LUNGILE MASHELE: Renewables alone can’t sustain AI growth

The 2026 World Economic Forum (WEF) annual meetings in Davos marked a turning point in how global leaders talk about AI, energy systems and capital allocation.

If previous years focused on AI’s promise, Davos 2026 confronted its physical and energy consequences head‑on. From CEOs to asset managers, the message was consistent: AI’s explosive growth is triggering a structural surge in electricity demand that existing energy models are not prepared to meet.

Honeywell chair and CEO Vimal Kapur cautioned that the energy requirements of AI‑driven data centres are colliding with the laws of physics. Kapur argued that the prevailing assumption that data centres can be reliably powered by intermittent renewables alone “doesn’t align with the reality of how these facilities operate”. He emphasised that AI data centres require continuous, high‑quality power, not systems that fluctuate with weather conditions.

Kapur stressed that the rapid build‑out of compute infrastructure is altering demand profiles across grids globally. Data centres are no longer marginal loads; they are becoming anchor customers for national power systems, forcing utilities, governments and technology firms to rethink industrial‑scale energy planning.

That reality was echoed by Larry Fink, chair and CEO of BlackRock, who used the 2026 Davos platform to deliver what many described as a “power reality check”. Fink warned that AI data centres “cannot depend on intermittent power like wind and solar alone” arguing that these facilities require dispatchable energy to operate at scale.

Fink went further, stating that the world is heading toward a shortage of reliable power if investment in firm generation sources does not accelerate alongside AI growth. In his words, the expansion of AI infrastructure forces a reframing of energy security from a geopolitical issue to a precondition for competitiveness.

Nvidia CEO Jensen Huang described AI as the foundation of “the largest infrastructure build‑out in human history,” spanning energy systems, semiconductor manufacturing, data centres and software platforms. This framing resonated throughout the conference: AI is no longer primarily a software revolution, but a compute‑and‑energy revolution.

AI models operate continuously, require dense chip clusters and demand intensive cooling, dramatically increasing electricity consumption. As a result, energy supply must evolve from incremental decarbonisation toward resilient, energy-intensive, scalable and firm power solutions, blending renewables with nuclear, gas, storage and grid-stabilisation technologies.

In 2025 Meta confirmed capital expenditure plans of up to $65bn for AI and data centre infrastructure. Meta’s strategy mirrors similar commitments by Microsoft, Google, Amazon and joint ventures such as Stargate, collectively signalling an industry-wide AI arms race driven by capital-intensive infrastructure.

AI bubble fears

CEO Mark Zuckerberg described the spending as necessary to secure long‑term competitiveness in AI. Meta’s structure involves special purpose vehicles to secure AI financing while concealing debt, a structure some have likened to Enron’s.

So, it was unsurprising that financial disclosures revealed that former US House of Representatives speaker Nancy Pelosi reduced direct holdings in several major tech stocks while maintaining options‑based exposure to AI leaders such as Nvidia and Apple and adding exposure to energy‑linked firms. Signalling perhaps that the end of the AI drive is imminent.

This raises questions about how expectations of AI growth and its energy implications are shaping both markets and policy scrutiny. Pundits warn of an imminent AI bubble burst, even as billions pour into new AI and energy projects that risk becoming stranded assets. The fallout will be huge, leaving little room for avoiding the inevitable.

While AI and energy are exciting, their intersection with politics and capital, highlighted at Davos, reveals an unavoidable truth: the success of AI hinges not only on advanced algorithms but also on the ability to provide secure, scalable and reliable power. The question is: who will provide it?

• Mashele, an energy economist, is a member of the board of the National Transmission Company of South Africa.