The European Commission’s recent decision to classify ChatGPT as a Very Large Online Search Engine under the Digital Services Act has generated surprise and, in some cases, irony. For many observers, ChatGPT is first and foremost a generative artificial intelligence system, not a traditional search engine.
Yet the European decision has its own logic. The Digital Services Act does not necessarily require a page full of links to qualify a service as a search engine: what matters is the role the service plays in providing access to information. Through its search functions, ChatGPT reached an average of 159.1 million monthly active users in the European Union during the six months ending in March 2026, well above the threshold established for Very Large Online Search Engines.
Source:
https://www.lastampa.it/tech/2026/09/01/news/per_l_europa_chatgpt_e_un_motore_di_ricerca-15727495/
The more interesting issue, however, lies elsewhere.
This decision provides an opportunity to ask a much more important question: where does Europe actually stand today in the global artificial intelligence race?
And, more importantly: is it still possible to close the gap?
If we look at the data without ideological bias or excessive enthusiasm, the answer is more complex than the usual slogans suggest.
Europe Is Not Out of the Race. But It Is Not Playing the Same Game as the United States and China
It would simply be wrong to claim that Europe lacks expertise in artificial intelligence.
The European Union has world-class universities and research centres, a significant industrial base, advanced scientific capabilities, and companies holding strategic positions in several parts of the technology value chain.
Europe has also built one of the world’s most advanced regulatory frameworks through the AI Act and is investing in supercomputing, AI Factories, research and infrastructure.
At the same time, the European Commission itself recognises that one of the main challenges is the innovation and investment gap. The Competitiveness Compass, drawing in part on the diagnosis contained in the Draghi report, identifies closing the innovation gap as one of the fundamental priorities for European competitiveness.
Source:
https://commission.europa.eu/topics/competitiveness/competitiveness-compass_en
The situation becomes much clearer when we look at the most advanced part of the AI value chain: foundation models, the capital required to develop them, computing capacity, and the speed at which all these elements can be brought together.
This is where the comparison becomes considerably less favourable for Europe.
The Gap with the United States Is Above All a Gap in Capital
The AI Index 2026 from Stanford University’s Institute for Human-Centered Artificial Intelligence paints a picture of an extremely concentrated market.
The United States continues to dominate private investment, frontier model development and the ability to rapidly transform capital into AI infrastructure and products at global scale.
China, however, can no longer be regarded simply as a follower. The technological competition between the United States and China has become significantly tighter, while both countries have the financial, infrastructure and industrial resources required to sustain a long-term race.
Source:
https://hai.stanford.edu/ai-index/2026-ai-index-report
This highlights a fundamental difference between Europe and the United States.
Having excellent researchers is not enough.
You also need the capital required to turn research into companies, products and infrastructure at global scale.
Frontier AI has become an extremely capital-intensive technology. Training and operating increasingly sophisticated models requires enormous amounts of computing power, energy, infrastructure and capital.
The competition is therefore not simply about who develops the best algorithm.
It is about who can build the entire economic ecosystem required to support it.
China Is Not Simply “Behind”
The second mistake would be to view the situation as a simple race between the United States and a Europe that merely needs to catch up.
China is proving to be a first-class competitor.
According to the Stanford AI Index 2026, the gap between leading US and Chinese models has narrowed considerably. Models from the two countries have repeatedly alternated at the top of performance rankings, and the distance between the leading systems has become much smaller than it was in the past.
China also has a powerful industrial base, enormous quantities of data, extensive manufacturing capabilities and a strongly technology-oriented industrial policy.
Source:
https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
It is therefore more accurate to describe the situation as a primarily US-China competition at the frontier of AI, rather than imagining Europe as an equivalent third competitor.
And this is precisely what makes Europe’s position more challenging.
Europe’s Problem Is Not Simply Creating a Better Model Than ChatGPT
It might seem that the solution is straightforward: finance a major European company and build a “European ChatGPT”.
