Last week I was in northern Germany, meeting companies involved in yacht and ship interiors. We discussed substrates, application methods, certifications and projects that can take months, sometimes years, to reach production.
These conversations eventually returned to the same practical issue. A material does not enter an industrial process simply because it is new or performs well in a laboratory. It has to be tested on the actual substrates, applied with the available equipment, documented, supplied consistently and accepted by both the producer and the final customer.
I had not initially connected those meetings with this week’s announcement that the European Union and the Philippines had reached substantial agreement on a free trade deal. Nor had I connected them with recent research into bio-based adhesives and the debate about whether parts of advanced AI development should sometimes be slowed down.
But they raise the same question for me: what will Europe actually be able to make and sell in the markets it is trying to open?
From a sales perspective, I welcome the EU–Philippines agreement. The two sides still have technical details to finalise, so this is not yet a trade agreement in force. If completed as announced, however, it should liberalise more than 94% of tariff lines and cover over 97% of bilateral trade.
I have worked with customers in the Philippines and know that fewer barriers can make a real difference. They can improve price competitiveness and make a market easier to serve. What they cannot do is create demand or build a distribution network.
A tariff may disappear through negotiation. Product qualification, technical support and trust do not.
Before treating an agreement as a commercial opportunity, I would want to understand which products gain a meaningful advantage, which local industries are growing, whether we have the right partners and whether we can support customers after the first shipment. Market access matters, but it becomes useful only when a company has something distinctive and reliable to bring through the door.
A recent development from KAIST in South Korea gives this question a more concrete form. Researchers engineered E. coli to produce aromatic PHA polymers from glucose and tested them as hot-melt adhesives.
One formulation achieved a higher lap-shear strength than a commercial EVA adhesive on stainless steel under the conditions used in the study. That does not make it a better adhesive in general. The same research reports lower peel resistance in another test, much longer open times and the need for further work before large-scale implementation.
Performance depends on the substrate, process and application—something anyone working with industrial adhesives learns quickly.
Still, the research matters. It suggests that functional polymers traditionally derived from petrochemicals could eventually be produced through biological processes and adjusted for different properties. The product may not be ready, but the direction is worth watching.
AI may shorten part of that journey. Researchers at Tohoku University recently described a possible closed-loop system connecting polymer databases, predictive models, AI agents and automated experiments. It is a framework for future development, not an autonomous industrial laboratory already producing market-ready materials. Its potential lies in reducing some of the slow and wasteful trial and error involved in finding promising candidates.
At almost the same time, OpenAI announced that it had paused reinforcement-learning work on some advanced models while strengthening its monitoring, security and alignment safeguards. It was a targeted pause, not a halt to AI research. The decision followed a security incident and preliminary evidence that a forthcoming model might possess critical cybersecurity capabilities.
I do not see these developments as inconsistent. A technology can be valuable enough to accelerate scientific discovery and powerful enough to require limits while its control systems catch up.
Managers face a less dramatic version of this tension whenever they decide whether a new material, process or technology is mature enough to introduce into an organisation. Moving too slowly carries a cost. Moving before we understand the consequences can create another one.
For Europe, greater access to markets and faster technological development should represent an opportunity. The harder question is how much of the resulting value we will retain.
European manufacturing is not dead. In 2025, the value of EU sold production increased by 2.9% in constant prices after two years of decline. Yet the picture is uneven: chemicals and fabricated metal products both fell by 1.4%. A rebound does not settle the longer-term question.
Our competitiveness will depend less on whether an invention originates in a European laboratory than on whether we can industrialise it at a viable cost. Research, regulation and branding are not enough if the knowledge needed to scale production, solve application problems and serve customers moves elsewhere.
Trying to preserve every existing process simply because it is European would be equally shortsighted. Europe does not need to manufacture everything. It does need to decide which capabilities it cannot afford to lose and in which emerging fields it wants to build an advantage.
That decision will also have a physical effect on our cities and regions.
I was trained as an architect, so I tend to notice this part of the discussion. An industrial area is not merely a collection of buildings outside a town. It contains jobs, technical knowledge, suppliers, transport, training and a large part of the income that supports the surrounding community. Working in Veneto makes that relationship difficult to ignore.
If more industrial value moves towards laboratories, automated production, data infrastructure and highly specialised services, industrial districts will change with it. Some existing buildings may no longer be suitable. Others may be adapted for smaller and more flexible forms of production.
Logistics centres and data centres may occupy sites once used by factories, but replacing one building with another tells us little about whether the territory is better off.
We also need to ask what kind of employment is created, which skills remain locally, how much energy the new activity requires and whether it generates relationships with the surrounding economy. These are not only planning questions. They belong in industrial and investment decisions from the beginning.
This is why I eventually began to see a connection between a trade agreement, an experimental adhesive and a pause in AI development. Each sits somewhere between possibility and execution.
Opening a market creates an opportunity, but not the products with which to enter it. Research produces new possibilities, but not necessarily an industrial process. AI may accelerate discovery, while making the question of control more urgent.
What follows depends on decisions made much earlier than the first sale: what to research, what to produce, where to invest and which risks we are prepared to carry. It also depends on whether companies can connect technical innovation with actual customer problems.
The question is not simply whether Europe will continue to manufacture.
It is what kind of manufacturing can still create enough value to sustain its companies, its places and the communities living around them.