SCOTT TIMCKE: AI bubble? Big promises and bigger questions

Backed by a projected $1-trillion in capital expenditure, the push into artificial intelligence (AI) raises questions about its economic impact. Investors must demand clarity on how this capital allocation will translate into tangible returns.

The AI growth debate

AI promises to revolutionise business processes across several industries, from customer service to computer-assisted manufacturing. Yet as tech giants pour vast sums into AI infrastructure — chips, data centres and power supplies — the economic payoff remains speculative.

Proponents of AI tout potential cost savings of up to 60%, but such claims warrant scepticism. Academic economic forecasts paint a more modest picture, suggesting AI might contribute just 1% to GDP growth over the next decade in advanced economies. For emerging markets such as SA, with different economic structures, the benefits may be even less clear-cut.

Current estimates indicate that only about 5% of tasks — the most cost-effective ones — are prime candidates for AI automation. Complex tasks are proving more resistant to AI adoption, which should temper expectations of transformative change in the near term.

The promise of AI as a low-cost solution for business challenges another illusion. Once realised, efficiency gains are likely to be competed away, leaving firms scrambling for sustainable revenue growth. 

Considering the enormous investments flowing into AI one must ask what problems will it actually solve? What are the real goods being produced with all this investment? And are shareholders prepared to let executives bet entire business models on the hope that middling AI will improve with more spending on data and computing power?

The semiconductor squeeze

Compounding these challenges is the looming semiconductor bottleneck. Nvidia’s high valuation not only reflects extreme demand for its purpose-built AI chips, but also the limitations of production and supply. If AI continues to grow rapidly, the hardware needed to support it may not be able to keep pace, potentially slowing down AI adoption.

Further complicating matters is the geopolitical chess game unfolding around chip manufacturing. Major Western powers, wary of supply chain vulnerabilities exposed by recent global crises, have embarked on aggressive neo-mercantilist policies. These initiatives aim to repatriate or “friend-shore” semiconductor production, insulating them from great power competition. 

While such moves may bolster national security of Western countries, they risk fragmenting the global semiconductor market, potentially driving up costs. How might these factors shape African AI?

Powering the AI revolution

The energy demands of AI present another significant hurdle. The stepped-up power requirements of AI systems and data centres represent an increase in demand not seen for decades. This poses particular challenges for countries such as SA, where the power grid is already strained. 

There is a risk that the pursuit of AI could worsen existing crises. The energy demands of AI systems and data centres represent a significant increase in power consumption, raising concerns about climate impact. How will countries such as SA, already grappling with an ageing power grid, meet this surge in demand while pursuing clean energy goals?

This matter must be central to any “just transition” energy discussions, especially if the shift to AI threatens traditional careers and incomes. Substantial investments in infrastructure will be necessary, but these will take time to come online and will face fiscal and supply constraints.

A measured approach

As African countries rush to develop AI strategies, there is a risk of these becoming part of the iconography of a contemporary administrative state, regardless of their efficiency, effectiveness or enforceability. Indeed, few countries seem to be considering AI as a multi-asset strategy, or fully grappling with the challenges of operationalisation and funding. Like investors and executives, policymakers would do well to take a more measured approach to AI.

There is potential with AI, although we must all demand greater scrutiny of its promised returns and a clearer understanding of its true costs — both financial and social. As AI’s appetite for computing power grows insatiable, policymakers and investors must navigate this complex landscape with foresight.

The promise of AI is compelling, but we must ensure we are not creating more problems than we are solving, and whether the costs of achieving those capabilities are truly justifiable.

• Dr Timcke is a senior research associate at Research ICT Africa, a research associate at the University of Johannesburg Centre for Social Change, and an affiliate of the Centre for Information, Technology & Public Life at the University of North Carolina.