Nokia chief says supply limits data centre construction

Justin Hotard says customers would build data centres twice as fast if memory and energy supply allowed. Nokia’s data centre equipment sales doubled to EUR 446M in the second quarter. His comments frame a supply constraint, while the scale of future demand remains contested.

Hotard’s account is that customers are prepared to build faster, but cannot obtain enough memory or energy. He said demand could continue to grow even if no new frontier model arrived for three years. This makes the question relevant to infrastructure planning as well as technology development.

Nokia’s sales reflect infrastructure demand

Nokia supplies equipment that connects racks within a data centre and links data centres to one another. The company’s sales of that equipment doubled to EUR 446M in the second quarter, against group sales of EUR 4.82B. The figures offer a measure of the business Nokia is already doing as customers expand computing infrastructure.

Nokia is headquartered in Espoo, while most of its customers are elsewhere. Its equipment is used in the connections between computing systems, rather than in the chips themselves. That position gives Nokia exposure to construction activity without making it a direct measure of the full cost of building data centres.

European support has a committed portion

The EU opened bidding in July for seven AI gigafactories valued at EUR 30B. About EUR 1B of public money is committed, according to the source account. The distinction matters for readers assessing announced infrastructure ambitions against money that has actually been allocated.

Nokia’s fastest-growing business sells into a buildout largely taking place outside Europe. The source gives no comparable European estimate for total data centre investment. It therefore does not establish how much of the wider construction opportunity will be located in Europe, or how much Nokia may supply there.

Demand and financing remain disputed

Hotard said existing technology could support substantial progress even without another frontier model in the next three years. The argument is that deployment of current systems may sustain demand for infrastructure. It is a statement of his view, rather than a confirmed forecast of future customer spending.

The source also describes contrary assessments of whether revenues will cover the infrastructure being built. Bain estimated that AI would need $6T annually by 2031, while products available today might generate $1.2T to $1.8T. A paper presented at the Brookings Papers on Economic Activity puts investment at $10.3T between 2025 and 2032, and warns that financing is shifting into joint ventures, private credit and special purpose vehicles.

Investment totals leave open questions

The account identifies memory and energy as the constraints Hotard named, but supplies no timetable for easing either shortage. It also does not give a forecast for Nokia’s equipment sales beyond the second quarter. The difference between planned construction and the revenue available to fund it remains an open issue.

The EU bidding process and its committed public contribution provide one policy reference point, while Nokia’s sales show current commercial activity. The source does not describe a regulatory decision or include a separate response from Nokia to the doubts about future demand. Further evidence would need to show whether supply expands, projects proceed, and revenues meet the financing needs described.