What is an AI factory?
While the first industrial revolution brought us coal-fired factories to make work more efficient and the telegraph to empower wider communication, this latest revolution is spurred by the most computationally demanding task to ever face humanity – generative AI. Generative AI enables users to quickly create new content based on a variety of inputs, such as text or images. Because of the massive amounts of data this entails, our current computing infrastructure simply won’t suffice. European nations must prioritize the creation of sovereign AI infrastructure to meet demand. In practice, this means the creation of AI factories.
At a basic level, an AI factory is where data comes in and intelligence comes out. It’s an entirely new generation of data center that uses a full-stack accelerated computing platform to perform the most intensive computational tasks. Much like heavy machinery is needed to refine raw materials into more useful resources, substantial computing power is required to turn enormous amounts of raw data into intelligence. The AI factory will become the bedrock of modern economies across the world.
Currently, the world’s most powerful supercomputers are clustered, with the majority of AI computing power in prestigious universities, research labs and a handful of companies. This landscape prevents many nations from creating generative AI that takes advantage of valuable local data to understand the local language and its nuances. The Future of Compute Review, commissioned by the UK Government, found that for the UK to project its global power as a science and technology leader, it needed to ensure its own sovereign computing capability.
Cooperating with national champions
The sovereign AI race is already underway. Japan, India and Singapore have already announced plans to construct next-generation AI factories. While these countries are enjoying a head start, the race is far from over. Real progress is already starting to be made in Europe, as the European Commission has recently announced its support for a network of AI factories.
However, governments are unable to power this new industrial revolution alone. Generative AI development on this scale requires vast resources in material wealth and technical skills, so partnering with the private sector will be critical to success. Every country already has its own strong domestic sector, filled with local technology champions. Making the most of their expertise and capabilities is the first step to success.
The telecommunications industry is one such industry that is well-positioned to support generative AI infrastructure efforts by evolving into AI factories. Leading telecom operators, such as Orange in France or BT and EE in the United Kingdom, are trusted service providers with large in-region customer bases. The demands of the telco industry have prepared these companies to effectively assist the generative AI infrastructure revolution. Telcos are already used to intensive investment and infrastructure replacement cycles, such as recent rollouts of 4G and 5G solutions. Moreover, they have access to secure, high-performance distributed data centers located close to large metropolitan areas, which helps to combat latency issues.
If Europe is to sit in the driving seat of the latest industrial revolution, rather than just be a passenger, European countries must make AI infrastructure investment an absolute priority.
A new understanding of sovereignty
Although we are in the midst of a generative AI boom and interest keeps growing, development and deployment tools remain limited in terms of their accessibility. Most, if not all, of the most popular AI tools are primarily available in the English language. In a geographical area as culturally and linguistically diverse as Europe, AI tools need to be accessible to all – not only those who happen to speak English. Making this a reality means using local data, implementing local languages and, most of all, bringing the translation capabilities to do so within one’s own borders.
Changing the perception of sovereignty to include computing power is no small feat and is certainly not achievable without action. The shift to sovereign data centers both preserves cultures and native languages in AI tools and ensures that GenAI applications can function accurately within their specific context. But it will require generational investments and ongoing support. The AI infrastructure that tomorrow’s economies will be built upon simply does not exist yet, and those who begin building first will stand to have the most to gain.
Vice President of Worldwide AI Initiatives at NVIDIA
Securely store your business data with the best business cloud storage.
