Notebook: Europe’s Sovereign AI Needs Customers, Not Just Money
Despite record venture funding for European AI startups, experts at the HumanX conference emphasize that capital alone is not enough. To achieve true sovereign AI, European enterprises and governments must actively purchase local technology solutions.
While European AI startups are securing record levels of investment, capital alone will not guarantee the region greater command over its artificial intelligence destiny. Discussions at the recent HumanX conference in Amsterdam highlighted a critical missing link: large enterprises and governments must also purchase the solutions these startups are creating.
Data from a joint report by Crunchbase and HumanX reveals that European AI-focused startups pulled in $23 billion during the first half of 2026. This marks a 130% year-over-year increase and represents 55% of the region’s overall venture funding. This investment boom brings the core debate of sovereign AI into sharp focus: Where ought nations to direct their resources to secure data control and capture the economic value of the technology?
Achieving sovereignty, however, does not demand ownership of every single tier of the AI stack. This insight was shared during an onstage panel discussion by Axelera AI founder and CEO Fabrizio Del Maffeo and AI71 chief product and technology officer Mehdi Ghissassi.
Representing perspectives from Europe and the United Arab Emirates respectively, their insights pointed toward promising opportunities in applications and specialized chips. They further emphasized that government mandates, purchasing choices, and compute access ultimately dictate whether AI ambitions turn into tangible growth.
Five-layered cake
Nvidia CEO and President Jensen Huang has previously outlined a “five-layer cake” required for delivering artificial intelligence, encompassing energy, chips, infrastructure, models, and applications.
For Del Maffeo—whose firm builds inference chips—the primary opportunity rests within the semiconductor layer, especially as AI processing shifts away from centralized data centers and out toward physical devices.
“Artificial intelligence will expand from cloud computing, from centralized data centers, to devices closer to us in the physical world,” he explained. “To enable this, you need specific chips which can run efficiently, at a lower cost, to connect these networks that today are running in the cloud. We are here to solve this problem.”
Now five years old, Axelera AI supplies two chip generations to roughly 600 clients. The business is currently preparing to branch out into decentralized cloud computing products.
AI transformation
Headquartered in Abu Dhabi, AI71 collaborates with government agencies and major organizations to help guide their AI transformations.
According to Ghissassi, the UAE possesses the world’s highest compute capacity per capita alongside a strict three-month mandate requiring all government bodies to establish agentic processes for citizens. Energy resources there are cheap and plentiful. The state aims to have half of all citizen-government interactions managed by an AI agent within the next two years, he noted.
Ghissassi argued that attempting to compete at the foundation model layer is impractical due to the massive capital requirements and rapid commoditization, leaving room for only two, three, or four major players worldwide.
Instead, Ghissassi—who formerly directed product development at Google DeepMind—stressed that the application layer is vital for data sovereignty. “You want to make sure that it stays with you, be it that you’re a government or an enterprise. If you’re giving away your trade secrets and know-how, nobody stops whoever is being a provider to you today, from replacing you.”
In the UAE, “what helps is the mandate, the pace at which things happen, the availability of compute, both in terms of sovereign clouds, or on-prem, or global clouds, and then the amount of capital that is being invested to help transform these companies to benefit from these technologies,” Ghissassi remarked.
Conversely, Europe faces headwinds as a net importer of energy. While it maintains capabilities in semiconductors, it lacks the full computing scale required for advanced neural networks and frontier labs.
“We should not be obsessed with controlling the entire stack,” Del Maffeo stated. He added that Europe must focus on generating value rather than merely paying for services. “Creating value means creating a wealthy economy.”
Europe’s core advantages feature vast research talent pools and a population of 440 million potential AI beneficiaries. Nevertheless, Del Maffeo pointed out that large corporations lack an established culture of buying from startups, depriving the region of the growth flywheel seen in the United States.
“What worries me is that we are a little bit lagging behind, and therefore we are missing this value creation, and this will weaken the economies of Europe,” he warned.
Related Crunchbase query:
- Europe AI Funding In 2026
Related reading:
- Crunchbase & HumanX 2026 European AI Economy Report: Funding, Innovation and Growth
- Europe Posted Its Strongest Venture Funding Quarter In 4 Years As UK Gains, M&A Holds Up
Illustration: Dom Guzman
Frequently Asked Questions
01What is sovereign AI and why is it important for Europe?
Sovereign AI refers to a country or region’s capability to control its own artificial intelligence infrastructure, data, and technology stack. For Europe, it is viewed as vital for retaining economic value, protecting sensitive data, and preventing over-reliance on foreign technology providers.
02Why is funding alone not enough for European AI startups?
While European startups are raising record sums of capital, they lack a strong domestic customer base. Without large businesses and governments actively purchasing and adopting their products—similar to the commercial flywheel observed in the U.S.—startups struggle to scale and create lasting economic value.
03What are the key layers of the AI stack?
According to industry leaders like Nvidia CEO Jensen Huang, the AI stack consists of five layers: energy, chips, infrastructure, models, and applications.
04Why do experts suggest Europe shouldn’t try to control the entire AI stack?
Building and maintaining frontier models and massive compute infrastructure requires extreme amounts of capital and abundant energy resources. Industry executives argue that Europe should instead focus on areas of comparative advantage, such as specialized semiconductor chips and specific application layers.