But that would be an oversimplification.
The real competition concerns an entire value chain.
It requires:
- fundamental research;
- researchers and engineers;
- semiconductors and accelerators;
- computing capacity;
- data centres;
- energy;
- cloud infrastructure;
- datasets;
- venture capital;
- sufficiently large markets;
- companies capable of adopting technology rapidly;
- startups capable of scaling into major businesses;
- management expertise;
- digital infrastructure;
- and the ability to move research rapidly into market-ready products.
It is a chain in which weakness at one point can undermine the others.
And this is where one of Europe’s main vulnerabilities becomes apparent.
European Capital Exists. The Problem Is Turning It into Capital for Technological Growth
Europe is not short of capital.
The problem is its ability to mobilise that capital and direct it towards technology companies capable of scaling into global players.
This is one of the issues highlighted by the European Commission in its analysis of financing the AI economy.
The European ecosystem is relatively effective at the early stages of startup development, but it faces greater difficulties when companies require much larger funding rounds to scale rapidly.
The risk is significant.
A startup may be founded in Europe, develop its technology using European researchers, use European infrastructure and then, precisely when it becomes economically attractive, obtain financing from foreign investors, be acquired or shift its growth towards foreign capital markets.
The risk is not necessarily that innovation will disappear from Europe.
The risk is that the economic value created by that innovation will be captured elsewhere.
The Question Is Not Only Who Invents AI. It Is Who Captures Its Value
This may be the most important question for Europe.
Imagine that over the next ten years artificial intelligence dramatically increases the productivity of companies, professionals and public administrations.
Who will sell the models?
Who will provide the cloud infrastructure?
Who will build the data centres?
Who will manufacture the chips?
Who will control the platforms?
Who will own the intellectual property?
Who will capture the margins?
And, above all, where will the companies that benefit most from this transformation be headquartered?
If the answer is predominantly “in the United States and China”, Europe could capture some of the benefits of AI as a consumer and user without capturing an equivalent share of the industrial value it generates.
That is far from a theoretical distinction.
AI Could Become a Productivity Issue
At this point, the issue moves decisively beyond technology.
Artificial intelligence is a general-purpose technology, with transformative economic potential comparable, in its nature, to previous major innovations such as electricity, computing and the Internet.
The European Central Bank has examined precisely this aspect: AI’s impact on the European economy will depend not only on the quality of the models, but also on how quickly companies adopt the technology and integrate it into their production processes.
Source:
https://www.ecb.europa.eu/press/key/date/2026/html/ecb.sp260323_1~1e06784a89.en.html
And this is where the latest data allow us to correct an excessively pessimistic narrative about Europe.
According to a new ECB analysis published on 26 August 2026, the share of euro-area workers reporting that they use AI at work increased from 26% in 2024 to 41% in 2025, reaching 52% in 2026.
AI, therefore, is by no means a marginal technology in the European labour market.
Source:
https://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260826~e1c1a89999.et.html
The problem is that adoption and productivity are not the same thing.
The ECB also finds that the median user reports saving around three hours per week thanks to AI. But the time saved only translates into a genuine productivity increase if those hours are actually converted into greater output or higher-value activities.
This is a fundamental distinction.
Using ChatGPT to write an email more quickly is useful.
Redesigning a business process in which AI automates entire operational stages is something entirely different.
Europe’s Real Problem May Be Deep Adoption
ECB research published in 2026 highlights precisely this paradox.
A large majority of euro-area companies report some form of AI use, yet only a small minority consider their adoption significant.
Source:
https://www.ecb.europa.eu/home/html/index.en.html/pub/pdf/scpwps/ecb.wp2713~91ddff9e7.sk.html
Earlier ECB analysis also indicates that companies using AI in core business processes are more likely to obtain meaningful benefits than those using it only for peripheral or occasional tasks.
Source:
https://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260624~44f70da110.lt.html
This fundamentally changes the question.
