Almost every day, we can see advancements from major AI players in the US and China, citing increasingly complex and powerful AI models. Europe, on the other hand, often gets left behind in the coverage, lacking any major AI players (at least when it comes to models themselves). In today’s blog, we investigate the European AI landscape, its startups, as well as what could be done in the future to improve the situation and put Europe on the EU map as well.

The AI industry has experienced one of the fastest growths of any industry in history. Since the initial models were introduced by OpenAI a couple of years ago, technology has advanced to such an extent that it overtakes human capabilities on many metrics.

This growth has been primarily driven by US companies, such as OpenAI, Anthropic, and xAI, but also increasingly by Chinese companies. Europe, on the other hand, has been on the passenger side, with no major players present, often relying on US companies to automate many processes. However, while the continent is “missing” major names, some older companies and new startups are capturing segments of the market in their own niches.

Select list of European AI companies incl. their latest credible valuations (USD bn)

Source: Euronext, Financial Times, The Times UK, Wortins, Business Insider, InterCapital Research

Sorted by valuation, the largest one is a German company, Helsing, which has the latest valuation of approx. USD 18bn. To put things into perspective, OpenAI is valued at over USD 852bn as of the latest valuation, and Anthropic at USD 965bn, meaning that European companies, at least in terms of size, are far behind their American counterparts. Helsing itself specialises in Defence AI, sensors, autonomous systems, drones and underwater systems. The next on the list, a London, UK-based company, Nscale, specializes in AI cloud, GPU infrastructure and sovereign AI data centres, valued at USD 14.6bn. Mistral AI, the most-well known European AI company, which is directly comparable to US heavyweights, is valued at only USD 13.7bn (September 2025), although newer talks, not yet official, put it closer to >USD 20bn.

Other companies, such as the Swedish Lovable, are valued at USD 13.3bn, focusing on Natural-language software and web-app creation. Among the other companies on the list, we have ones that specialize in AI voice, speech synthesis and audio generation, embodied AI and autonomous-driving software, drones, agentic legal AI, among many others. Many of these companies have also expanded rapidly, attracting far higher valuations during short periods of time.

In other words, Europe does not lack creativity or talent when it comes to AI; what it does lack is scale.

Private AI, Private generative-AI investment in 2025 (USD bn)

Source: Stanford AI Index 2026

When looking at the levels of private investments, be it in AI companies in general or generative AI (read: LLMs like ChatGPT or Claude), Europe is lagging far behind the US. US private AI investments amounted to almost USD 286bn in 2025, with generative AI investments at almost USD 164bn, while in Europe, Private AI attracted only about USD 20.9bn, and generative AI far less, at only >USD 3.2bn. In other words, Private AI investments are 13.7 times higher in the US than in Europe, while generative AI investments are about 51 times higher.

Chinese numbers, while appearing modest, cannot be directly compared to American or European numbers, due to government-guided funds and state capital, which is significant within the country. Furthermore, many of the Chinese AI companies are part of the largest players such as Alibaba or ByteDance, and as such, already have significant funding available as compared to newly opened startups.

Even within Europe, the majority of investments into private AI are in the UK (USD 5.9bn in 2025), meaning that a country outside the EU makes up more than ¼ of all AI investments. The discrepancy is also visible in the number of newly funded AI companies.

Number of newly funded AI companies in 2025

Source: Stanford AI Index 2026

US itself created more than 1.95k companies in 2025; the UK followed with 172, China with 161, and India with 108, while Europeans began with Germany at 92 and France at 84, while the rest are lagging behind Canada, Israel, South Korea, Japan, etc. In other words, the US itself has funded dozens of times higher numbers of AI companies as the EU. Even a country which is considered “poor” by global standards, India, has more AI-funded startups than the largest economy in the EU!

But even ignoring these numbers, one big issue that Europe might run into in the future is its high dependency, which, as we saw with the recent turmoil in the world, is something that one of the world’s largest economic blocks cannot afford. In the European Commission’s 2026 Cloud and AI Development Act proposal, it is stated that the EU providers’ share of the European cloud market fell from 29% in 2016 to 15% in 2022 and subsequently stagnated, while three non-EU hyperscalers now control more than 70% of the European cloud market.

