According to Nokia, demand for new AI data centers exceeds what the market can currently supply. It is not a lack of interest, but shortages of resources such as energy and memory chips that are limiting the pace at which new capacity becomes available.
Nokia CEO Justin Hotard states that customers would likely build data centers twice as fast if the necessary resources were available. In an interview with CNBC, he therefore says he sees little merit in the idea that too much AI infrastructure is currently being built. Rather, supply-side constraints are obscuring just how much capacity companies actually want to add.
This development is relevant to Nokia because the company is increasingly focusing on infrastructure for AI data centers. It provides network technology for communication between systems within a data center, as well as for connections between separate data center locations. As a result, the company benefits from a market in which computing chips, memory, network capacity, and electricity are key drivers of further growth.
Energy becomes a limiting factor
In particular, energy availability is increasingly becoming a barrier to data center expansion. AI systems require large amounts of electricity, while new power connections are not readily available everywhere. At the same time, suppliers are facing shortages of certain types of memory required for AI accelerators.
Hotard does not view these constraints as an indication that the investment wave has peaked. In his view, the AI rollout is still in its early stages. This contrasts with the growing debate over whether the billions invested in data centers will ultimately yield sufficient returns.
These amounts are substantial. Research presented at the Brookings Institution estimates AI infrastructure investments between 2025 and 2032 at $10.3 trillion. In addition to internal funds, companies are using debt and joint ventures to finance this. This increases dependence on sustained high demand for AI services.
Existing AI is already sufficient for growth
Hotard also believes that the need for infrastructure will not disappear immediately if the development of new AI models slows down. In recent weeks, this has once again become a topic of discussion following warnings about the risks of increasingly powerful models.
According to the Nokia CEO, however, significant capacity is still needed to deploy existing technology at scale. Even if no new generation of frontier models were to emerge for three years, he believes the application of current models could still yield substantial growth.
As a result, the discussion about AI infrastructure is shifting in part from the models themselves to the physical conditions under which they can be deployed. The availability of power, memory, and network capacity may ultimately affect the pace of construction more than companies’ willingness to invest in new data centers.