one.
I see it as a division of roles. There are a few conditions that companies cannot create on their own: growth-oriented regulation, a flexible labor market and energy at competitive prices. In Germany, net investment as a share of GDP fell from around 2 percent to 0.2 percent in 2024. Our analysis does not attribute that to a lack of willingness among companies to invest, but to a structural cost disadvantage. How to address it is for policymakers to decide. For the executives we surveyed, regulation and bureaucracy are the greatest external obstacles to implementing bold strategies, so the pace of implementation matters as much as the measures themselves.
On the business side, the room to maneuver is greater than the debate suggests. The 29 champions did not wait for perfect conditions; they made their decisions within the conditions they had. And there are initiatives in which companies and the public side work together on concrete commitments, such as Made for Germany, where many large companies have pledged investment. That is the kind of alliance that turns intentions into numbers.
At a summit about alliances, I would add one more: along the value chain. Many of the new growth fields, physical AI among them, are being built through co-development between technology providers, component suppliers and the companies that deploy the technology. In robotics, our colleagues see exactly these partnerships as the way through the current supply bottlenecks.
The question would be: which decision will you take in the next twelve months that your company, or your country, will still show in 2035?
As for the end of the decade, I am a data person, so I would look at three numbers. First, breadth: that productivity growth no longer rests on 0.2 percent of companies, and the circle of champions has grown well beyond 29. Second, direction: that the share of revenue European companies earn in new growth fields has moved clearly beyond today's 6 percent, and net investment is well above the level that only maintains what exists. Third, results from technology: that a majority of companies using AI can show it in their earnings, not just in their employees' daily work, and that some of the companies supplying and deploying the robots of the next decade are European.
We do not need everyone to move. We need enough companies to decide, and others will follow.
Andreas Venus is a Senior Partner at McKinsey & Company and a Managing Partner for Germany and Austria. He previously led the firm's Advanced Industries Practice in Germany. Since joining McKinsey in 2009, Andreas has worked with companies across automotive, industrials, aerospace and defense, and semiconductors, with particular depth in strategy, purchasing and product development. His focus is on turning data and analytics into measurable bottom-line results. Before McKinsey, he gathered early experiences ranging from JPMorgan to Nike to the United Nations. He holds a PhD from TU Dortmund and studied in London, Paris and Berlin.
Share on X
Share on LinkedIn