Within the same report, the Commission stated that insufficient domestic data-centre capacity and dependence on a small set of third-country suppliers is one of the major constraints on European AI development. To be fair, this disadvantage goes back before AI, as Europe did not create equivalents to AWS, Microsoft Azure, or Google Cloud. As AI shifted from a mostly software problem to a combination of models + chips + data centres + electricity + cloud distribution, the lack of European hyperscalers became much more expensive. Mario Draghi, former Italian PM and President of the ECB, also stated in his competitiveness analysis that Europe had missed significant parts of the earlier digital revolution and imports more than 80% of its digital technologies.

To be fair, China does have its own constraints, such as US restrictions on access to the most advanced chips, but unlike Europe, it possesses enormous domestic digital platforms, cloud providers, device manufacturers and a state-supported semiconductor strategy. Even with these restrictions, China was able to close the gap to US frontier models, as demonstrated by its latest releases (DeepSeek, Kimi, Qwen, GLM, among others), which are in many cases comparable to the top dogs from the US.

However, not all is doom and gloom for Europe. As a continent, it is far more competitive in physical and industrial AI, with companies such as Wayve (autonomous driving), Helsing and Quantum Systems (defence autonomy), Neura Robotics (robotics), PhysicsX (engineering), and CuspAI (materials). However, in the physical-AI space, China is also powerful, as it installed roughly 54% of all industrial robots globally in 2024.

As such, Europe’s industrial installed base and engineering expertise represent both a comparative advantage and a battlefield on which Chinese competition may become even more intense. Meanwhile, the American advantage is that it can combine software leadership with capital and hardware. The frontier-model race is increasingly belonging to industrial organizations, with more than 90% of notable frontier model releases in 2025, showing just how difficult it is for university labs or lightly capitalized startups to remain at the frontier.

Knowing that the EU has the talent and the industrial base to take this to the next level, what’s stopping it? There is no singular answer, but a combination of factors.

Firstly, funding. EU has a much smaller risk-capital (venture capital, VC) pool than the US, estimated at EUR 150bn vs. US’s EUR 930bn. This issue becomes particularly severe in later-stage funding rounds, precisely when a successful AI company needs hundreds of millions, or billions of euros for compute, hiring, and international expansion.

As a point of comparison, European and UK VC funds raised less capital on average in 2020 – 2025, at about EUR 73m per fund per year across 2,321 funds, versus EUR 103m across 7,433 US funds. The US advantage, therefore, consists of both more funds and larger funding available. This is an old European problem, especially in the context of AI, as financing requirements have changed dramatically. For example, OpenAI’s USD 122bn March 2026 financing and Anthropic’s USD 65bn May financing are orders of magnitude beyond the scale of conventional software venture rounds. A region can have excellent researchers and seed-stage VC and still be functionally excluded from the frontier race if it cannot finance at this scale.

This issue also creates a feedback loop. European companies take US money; investors encourage US expansion; management and sales functions move closer to the US market; eventual exits occur in the US; the proceeds, experienced executives and institutional expertise are then recycled into the US ecosystem.

The problem is not simply that Europeans are poor or lack savings. European finance remains more bank-centred and nationally fragmented than the US financial system, while venture investing requires equity investors willing to accept high failure rates and long-duration illiquidity. An IMF analysis argues that this bank-dominated model is structurally less suitable for young companies with few tangible assets. Furthermore, US pension funds and foundations play a much larger role in VC. The ECB did recommend increasing institutional participation, particularly from pension funds, and integrating European capital markets to help mitigate this.

The difference here is huge. A foundation model, autonomous vehicle stack or robotics company often has little conventional collateral but can require billions before producing stable cash flows; bank credit is an imperfect instrument for that risk.

Lack of hyperscalers, as described above, is also an issue. A European startup frequently trains on US-designed NVIDIA GPUs, inside infrastructure controlled by US cloud providers, using capital partly supplied by US investors, before selling into a market whose largest technology customers are also US-based. In other words, a large share of the AI value chain remains outside European ownership.

