Imports of office equipment and computers totalled ISK 44bn in both 2022 and 2023, ISK 93bn in 2024, and ISK 171bn in 2025, or about 3.5% of GDP. The figures cover all computers, but growth is attributable for the most part to equipment for data centres. In a small economy, a single project can profoundly affect such figures: In 2025, atNorth disclosed that the company’s customers invested in equipment in Akureyri for nearly ISK 100bn, or 2% of GDP for the year. In the second investment cycle, computer importation came to ISK 67bn in Q2/2026, more than in any single quarter since 2015.

First comes the equipment …

Investment figures are affected first. In the August issue of Monetary Bulletin, the Central Bank (CBI) projected that general business investment (excluding energy-intensive industry, ships, and aircraft) would grow by more than 6% this year. If computer equipment for data centres is set aside, however, that number falls to just over 1%. As these numbers indicate, growth is due almost entirely to data centre computers.

The chart, which is based on the CBI’s most recent figures, shows how strongly this can affect the statistics. From end-2024 through mid-2025, general business investment grew by 16-21% but would have contracted by 7-18% if data centres’ computer equipment were excluded. The pattern reversed at the end of 2025, with a contraction in general business investment despite an increase in underlying investment.

The equipment is recognised both as investments and as imports. In GDP data, the two items offset one another, as we discussed here, so the equipment has far less impact on GDP growth than the investment figures alone would indicate. On the other hand, the current account balance measures external trade. Imports show in the current account as expenditures, but the investment in Iceland does not offset them, and the current account deficit widens by the amount of the imports.

Iceland’s CA deficit measured 9.3% of GDP in Q2/2026, its largest since autumn 2008, if the failed banks’ calculated interest expense is set aside. But the data centre equipment makes the CA deficit look worse than the underlying position warrants: The CBI estimates that excluding data centres, the current account would be broadly in balance for 2026 as a whole, instead of showing a deficit of just over 3%, as both we and the CBI project. Furthermore, the deficit does not weaken the ISK, as the data centres’ foreign owners and customers pay for the equipment and, as we have discussed in the past, no foreign currency is exported from Iceland in connection with the transactions. The deficit shows as a debt owed to the foreign parent company, which increases Iceland’s external liabilities accordingly.

… and then come the revenues

The second part of the cycle involves the exportation of other computer services, which the CBI links to data centre activities. These exports totalled ISK 8-9bn in H1/2025 but have averaged about ISK 21bn per quarter since Q4/2025. In H1/2026, the exports in question grew by nearly ISK 25bn year-on-year, while total export revenues grew by ISK 20bn. In other words, the data centres accounted for all of Iceland’s export growth and more besides, as other exports contracted slightly.

The intellectual property industry’s share in exports has grown by leaps and bounds. It measured 11% in 2018 and, according to our macroeconomic forecast, will exceed a fifth of total exports by 2028. Growth has been particularly rapid in ICT services, which have surged 65% YoY in the past twelve months. Much of this is due to data centre development, as we have discussed recently.

The revenues boost GDP growth, but their impact on GDP will taper off when exports stop increasing, even if activity in the sector holds its ground. YoY growth in other computer services measured 159% in Q2, but if revenues hold steady, the growth rate will fall to 32% in Q3 and turn negative in Q4. Slower export growth in the next set of figures will not necessarily indicate that data centre performance is weakening.

Who owns the data centres, and where do the revenues go?

The cycles described above are largely temporary. The lasting benefit lies in the revenues that remain in Iceland, and this is where ownership is important. Export revenues from data centres totalled about ISK 80bn in the last four quarters, and the full amount is included with GDP, as the service is produced in Iceland. That said, the revenues revert to Icelanders only in part. The data centres are foreign-owned to a great extent, and a large share of the equipment is owned by foreign AI companies. In Akureyri, atNorth has invested for ISK 12-13bn, whereas its customers have invested in equipment to the tune of nearly ISK 100bn.

The profit belonging to the foreign owners is therefore considered their revenues and not Icelanders’ revenues, and as a result, it is deducted from national income but not from GDP. In the current account, the profit is recognised as expenditures, no matter whether it is exported from Iceland or reinvested here.

In this respect, the data centres are in the same position as the aluminium smelters, which are wholly owned by foreign entities. The difference is that the smelters import raw materials for their production on an ongoing basis, while the data centres mainly import equipment for the investment, usually in big waves. Foreign inputs are therefore a negligible part of data centre activities unless the companies are scaling up their operations.

Development in data centres is roughly of the same magnitude as development in aluminium smelters two decades ago. Investment in data centres and other information technology accounted for about 16% of all investment in 2025, while investment in aluminium manufacture peaked at 15% in 2006. The upward cycle is quite different, however. The aluminium smelters required massive construction in Iceland, including the Kárahnjúkar Power Station. Investment in data centres, on the other hand, is limited to imported equipment, which generates little activity in Iceland. Investment in energy and utilities is still less than half of what it was in 2006.

What stays behind in Iceland?

What the data centres leave behind in Iceland is primarily the money they pay to domestic entities: they buy electricity, pay wages, buy domestic services, and pay public levies. Presumably, electricity is the largest of these items. We estimate that when the data centres’ additional service capacity is fully operable, these items will return tens of billions of Icelandic krónur per year to the domestic economy.

This is the long-term benefit from the data centres. It is part of the ISK 80bn in export revenues over the past four quarters but is dwarfed by computer imports, which totalled ISK 170bn in 2025. The big numbers that have affected the statistics exaggerate the benefit to Iceland, but nevertheless, in an economy the size of ours, tens of billions per year make a difference.

The amount that remains in the country depends not least on energy. According to the Federation of Icelandic Industries, the data centres used an average of 70 MW in H1/2026, or about 3.7% of Iceland’s electricity consumption. This is small in comparison with the metals industry, which used about twenty times that amount. The new data centre that Verne plans to build in Reykjanesbær will consume 120 MW all by itself, however, if the project materialises. But energy supplies are limited, and capacity constraints pose the greatest barrier to further data centre growth.

What comes next?

We assume that data centre development will taper off in 2027. According to our macroeconomic forecast, business investment will contract by 3.2%, not least because of that slowdown. On the other hand, equipment imports will subside, so the contraction in investment does not necessarily mean the economy is weaker. The CA deficit will shrink from 3.3% of GDP in 2026 to about 1% in both 2027 and 2028.

Three main factors will determine how long the data centres will affect the figures.

• First is the supply of energy: the CEO of atNorth said this spring that the company could doubtless attract more customers if more energy were available.
• Second is AI-related demand, as a large share of the equipment is owned by AI companies. That demand is not a given: The CBI has pointed out that it is uncertain whether massive investment in AI and high expectations about its future will prove warranted.
• Third is equipment renewal, as hardware becomes obsolete quickly, so computer imports could spike again without further expansion of the industry.

Headlines about record-breaking investment, historically large CA deficits, or abrupt turnarounds in exports should be interpreted with caution. Underlying the cycles is the overriding picture we sketched out in our macroeconomic forecast: growth will be relatively subdued but will gain pace, measuring 1.2% in 2026, 1.8% in 2027, and 2.3% in 2028.