Google plans €13bn Finnish AI infrastructure programme as environmental review pauses site work

Google has announced plans to invest €13 billion in Finnish AI infrastructure in 2027–2028. The planned programme covers the development of data centers at three locations in central and northern Finland, alongside investment in Google’s existing facility in Hamina.

The investment remains a forward plan rather than confirmed expenditure. No capacity figures in MW or GW, commissioning dates, or allocation of the €13 billion between the new locations and Hamina have been disclosed in the available material. The full list of the three proposed new locations has also not been identified.

At two of the projects, in Kajaani and Muhos, Finland’s licensing and supervisory authority has ordered a halt to parts of the preparatory works until the required environmental impact assessment has been completed. The reported action concerns preparatory activity, not cancellation of the wider investment programme or all work across the portfolio.

According to the Finnish Nature Conservation Association, nearly 200 hectares had been cleared at Kajaani and about 330 hectares of forest felled at Muhos. Those figures have not been independently documented in the material available for this report, and no Google response on the reported cleared areas is included.

Google has also signed a 22-year power purchase agreement with Fortum, covering up to half of generation from the Loviisa nuclear power plant. Fortum said the agreement would support continued operation of the plant through 2050. For data-center operators, the PPA illustrates the role that long-duration power procurement can play in underpinning large AI infrastructure plans where reliable, lower-carbon electricity supply is central to site selection.

The planned construction phase in 2027–2028 is expected to support more than 37,000 jobs and contribute an average of €3.6 billion annually to Finland’s economy, according to the source. These are projected construction-phase effects rather than operating metrics for the planned data centers.

More broadly, the case suggests that Finland’s energy profile and ability to support long-term power contracts could remain attractive to AI workloads. However, it also indicates that environmental assessment, land preparation and local regulatory processes can affect deliverable schedules for large-scale data-center campuses even where the power proposition is strong.