The AI Sovereign Wealth Fund: A Global Mirage?

Examining AI-induced inequality on global level reveals why distribution may not be the answer.

Source: Financial Times, © Jason Redmond/AFP/Getty Images

An AI Sovereign Wealth Fund

Recently, Sam Altman proposed allocating a 5% equity stake in OpenAI to the Trump administration; the initial step to distributing AI-generated wealth among the American public through a newly established US sovereign wealth fund. Under this proposal, other leading AI development labs would be invited to follow suit, contributing their own equity to the public fund. The holdings within the fund would be invested into financial markets, with the dividends then shared amongst the general population.

Whilst an honourable proposition, questions remain regarding the effectiveness of such a programme. It seems unlikely that each person will receive an income proportional to the aggregate riches AI will generate, meaning the inequality persists.

For the transfer to be meaningful, it would require co-operation on a similar, if not larger, scale from the other technology companies; something they may be unwilling to do, especially if the contribution rises from 5% and the return from huge investments are taking time to materialise.

A feasible alternative is presented in the Sanders Bill - legislation designed to give the public a direct ownership stake in the largest American AI companies through a one-time 50% tax on their stock, which would be deposited directly into a sovereign wealth fund. 

The influence obtained through these massive stock holdings creates an interesting dynamic: a commission would be established to use this voting power to steer corporate decisions in the best interest of the general public, resulting in a significantly higher benefit to the public compared with the voluntary initiative. Furthermore, given that no other AI companies have signed onto Altman’s proposal, and that this bill has been formally put forward by a sitting senator, the Sanders model appears more politically viable.

Details aside, the fact that the idea has been raised on multiple fronts speaks volumes; it is a clear recognition that without direct intervention, the gains of AI will inevitably concentrate and divergence will ensue.

However, this remains a discussion internal to America; yet to be heard elsewhere. Operating at the technology’s frontier, the US is already well-positioned to maximise the public benefits of the AI transition. Most nations are not so fortunate.

Therefore, what about the disparity AI may generate on a global level?

An AI-amplified Global Disparity

In their 2021 paper, Korinek and Stiglitz warn that AI could severely impact developing nations by driving a wave of innovation that is simultaneously labor-saving and resource-saving. This combination effectively strips the value of the structural advantages - like abundant cheap labour and rich natural resource deposits - that have fueled the growth of emerging economies over the past half-century. Because AI-driven automation compresses global demand for human labour while allowing manufacturers to generate more output with fewer raw inputs, it fundamentally disrupts the comparative advantage held by emerging economies.

For instance, as AI-driven automation increases efficiency and shifts industrial demand away from traditional commodities toward rare earth elements and advanced synthetic materials, lower demand means traditional exporters like Zambia and Angola could face a devastating deterioration in their terms of trade. The rapid deployment of generative AI across Western corporate infrastructure is already repatriating low-skill labor; when an AI assistant can instantly resolve customer service queries or manage back-office logistics, the economic incentive to offshore these operations to emerging markets vanishes.

Modifying a natural competitive edge to fit a changing environment is difficult for any nation; especially for those that physically and financially lack the means to do so. This indicates a period of economic reversal, as developing countries struggle to adapt to a globalised world that doesn’t necessarily require them in the chain, threatening their ability to sustain economic progress and ‘catch-up’ with the developed world that is benefiting from the technological progress that harms them. 

Whilst the situation for developed countries is not so dire, the implications for relative economic standing reveal another dimension of widening disparity.

The Concentration of AI development

The principal reason why an equitable, global diffusion of AI’s potential gains is so difficult is because the technology’s development is essentially harboured in two locations: the US and China. And within these countries, the development is concentrated further within a few ‘superstar industries and companies’ ; the asymmetry that motivates proposals such as the sovereign wealth fund.

As Korinek and Stiglitz suggest, this concentration threatens to turn these leading nations into "superstar countries" that reap all the economic rents of AI development, leaving the rest of the world behind as a monopolisation of knowledge blocks the traditional economic catch-up process. 

Yet, while this monopolisation of knowledge is a valid concern, equally pressing is a lack of infrastructure and capacity to actually use that knowledge once it trickles down. If nations lack the necessary physical and social frameworks to effectively integrate the technology into their existing economic systems, the resulting costs and frictions contribute to development lags, further widening the gap between countries that have adaptive capacity and those that don’t.

Reactive is the New Proactive

The impact of technological change is highly dependent on pace and scale. As recursive self-improvement becomes a reality and an unprecedented level of investment fuels innovation, AI development is bound for exponential, rather than linear growth, with the gap between countries unfolding in a similar manner.

Domestic policies are futile when the problem is between, rather than within, borders. The utopian solution would be something like a supranational level technology and wealth transfer scheme. But, the multilateral governance and co-operation required for this to be possible is a distant dream.

Source: abcNews, ‘Your countries are going to hell’: Trump bashes United Nations, world leaders in speech

Being locked out of the frontier and reliant on diffusion yields a natural conclusion for non ‘superstar countries’: the need to hone adoption-led innovation. Being able to tailor technologies to suit domestic purposes is a major step towards benefiting, rather than losing out from developments in AI. Prioritising this is crucial to building the adaptive capacity needed to thrive in this era.

The disruption to the comparative advantage of developing countries is a more difficult issue. However, the time to start considering it is now, before the impact has fully materialised. Global support to help these countries begin to shift to less exposed sectors would go a long way; as would a period of trade guarantees and preferential agreements, easing the friction as they attempt to adapt.

Building economic resilience and absorptive potential is a far better approach to distributing the gains from AI; perhaps there is a lesson to be learnt here for the US and its sovereign wealth fund.