The single market also presents an issue. While the EU is a nominal home market for approx. 450m people, startups encounter different corporate-law regimes, employment rules, option taxation, insolvency regimes, financial markets and administrative practices. According to research by the IMF, fragmented regulation, inefficient financial intermediation, limited labour mobility, and fragmented energy markets are the four binding constraints on EU firms scaling across borders. In the same research, 60% of EU exports reported having to comply with differing standards and consumer-protection rules across member states. Even where rules are formally harmonized, national implementation and company infrastructure can still create multiple fixed costs.

Compare that with a software startup in California: once it has product fit for market, expansion across the US does not require establishing 27 different corporate structures or understanding 27 national option-tax regimes. China similarly gives domestic companies an enormous national market, even though Chinese regulation creates other constraints.

EU’s underinvestment in private digital R&D before the AI boom is also posing an issue. According to Mario Draghi, European companies spent around EUR 270bn less on R&D than US counterparts in 2021, with much of the difference explained by industrial composition. EU’s leading corporate R&D spenders remained concentrated in automotive, while the US shifted towards tech companies.

This matters as today’s frontier AI ecosystem was built on yesterday’s cloud, semiconductor, search, advertising, mobile, and developer-platform profits. Europe does have outstanding companies such as ASML and world-class industrial engineering, but far fewer cash-rich consumer/cloud software platforms able to spend tens of billions annually on speculative AI R&D.

Energy also plays a role in this context. AI has converted electricity policy into technology policy. Training and serving models requires huge data centres, while semiconductor fabrication is similarly energy intensive. Draghi’s competitiveness analysis found that European industrial electricity prices were often around 2 to 3 times those in the US or China. Europe’s fragmented energy market and high, volatile prices also pose a constraint on investment and scaling.

At the same time, the EU does have a talent problem, although a very specific one. According to a 2025 analysis supported by the EC, EU institutions offer approx. 35% of AI-related master’s programmes globally, while the EU’s AI workforce more than doubled between 2016 and 2023. The problem here does not lie in the education, but in what happens to talent after training: where researchers can obtain the best compute, equity upside, research budget, and entrepreneurial environment. Capital, infrastructure, and talent are complementary; fixing the first two can materially improve the third.

Regulation does play a role here as well, but it did not create the historical gap with the US. EU’s technology gap predates the AI Act by many years. Europe had already failed to produce equivalents to Google, Amazon, Microsoft, Meta, or major cloud hyperscalers long before the AI Act was adopted. As such, it could be said that the regulatory complexity can exacerbate an existing scale disadvantage, but capital, market structure, compute, and commercialization are deeper historical causes.

So what can be done?

As the issue did not come from a single cause, you cannot treat it with a single “cure”. The key is to attack several bottlenecks simultaneously so that capital, compute, talent, demand, and market scale reinforce each other.

Among the things that could be done, we find the following most plausible: firstly, scaling up capital through funding large pan-European growth funds, and unlocking pension/insurance capital for VC and growth equity. Secondly, enacting an EU-wide corporate regime, with common company formation, options, and digital administration. Next up, completing AI gigafactories quickly and guaranteeing startup/research access rather than concentrating capacity in the incumbents.

Energy also needs improvement, and this could include accelerating grid buildout, generation, interconnectors and data-centre permitting. Cloud also needs to be tackled by creating conditions for competitive European cloud/AI infrastructure without forcing inefficient autarky.

Semiconductors are also an area of interest. EU should focus on realistic European advantages, i.e. equipment, advanced packaging, R&D, specialized AI accelerators and strategic foundry capacity. Talent’s position could also be improved through better researcher visas, startup equity taxation, university founder pathways and AI education.

There are other areas that can be improved as well, such as public procurement: by using defence, healthcare, energy, and administration as coordinated launch customers, industrial data could be used to create secure sector data pools across industry, etc.

In other words, there is a lot to be done, but it’s not impossible; it just requires will and direction. Lacking this, Europe will continue to fall behind the US and China, something that can become a deadly game if continued for long enough.