How the AI Boom Could Trigger a New Dutch Disease
The technology sector’s debt-fuelled expansion risks squeezing investment elsewhere in the US economy. Traditional remedies offer little help. A commentary by Ignazio Angeloni
As Wall Street prepares for Anthropic’s blockbuster flotation, with $2 trillion in value created in just over five years and a potential market estimated at $40 trillion, artificial intelligence continues to inspire a mixture of enthusiasm, fear and doubt. Recently, fear and doubt have gained ground over enthusiasm.
Progress will not stop, that much is clear, and AI is outstripping all existing computing technologies. But attention is increasingly turning to its possible downsides and ways to address them across a wide range of fields: employment, security, culture, education, scientific research and beyond.
One aspect has received relatively little attention: the effects on the rest of the economy. A sectoral boom does not remain confined to the sector itself. Once it reaches a certain scale, it sets in motion direct and indirect mechanisms that reverberate throughout the economy and engage macroeconomic policy, whether fiscal, monetary, trade-related or otherwise.
Beyond a certain point, one sector’s success can put others under strain, undermining a structural balance that depends partly on economic diversification. Industrial policy, normally distinct from macroeconomic policy in both its objectives and its instruments, ends up intersecting with it.
For those old enough to remember, all this recalls “Dutch disease”, a subject that has largely faded from economic debate but was a dominant concern in the 1960s and 1970s. The reference to the Netherlands is not entirely justified, since the phenomenon extends far beyond the country, but the label has stuck.
In 1959, natural gas deposits were discovered in the plains around Groningen, in the north-east of the Netherlands. State revenues surged, while households and businesses benefited from abundant, cheap energy. The windfall was not an unqualified blessing. The Dutch guilder, before the euro existed, appreciated, eroding manufacturers’ competitiveness in an economy highly open to trade.
The external accounts remained healthy thanks to the energy sector, but the domestic economy was in dire straits: high costs, unemployment, a rising cost of living, industrial decline and the resulting erosion of physical capital and know-how. When Britain discovered oil in the North Sea a few years later, the problem spread across the Channel. The Economist identified the problem and coined the term. As the Netherlands began to recover, Britain entered a period of economic distress that, together with Margaret Thatcher’s free-market policies, plagued the country until the mid-1980s.
Before drawing a partial parallel with developments today, it is useful to recall the remedies economists proposed and some governments adopted at the time. The first was to insulate the energy sector from the rest of the economy by saving its revenues to reduce their expansionary impact. This is an appropriate response if the shock is considered temporary and it therefore makes sense to build up reserves. Some energy-producing countries, including Norway and Saudi Arabia, have followed this path.
The second strategy involved defending non-energy sectors through subsidies and other measures. The Netherlands itself pursued this approach, as did Britain before Thatcher.
The US, today’s main centre of AI development, and its Californian microcosm, itself hardly small, are experiencing developments that in some respects resemble those triggered by Dutch gas, although the mechanisms differ.
The variable transmitting these effects to the wider economy is not the exchange rate: over the past two years, the dollar has depreciated rather than appreciated, and for different reasons. It is interest rates, especially at longer maturities, which have been driven up by the mountain of borrowing generated by the high-tech sector. Outside technology, high rates discourage investment and growth.
Data published by Oxford Economics document a sharp divergence in US investment: robust in information technology, to finance data centres and other AI infrastructure, but stagnant or falling elsewhere.
The boom’s impact on the wider economy is limited by the information technology sector’s very low demand for labour. Investment returns are astronomical but concentrated in very few hands, and therefore do not generate aggregate demand.
The slowdown in investment, growth and innovation in other sectors remains, and its effects are substantial. The economy becomes polarised and less diversified.
The process is still in its early stages. Where will it lead? Its direction is relatively clear; its scale and eventual outcome are much less so.
Traditional remedies seem ill-suited to what is happening today. Dutch gas revenues flowed largely to the state and could temporarily be set aside. Today, the gains accrue to a small group of individuals in the form of capital gains on equity holdings that are not yet publicly listed. The sovereign wealth fund route is unavailable.
Protecting manufacturing, exposed to competition and hurt by financial conditions, remains possible and is probably desirable. Might Trump’s tariffs and the dollar’s depreciation, encouraged in various ways by his administration, therefore have an economic rationale?
Unfortunately, this ex post justification does not stand up to scrutiny. In their sectoral and geographical distribution, as well as their scale, the tariffs announced on the so-called “Liberation Day” bear little relation to a rational strategy of containment and rebalancing.
By contrast, the rise in borrowing costs that is choking much of the economy has a great deal to do with this administration’s unpredictable policies.
A previous version of this article was published in the Italian daily Il Foglio
IEP Bocconi does not express opinions of its own. The opinions expressed in this publication are those of the authors. Any errors or omissions are the responsibility of the authors.
