Artificial intelligence, production potential at 570 billion in Europe

The race is on and, according to the authors of the report 'Time to place our bets: Europe's AI opportunity', by McKinsey Global Institute, the task facing Europe is clear: focus on several key areas in order to be more competitive on the artificial intelligence front and catch up with the gap in the adoption of new technologies that organisations on the Old Continent complain about vis-à-vis their US counterparts. The opportunities at stake, it is well known, are enormous. And this is confirmed by estimates that AI could help the European ecosystem achieve an annual productivity growth rate of up to 3% by 2030, worth more than USD 570 billion.

To achieve this goal, however, companies and governments will have to support the design and development of cloud infrastructures and AI semiconductors and supercomputers at a faster pace. In fact, AI will accelerate not only the transformation of processes in entire sectors but also the operation of data centres where applications and data are hosted and processed (and where large language models are trained), bringing the demand for electricity needed to run them to weigh more than 5 per cent of Europe's total needs by 2030.

Environmental sustainability, as it happens, is one of the AI-related trends that Sas executives have outlined to describe the business and technology scenario that awaits us in 2025. The faster training of Llm models will have a direct repercussion in terms of energy, because - as Bryan Harris, Chief Technology Officer of the American company specialising in data analytics, observed - 'speed and algorithmic efficiency cannot be ignored as key levers for reducing the consumption of cloud resources'. Beyond the progressive adoption of alternative sources (such as nuclear power), in short, an important contribution to cutting consumption should come from the growing demand for less energy-intensive AI models. Sharing the responsibility for greening the large-scale adoption of generative technologies, again according to the vision of Sas managers, will however not only be the large cloud providers and hardware suppliers but also the end users of artificial intelligence, and in particular those who manage data and workloads. Greater efficiency in the development of AI models through platforms (of data and algorithms) optimised for computing in the cloud will, in other words, be the way to reduce duplication and unnecessary waste, while minimising consumption.

Finally, there is a further rationalisation and efficiency-boosting element linked to the acceleration of Ai (and simultaneously of the cloud), and it will closely affect IT infrastructures. If until now companies have operated with isolated systems, each dedicated to a different function or customer segment, the decisive turn towards the modernisation of existing IT resources in a flexible key will reward (in terms of speed, cost savings and data management capacity) those companies that bet on 'cloud-native' and AI-enhanced platforms capable of supporting multiple functions.

The conviction of executives, projecting towards 2025 and beyond, sees on the one hand the need to exploit Gen AI for business value creation and on the other hand the fact that, already within the next 12 months, large language models will become a commodity while open-source Llm will grow significantly. The consequence? The possible collapse of prices (since many basic functionalities will be available for free) and the shift of the true value of this technology towards specialised services and applications for individual sectors. An example? In 2025, as speculated by Sas, marketers will move from the simpler applications of generative AI (focused on productivity and content creation) to more advanced capabilities capable of delivering more personalised experiences and tangible revenue benefits. We are talking about synthetic data and digital twin, which will be deployed in close synergy with established technologies such as machine and deep learning.