The Great AI Gamble, Billions Pour In, But Will the Productivity Payoff Follow?

The Great AI Gamble, Billions Pour In, But Will the Productivity Payoff Follow?

The global economy is in the throes of a seismic shift, one driven not by traditional industries or resource extraction, but by the relentless, data-hungry engine of Artificial Intelligence. We are witnessing the dawn of the AI Gold Rush, a period of unprecedented capital investment where nations and corporations are staking their futures on the promise of intelligent machines. However, beneath the glittering surface of multi-billion-dollar data centres and soaring stock valuations lies a critical, unresolved question: Is this a foundational transformation of our economic paradigm, or the largest speculative bubble in modern history?

The stakes are astronomically high. The choices made today—by policymakers, investors, and corporate leaders—will reverberate for decades, shaping global economic hierarchies and redefining the very nature of work and productivity.

The Global Building Site: Data Centres as the New Factories

The most visible and tangible manifestation of this rush is the explosive growth in data centre infrastructure. These vast, power-intensive facilities are the factories of the digital age, and their construction is proceeding at a breakneck pace across the globe.

The recent announcement of Google’s commitment of $15 billion over five years to build a new AI-focused data hub in Visakhapatnam, India, is a case in point. This single project is expected to generate nearly 200,000 new jobs and have a “non-trivial impact” on India’s $4 trillion GDP. But Google is far from alone. The article highlights a coordinated surge from other global behemoths like Amazon and Microsoft, alongside domestic giants such as Airtel and Reliance, all channeling colossal sums into building out India’s digital backbone.

This phenomenon is a microcosm of a global trend. In the United States, the epicentre of AI development, the scale of investment is even more staggering. Jason Furman, a prominent Harvard economics professor, provided a startling statistic: AI-related investments accounted for a staggering 92% of the United States’ GDP growth in the first six months of 2025. This capital is not just for servers; it is being deployed across the entire ecosystem—networking equipment, advanced cooling systems to manage immense heat loads, dedicated fibre optic cables for lightning-fast data transfer, and, most critically, new power generation capacity to feed these energy-guzzling behemoths.

This infrastructure build-out is undeniably real. The steel is being erected, the chips are being manufactured, and the jobs are being created. It represents a classic Keynesian stimulus, injecting vast amounts of capital into the economy. But herein lies the first layer of the puzzle: while the investment in the means of production is clear, the output and productivity gains from what that production creates remain shrouded in uncertainty.

The Guessing Game: Wildly Divergent Projections on AI’s Economic Impact

When it comes to predicting the long-term economic payoff of AI, the world’s leading institutions are playing a high-stakes game of estimation, and their projections are all over the map. The confidence in the potential is universal, but the specifics vary wildly, reflecting the profound uncertainty of this new frontier.

On the more conservative end of the spectrum, we have Goldman Sachs projecting that AI will increase global GDP by $7 trillion over the next decade—a substantial figure representing about 7% of current global output. The International Monetary Fund (IMF) adds a layer of social impact, warning that AI will affect 40% of all global jobs, through either augmentation or outright replacement.

Then, we have the ultra-bullish forecasts. Consulting giant McKinsey posits that AI could grow the global economy by an eye-watering $17.1 to $25.6 trillion. To put that in perspective, this upper range is larger than the entire annual economic output of China, the world’s second-largest economy. These are not marginal gains; they are predictions of a fundamental reordering of global economic capacity.

India-specific projections are equally ambitious. The NITI Aayog, the Indian government’s premier policy think tank, estimated in September that AI could inject an extra $500 to $600 billion into the Indian economy by 2035. This growth is expected to be most pronounced in sectors like information technology and finance, but the ripple effects are anticipated to accelerate growth across a swathe of industries, from agriculture and logistics to healthcare and education.

However, a crucial disconnect is emerging between these grand, long-term forecasts and the current economic data. This brings us to the critical issue of measurement.

The Measurement Conundrum: Why GDP Might Be Missing the Picture

A fascinating insight from the Goldman Sachs report, compiled by a team of three analysts, is the distinction between “official GDP” and “true GDP” contributions from AI. Since the advent of ChatGPT in late 2022, US firms have invested approximately $400 billion in AI infrastructure. After adjusting for imports and inflation, Goldman Sachs calculates that this activity contributed $160 billion to the US “true GDP” from 2023 to June 2025.

