The strong enthusiasm surrounding artificial intelligence has been a key driver of stock markets in recent years, but underlying inefficiencies suggest turbulence could be ahead. Investor confidence has soared since 2022, fuelled by liquidity and economic optimism. However, changing conditions may soon turn the current AI-driven gains into losses. Warning signs point to cost structures that are not sustainable, and this could affect equity markets more severely than expected.
AI momentum has delivered notable gains, with leading firms rising sharply in value and securing major positions in key market indices. Much of this growth has been driven not only by innovation but also by an influx of capital supported by central banks and broad liquidity. As interest in scalable technology coincided with this capital, companies linked to generative AI saw their valuations skyrocket. One top AI chipmaker, for instance, reached a valuation of $7.6 trillion despite forecasting only $100 billion in revenue by 2026.
The economic model behind AI is markedly different from earlier software-based technology. Traditional platforms, such as cloud software, offer recurring revenue with very low delivery costs—code is developed once and then sold repeatedly. AI products differ, as they bring persistent and rising expenses. Each chatbot response or content generation request demands costly hardware and a large amount of energy. Unlike pure software, AI incurs a cost with every user interaction.
The infrastructure needed to support AI, including data centres packed with GPUs that use vast amounts of energy and water, does not scale effectively. AI tools like ChatGPT reportedly consume 10 to 15 times more energy than a standard web search, pushing operational costs up by hundreds of %. This increases pressure on power grids and raises expenses throughout the supply chain. Projects aimed at expanding capacity by hundreds of gigawatts or acquiring tens of millions of GPUs each year push up against both physical and environmental boundaries.
If these costs start to catch up with valuations before profits materialise, investor sentiment could shift quickly. Confidence in AI has so far been supported by hype and capital flow, but tighter monetary policy, slowing growth and rising infrastructure costs may reverse this trend. Historically, markets tend to react before companies report poor results, which means selloffs could happen early.
This is not a prediction of collapse, rather it signals a phase of recalibration that may bring more volatility and weaker returns compared to recent years. The market highs from 2023 to 2025, powered by AI, could give way to a more uncertain 2026 as investors adjust to the expensive and less scalable realities of artificial intelligence.