It is no longer simply:
“How many European companies are using AI?”
The question becomes:
“How many European companies are actually transforming their processes through AI?”
And this is probably where a significant part of the competitive race will be decided over the coming years.
The European Paradox: Leading in Regulation, Lagging in Technology
This is where the debate risks becoming ideological.
Criticising Europe simply because it regulates AI would be simplistic.
The AI Act addresses a real issue: a technology capable of influencing information, employment, security, fundamental rights and decision-making processes requires rules.
The Digital Services Act addresses legitimate concerns as well.
The classification of ChatGPT as a VLOSE reflects a genuine concern: if millions of people use an AI system as a primary intermediary for accessing information, that system acquires enormous power over how information is selected and presented.
The issue is therefore not:
regulation or innovation.
The issue is finding a balance in which regulation does not prevent innovation from being created, scaled and made competitive.
There is also an interesting development here: Europe appears to have begun correcting some aspects of its approach.
The AI Omnibus, which entered into force in July 2026, introduces simplifications, extends certain deadlines, expands experimentation mechanisms and aims to make compliance more proportionate, particularly for smaller companies.
Source:
https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force
This suggests that European policymakers have recognised at least part of the problem.
Europe Is Responding. But Is It Responding Fast Enough?
It would therefore be unfair to claim that the European Union is doing nothing.
In 2025, the European Commission launched InvestAI with the objective of mobilising up to €200 billion for AI investment, including a €20 billion European fund dedicated to AI Gigafactories.
The AI Continent initiative also includes plans for at least 19 AI Factories across Europe and aims to substantially expand computing infrastructure.
Source:
https://commission.europa.eu/topics/competitiveness/ai-continent_it
These figures matter because they demonstrate that the problem has been recognised at institutional level.
But there is another question that only time can answer:
How quickly will these investments translate into technological and industrial capabilities?
The issue is not simply how much is spent.
It is how quickly capital becomes infrastructure, infrastructure becomes research, research becomes products, and products become competitive companies.
Computing Has Become a Strategic Resource
One concrete example is computing infrastructure.
On 31 August 2026, the European Union awarded a €387.8 million contract for the construction of LUMI-AI in Finland, a new supercomputer dedicated to AI workloads.
The system is expected to become operational in the second half of 2027.
This is an important investment.
But it is also a snapshot of the challenge.
Europe is building today computing capacity that the United States and China began developing on a massive scale several years ago.
That does not necessarily mean it is too late.
But it does mean that execution speed has become a strategic variable.
Time May Be the Most Important Factor
This is probably the most underestimated part of the problem.
In technology, five years can represent an entire generation.
A company that today lacks computing capacity, capital and expertise may not simply be “behind”.
Five years from now, it could find itself facing an ecosystem in which:
- the best researchers work elsewhere;
- the strongest startups have been financed by foreign capital;
- foundational models belong to non-European companies;
- cloud infrastructure is controlled by a handful of major players;
- computing infrastructure is concentrated outside Europe;
- European companies predominantly use technology developed elsewhere.
At that point, catching up would become much more expensive.
This is the cumulative effect of innovation: those who start ahead can reinvest the revenue, data, talent and experience gained from one generation into the next.
The Risk of a New Form of Technological Dependence
This brings us to another dimension of the issue: technological sovereignty.
For years, Europe has debated energy dependence.
Today, it is also discussing dependence on semiconductors, cloud infrastructure and digital platforms.
AI could add another layer.
If an increasing share of the European economy depended on AI models, cloud infrastructure and computing capacity controlled by foreign companies, Europe could find itself in a position where a technology fundamental to productivity and economic competitiveness is not genuinely under its own control.
This does not necessarily mean building a completely autonomous European AI ecosystem.
It means having credible alternatives.
That distinction matters.
Technological sovereignty does not necessarily mean technological autarky.