Yet, the official government calculation for AI-related activity over the same period was a mere $45 billion. Why this massive discrepancy? The answer lies in the arcane rules of national accounting.

GDP primarily measures final demand—the value of goods and services consumed by end-users. However, a significant portion of AI investment is classified as “intermediate” spending. When a company buys servers from Dell or cloud capacity from Amazon Web Services to build its AI capabilities, this is an intermediate business cost, not a final product sold to a consumer. Consequently, a large part of this investment boom does not directly show up in the headline GDP number. Its contribution is captured only indirectly, through the increased productivity and output of the industries that eventually use the AI tools.

This is why, despite AI capital expenditure contributing 1.1% to US GDP growth in the first half of 2025, the official annualised contribution remains a paltry 0.1% of the $29 trillion economy. The infrastructure is being built, but the transformative economic activity it is supposed to enable is not yet fully visible in our primary economic scoreboard. This lag and mismeasurement can create a perception gap, where the felt reality of an AI-driven boom clashes with official statistics.

The Crux of the Debate: Acemoglu’s Sobering Counter-Narrative

Amid the unbridled optimism, a voice of nuanced caution has emerged from one of the world’s most respected economists. Nobel Laureate Daron Acemoglu, in his recent paper “The Simple Macroeconomics of AI,” presents a contrarian and sobering analysis.

Acemoglu’s argument cuts to the heart of the productivity question. He estimates that only about 5% of all current tasks performed in the US labour market can be profitably automated by AI over the next decade. The key word is “profitably.” While AI may technically be capable of performing a much larger set of tasks, Acemoglu argues that for the vast majority, the cost of developing, implementing, and maintaining AI systems will exceed the savings from replacing human labour. The business case, in short, may not be there for many applications.

From this premise, he guesstimates that AI will add only around 1% per annum to US GDP growth. While significant and welcome in any mature economy, this is a far cry from the double-digit transformational boosts that the bulls are predicting. It suggests a future where AI is a valuable productivity tool for specific, high-value tasks, not a general-purpose technology that rewrites all economic rules overnight.

If Acemoglu is correct, the implications are profound. The current tidal wave of investment, predicated on astronomical returns, would be fundamentally misplaced. We could witness a dramatic “cooling off” in AI investments as the gap between expectations and reality becomes apparent, potentially leading to a significant market correction. The gleaming data centres, built at great cost, could become the digital equivalent of “bridges to nowhere”—impressive artefacts of a bygone hype cycle with limited utility.

The Other Side of the Coin: Betting on the Exponential

Of course, the entire tech industry and its investors are betting that Acemoglu is wrong. Their faith rests on several key assumptions:

1. Positive Externalities and Spillover Effects: The true value of AI may not be in simple task automation but in driving innovation. AI is already accelerating scientific discovery in fields like drug development, material science, and climate modeling. These breakthroughs have the potential to create entirely new industries and value streams that are impossible to quantify today.

2. Exponential, Not Linear, Growth: Acemoglu’s analysis may be based on a linear projection of current AI capabilities. Proponents argue that AI progress is exponential. What is unprofitable today may become trivial and cheap tomorrow, unlocking automations and efficiencies we cannot currently conceive of.

3. The Creation of New Demand: Every major technological revolution has created new forms of demand that were previously unimaginable. The internet gave rise to the app economy and streaming services. Similarly, AI could spawn new categories of entertainment, personalized services, and business models, generating final demand that will be captured in GDP.

Conclusion: Navigating the Uncharted Territory

The world is thus at a crossroads. On one path lies the cautious, incremental growth projected by Acemoglu, where AI is a useful tool but not a revolution. On the other lies the transformative, high-growth future promised by the likes of McKinsey and Goldman Sachs, fueled by a technology that fundamentally reshapes productivity.

The truth likely lies somewhere in between, but the journey to find out will be characterized by volatility, mismeasurement, and fierce debate. For countries like India, the Visakhapatnam data centre is not just an investment; it is a strategic positioning in the new global AI order. The jobs and immediate economic stimulus are undeniable benefits. The larger question is whether these investments will be the seeds for a sustained, productivity-driven economic miracle, or whether they will be modern-day pyramids—monuments to a hope that outstripped reality.

