The artificial intelligence boom has entered a phase in which technological ambition is running ahead of the infrastructure required to sustain it. Global AI investment is expected to cross $1 trillion in 2026, following private investment of more than $500 billion in AI companies in 2025. Yet, even as advances in automated reasoning and autonomous agents expand the possibilities of the technology, the investment boom has slammed into an energy wall.

Acute power constraints are beginning to bite. In the United States, local electricity grids are unable to supply enough power to data centres, resulting in project delays and cancellations. The problem is becoming global. Local communities are protesting over residential electricity supplies, rising utility bills, water scarcity, environmental damage and noise pollution.

These developments raise questions about the sustainability of the AI investment cycle. They do not diminish the transformative potential of the technology. Rather, they suggest that the physical and financial infrastructure supporting AI may be emerging as an important constraint on its future development.

Investment Frenzy
The year 2024 marked a watershed in the modern deep-learning boom that began around 2012. AI won two Nobel Prizes, one in physics for foundational discoveries enabling machine learning with artificial neural networks and another in chemistry for advances in computational protein design and protein structure prediction.

Equally important was the shift from speculative experimentation to mainstream enterprise adoption. While the period up to 2023 was characterised by public fascination with text chatbots, 2024 witnessed the embedding of AI into operational infrastructure, the processing of multiple forms of information simultaneously and early advances in autonomous agents capable of performing complex reasoning tasks and interacting independently with external software.

These advances generated extraordinary demand for capital infrastructure. Investment accelerated in 2025 as automated reasoning and autonomous super-agents progressed. The gap between American and Chinese capabilities narrowed as DeepSeek matched the capabilities of leading US models, while advanced reasoning engines became accessible to millions of users.

The economic promise is considerable. Digital technologies are transforming production, employment, consumer welfare and living standards. They are less costly to scale than many conventional technologies and are enabling innovators to become entrepreneurs. The integration of AI with robotics could usher in another wave of modernisation, comparable in significance to earlier technological revolutions.

The digital economy is estimated to account for more than 15% of global GDP. Generative AI alone could increase global GDP by $7 trillion–$10 trillion over the next three years, while large language models are estimated to raise worker productivity by 8–36%.

For emerging and developing economies, these technologies could offer a route out of the low- and middle-income trap. Yet, we do not know precisely how AI will deliver a better future. Its economic potential is substantial, but the path towards realising that potential remains uncertain.

Physical Constraints
The rapid expansion of AI infrastructure illustrates the tension between technological possibilities and the resources required to realise them. Data centres are central to the AI investment cycle, but, as stated earlier, their electricity requirements are testing the capacity of existing power systems. These are not peripheral considerations. If the infrastructure required to operate increasingly sophisticated AI systems cannot expand at the pace anticipated by investors, the economics of the investment boom could become more difficult.

The emergence of these constraints is particularly significant because AI investment has already attracted extraordinary amounts of capital. Expectations of technological progress and future economic gains are being translated into present commitments to infrastructure. The question is whether those commitments can be sustained if the physical infrastructure does not keep pace.

Financial Vulnerabilities
Doubts have begun to surface about whether the AI investment boom could develop into a bubble resembling the dotcom episode. Financial markets have already demonstrated their sensitivity to changes in sentiment towards the technology.

On June 23, a technology-stock sell-off spread across Asia. Korea's Kospi fell almost 10%, triggering a 20-minute trading halt after circuit breakers were activated. Equity markets in Japan, Taiwan and China also declined, as did the Nasdaq. India's Sensex fell 1% but recovered to close the week slightly above its opening level.

The episode illustrated how investment enthusiasm surrounding AI can influence market performance across countries. It also raised questions about the vulnerability of markets that have benefited most from the concentration of investment in technology.

A more pressing black swan event could arise from the financing of AI infrastructure itself. Data centres are largely funded through public debt markets, including instruments associated with AI investment. Defaults on debt-service payments could therefore transmit stress beyond the companies directly involved.

Already, credit default swap spreads are widening and debt-service coverage ratios are sliding towards one. These developments deserve attention because the financial obligations created during an investment boom must ultimately be supported by the economic returns generated by the underlying assets.

The risk is not simply that technology shares may become volatile. It is that difficulties in servicing infrastructure debt could create a broader financial shock.

India's Position
The AI investment boom has also influenced international capital flows. Foreign investment rotated away from India towards Korea, Taiwan and China in pursuit of opportunities associated with the technology. Korea's Kospi rose 75%, becoming one of the world's best-performing benchmark indices.

This raises an interesting question about India's position in the global AI cycle. Could its relatively muted exposure to the AI investment theme offer some resilience if enthusiasm for the technology were to reverse?

The contrasting performance of equity markets during the June 23 sell-off suggests that the question deserves consideration, although one episode cannot establish a durable advantage.

India is not standing apart from the digital revolution. The India Stack is expanding financial inclusion, strengthening banking infrastructure and improving public finance management. Digital payments, electronic marketplaces and AI-enabled financial services are extending the reach of technology across the economy.

The digital economy is estimated to account for a tenth of India's GDP, while forecasts suggest that generative AI could contribute $359 billion–$438 billion to GDP by 2029–30. Indian firms' integration of AI into production processes increased from 8% in 2023 to more than 25% in 2025.

The opportunity for India is therefore substantial, even if its financial markets have not participated in the global AI investment theme to the same extent as some other Asian markets.

There is another dimension to the risks accompanying AI's rapid development. Autonomous models are creating new cybersecurity and national-security concerns. A complex cyberattack on Hugging Face in July 2026, described as having been carried out by autonomous AI agents escaping their testing environments, illustrates the emerging challenges. Governments have begun treating advanced AI models as supply-chain and national-security risks.

The economic possibilities of AI remain extraordinary. But investment, infrastructure, financial markets and technological security are increasingly interconnected. As the technology advances, these connections require careful attention.

Looking over our shoulders at the evolution of AI, it is worth remembering that objects in the rear-view mirror may be closer than they appear.

Based on a lecture delivered at the National Institute of Bank Management titled “Artificial Intelligence and Decision Analytics for the Banking and Finance Sector”. 

Continues in part 2, which is about AI transformation in banking.