It can mean having enough industrial, infrastructural and scientific capacity not to be completely dependent on a single external ecosystem.
Where Europe Could Still Have an Advantage
The situation is not necessarily irreversible.
In fact, Europe has characteristics that could become significant advantages in several areas.
The first is industry.
Automotive, advanced manufacturing, energy, pharmaceuticals, aerospace, robotics, finance and industrial systems are all sectors in which Europe has highly relevant companies and expertise.
In these fields, the value of AI may not be determined solely by who develops the most powerful language model.
It may depend on who is best able to integrate AI, robotics, sensors, software, industrial data and domain-specific knowledge.
The second advantage is research.
The third is the single market.
The fourth is the enormous amount of industrial and scientific data that could become a strategic asset if properly organised and made usable.
The fifth is the possibility of building a European model of AI that combines technological capability with high standards of reliability, transparency and protection of fundamental rights.
None of these factors guarantees success.
But they are real assets.
The Most Important Opportunity May Be Applied AI, Not Necessarily General-Purpose AI
Perhaps the question Europe should be asking is not:
“Can we build the next OpenAI?”
It could instead be:
“How can we become the place where AI is applied most effectively to industry?”
That is a completely different strategy.
Europe may not necessarily need to win the race to develop the world’s most powerful general-purpose model in order to become a global leader in AI.
It could instead focus on robotics, manufacturing, medicine, pharmaceuticals, automotive, energy, logistics, finance, public administration and science.
The European Commission is moving in this direction through initiatives aimed at accelerating AI adoption across industry and public services.
Source:
https://commission.europa.eu/topics/competitiveness/ai-continent_it
It is probably one of the most realistic paths forward.
But it requires something that is often harder to obtain than funding:
speed.
The Real Cost of Inaction May Not Be Visible Today
This is perhaps the most important conclusion.
If Europe does not produce the world’s best language model tomorrow, nothing dramatic will happen.
The problem would emerge gradually.
One European company might use an American model.
Another might use a Chinese model.
Another might rely on the cloud infrastructure of a major international platform.
All of them could increase their productivity.
On the surface, everything would work.
But if the intellectual property, profits, cloud infrastructure, computing capacity and platforms enabling that growth were predominantly located outside Europe, a significant share of the economic value generated by AI would leave the continent.
That is the potential missed opportunity.
Not being the first to invent a technology is one thing.
Being forced to buy it for decades is another.
And This Brings Us Back to ChatGPT
The classification of ChatGPT as a search engine may therefore appear to be a technical or bureaucratic issue.
In reality, it represents something much more interesting.
It shows how a technology that began as a language model has simultaneously become:
- a personal assistant;
- a search tool;
- a software interface;
- a development environment;
- a business tool;
- an intermediary for information;
- and, increasingly, an agent capable of performing tasks on behalf of the user.
The speed at which these categories are converging inevitably makes it difficult for any regulator to design perfect rules.
And this applies to Europe as well.
The point is therefore not to accuse Brussels of failing to understand artificial intelligence.
That would be an overly simplistic conclusion and, more importantly, one not supported by the evidence.
The real question is more serious.
Europe appears to have understood the challenge. But it must now demonstrate that it can turn that awareness into industrial capability at the same speed with which the United States and China are turning AI into economic infrastructure.
Because the next phase of the competition will not simply be a race to build the best chatbot.
It will be a race to build the most productive economies, the most efficient industries, the most innovative companies and the most strategically important technological infrastructures.
And in that race, starting several years behind could have a much higher cost than we currently realise.
The real question, therefore, is not whether Europe should become American or Chinese in the way it approaches artificial intelligence.
It is whether it can build its own competitive model, one capable of combining research, industry, capital, infrastructure, innovation and regulation.
Because regulating a technology that others have already learned to build may be necessary.
But being able to build it, develop it and turn it into economic value is something else entirely.
And that is probably the race Europe cannot afford to lose.