The AI Gold Rush is undoubtedly on. Prospectors from California to Andhra Pradesh are digging in, armed with billions of dollars and boundless optimism. Whether they unearth fool’s gold or a vein that will enrich the global economy for a century to come is the defining economic story of our time. The drill bits are turning; we await the assay.

Q&A: Unpacking the AI Investment Boom

1. The article mentions that a large part of AI investment doesn’t show up in official GDP. Why is that, and what is the concept of “true GDP” in this context?

Official Gross Domestic Product (GDP) primarily measures the value of final goods and services consumed by end-users. A significant portion of current AI spending, such as a company purchasing servers or cloud computing power, is classified as an “intermediate good”—a cost of doing business, not a final product. Therefore, it doesn’t directly add to GDP. The “true GDP” concept, as used by Goldman Sachs, attempts to capture the total economic activity generated by the AI boom, including this intermediate investment. It represents the full value of resources being diverted into building the AI ecosystem, which official statistics, by their nature, undercount until the final AI-driven products and services are created and sold.

2. Nobel Laureate Daron Acemoglu offers a much more conservative outlook than other institutions. What is the core of his argument against overly optimistic AI projections?

Acemoglu’s skepticism is not about AI’s technical capabilities but its economic profitability. He estimates that while AI could technically perform many tasks, only about 5% of all tasks in the US labour market can be automated at a lower cost than using existing human-driven methods over the next decade. The core of his argument is a simple cost-benefit analysis: if implementing AI is more expensive than the current way of doing things, businesses won’t adopt it at a mass scale. This leads him to project a modest (though still significant) 1% annual boost to GDP growth, far below the more bullish forecasts.

3. How does the Google investment in Visakhapatnam exemplify the global “AI Gold Rush,” and what are its potential impacts beyond the immediate numbers?

The Google investment is a quintessential example because it showcases the scale, focus, and global nature of the rush. The $15 billion commitment is monumental, and its explicit focus on building an “AI-focused hub” signals a strategic bet on India’s digital future. Beyond the direct 200,000 jobs and GDP contribution, its potential impacts include:

• Skill Development: Creating a concentrated hub for AI will foster a local ecosystem of engineers, data scientists, and technicians.

• Attracting Further Investment: A vote of confidence from a giant like Google can attract a cascade of ancillary investments from startups and other MNCs.

• Infrastructure Boost: Such projects often force upgrades in local power grids, internet connectivity, and transportation, benefiting the wider region.

4. What are “positive externalities” in the context of AI, and why might they justify the high levels of investment even if direct task automation is limited?

Positive externalities are benefits that spill over to third parties or society at large, beyond the direct users of a technology. For AI, these could include:

• Accelerated Scientific Research: AI models can analyze vast datasets to discover new drugs, materials, or climate solutions, creating immense societal value not captured by the AI firm’s profits.

• Enhanced Decision-Making: AI tools can help farmers optimize crop yields, cities manage traffic flow, or doctors diagnose diseases earlier, leading to broader economic and social gains.

• New Innovation Platforms: Just as the internet enabled apps like Uber and Netflix, AI could become a platform for entirely new industries and services we cannot yet imagine. Investors are betting on these unpredictable, transformative spillover effects.

5. The article warns that data centres could become “bridges to nowhere.” Under what scenario could this happen, and what would be the economic consequences?

This dystopian scenario could unfold if the massive investments in AI infrastructure (data centres) far outstrip the actual, profitable demand for AI-powered services. This would happen if:

• Acemoglu’s Thesis is Correct: The cost of AI deployment remains high and its applicable scope remains narrow.

• Productivity Gains Disappoint: Companies find that integrating AI does not significantly boost their bottom line or productivity.

• A “AI Winter” Occurs: A loss of confidence and funding halts progress before major applications are found.

The economic consequences would be severe. It would lead to a massive devaluation of AI-related companies (like Nvidia), causing a stock market crash. The data centres would become underutilized “stranded assets,” representing a colossal misallocation of capital. This could trigger a broader economic recession, similar to the dot-com bubble burst of 2000, where over-investment in internet infrastructure led to a major market correction.