Special Analysis
The global AI market is entering a new phase of development in the first half of July. If the past two years were defined by a race for ever more capable models, the past week shows that competition is rapidly moving to another level: infrastructure, energy, specialised hardware and the operational application of AI in business. AI is increasingly seen not as an individual software tool but as strategic infrastructure that will determine the long-term competitiveness of companies and countries alike.
Several developments over the past week confirm this shift. The United Kingdom is accelerating the development of AI data centres, South Korea is investing hundreds of billions of dollars in semiconductor manufacturing, and new AI processor manufacturers are emerging, while the focus on energy sustainability, security and the governance of AI systems continues to grow. The market is entering a phase of operational discipline, in which the value of the technology is measured by business results rather than by demonstrations of what a model can do.
AI infrastructure becomes a strategic advantage
One of the most significant trends is the growing importance of physical AI infrastructure. Over the past week the United Kingdom announced accelerated approval for the construction of new AI data centres through simplified regulatory procedures, with the aim of increasing domestic computing capacity. At the same time, South Korea continued to implement an investment plan worth around 590 billion dollars, intended to consolidate its position as one of the world's leading centres for the manufacture of semiconductors and AI chips.
In parallel, companies such as FuriosaAI are introducing a new generation of AI accelerators designed for running large language models, showing that the market for specialised hardware is no longer confined to a handful of dominant suppliers.
These developments confirm that competition is no longer conducted solely at the level of models, but increasingly at the level of the infrastructure that runs them. Computing power, energy efficiency, cooling systems and the choice of hardware architecture now directly affect the security, scalability and long-term sustainability of AI systems. For companies, this means that infrastructure decisions are becoming as important as the choice of model itself.
The new arithmetic of AI investment
Alongside infrastructure development, the way AI investment is planned is also changing. A growing number of investment funds are directing capital towards projects that develop AI capacity and energy infrastructure at the same time. A further signal came from the Singaporean fund Temasek, which announced a significant increase in AI investment, with particular focus on the energy infrastructure that will support the growth of new data centres.
As the number of AI agents and automated processes grows, the total cost of running them day to day becomes as important as the quality of the model. Organisations will therefore increasingly assess AI projects through total cost of ownership, which includes computing power, energy, maintenance, security and the ability to scale over the long term.
Business value is determined by application, not by technology
An important lesson comes from the Chinese market. ByteDance introduced a subscription for Doubao, the most widely used AI application in China, and it lost 6.1 million users. They moved to Qwen, which remains free because Alibaba covers the losses from its cloud business. A week later, new regulation shut down companion features entirely on Doubao, Qwen and Yuanbao.
Users did not reject AI. The outcome was determined by price and by regulation.
The labour market is changing too
The changes are increasingly visible in the labour market as well. While traditional outsourcing models are slowing, demand is growing for specialists who understand business processes and know how to integrate AI into an organisation's existing systems. Roles are appearing more and more frequently whose task is not to develop models from scratch, but to connect them to a specific business environment.
Two hundred economists and AI researchers, among them sixteen Nobel laureates, have warned that AI could reshape the economy faster than the Industrial Revolution, and called for safeguards to ensure that it complements human labour.
A further challenge lies in the security risks of autonomous AI systems. The GhostApproval research demonstrated how AI agents can be induced to bypass existing approval and control mechanisms, confirming that the management of identities, privileges and AI agents is becoming a new field of corporate risk management.
Take away
Seen individually, the past week's developments concern different subjects, from data centres and chips to the labour market and the security of AI agents. Together, however, they point to the same conclusion: AI is entering a phase in which technological superiority is no longer the only source of competitive advantage.
For organisations planning a more serious application of artificial intelligence, the key questions are no longer confined to the choice of model or supplier. Equally important are the availability of infrastructure, the total cost of running AI systems, the quality and sovereignty of data, regulatory compliance, and the organisation's ability to develop internal competence in managing AI.
In other words, AI is no longer a product that is bought. It is also infrastructure that has to be planned, built and governed.