The corporate world, by and large, has gone absolutely bonkers over automation, pouring staggering sums into AI technologies. In 2025 alone, businesses globally, with a significant portion from the US, lavished an estimated $410 billion on AI initiatives. This colossal investment reflects a widespread belief that AI is a veritable productivity miracle. The prevailing logic suggests that AI tools should inherently make every worker faster, more efficient, and more capable, thereby reducing the need for extensive human labor as fewer staff can accomplish significantly more. In the long run, this would ostensibly translate into immense savings for companies, higher profit margins, and a powerful boost to overall economic output. This vision of AI-driven prosperity, characterized by lean operations and exponential growth, has become the dominant narrative emanating from corporate boardrooms and Silicon Valley.

However, in the hallowed halls of traditional banking, a distinctly different, and far more sobering, yarn is being spun. Goldman Sachs, a financial institution renowned for its deep economic analysis and often cautious outlook, has not only voiced carefully-worded warnings about the potential dangers of over-investing in AI but has now dramatically escalated its rhetoric. After months of subtle hints and nuanced observations, the bank’s analysts have now asserted, with remarkable clarity, that AI has had virtually zero impact on US economic growth throughout 2025. This audacious claim stands in stark contrast to the prevailing optimism, injecting a dose of cold reality into the heated discussions surrounding AI’s immediate economic benefits.

The significant and perplexing disconnect between the massive influx of AI investment and its negligible impact on measured economic growth can be largely attributed to two fundamental structural issues, which economic analysts are now scrutinizing with increasing intensity. The first is undeniably geographic. When US companies, particularly those at the forefront of AI development, purchase advanced semiconductor chips or other critical components from international suppliers, such as the behemoth foundries in Taiwan, that substantial capital outflow primarily boosts the economy of the exporting nation. While these chips are essential for powering AI infrastructure within the US, the immediate economic benefit of their purchase accrues to Taiwan’s GDP, not directly to America’s. This globalized supply chain, while efficient for sourcing cutting-edge technology, means that a considerable portion of the hundreds of billions invested by US firms circulates outside national borders before its ultimate deployment. The economic activity, job creation, and value-added associated with the production of these foundational AI components are therefore often realized elsewhere, creating a leakage in the domestic economic impact.

The second crucial issue revolves around productivity itself. While AI undeniably makes some individual workers faster and more effective within their specific roles, these localized gains do not automatically translate into broader, systemic efficiencies across entire supply chains or the national economy. So far, a significant portion of these nascent productivity gains appears to be largely "trapped" inside company walls. For instance, an AI-powered design tool might allow a single graphic designer to complete projects in half the time, or an automated customer service chatbot might reduce the workload on human agents. These are tangible, firm-level improvements. However, for these micro-level efficiencies to ripple out and affect macro-economic indicators like GDP, they need to facilitate more efficient resource allocation across industries, reduce overall production costs, or unlock entirely new markets and services on a large scale. The current evidence suggests that this broader, transformative effect on supply chains – from raw material sourcing to final product delivery – is not yet materializing in a way that significantly moves the needle on national productivity statistics. The implementation of AI is often a slow, complex process requiring complementary investments in training, organizational restructuring, and process redesign, meaning that the full economic benefits might only be realized with a considerable time lag, a phenomenon often referred to as the "productivity paradox" in economics.

This pushback on AI’s immediate economic impact marks a sharp, even jarring, break from the consensus that dominated even the most cynical analyses of 2025. During that period, even the "doomers" and skeptics of the tech bubble often credited AI technology with single-handedly providing a crucial, if perhaps artificial, buoyancy to US GDP growth. The prevailing narrative suggested that without the AI boom, the economy might have faced a much grimmer outlook. Though the broader market has yet to fully internalize Goldman Sachs’ stark assessment – investors are still projected to spend an even more colossal $660 billion on AI across 2026 – a growing chorus of respected analysts and economists are beginning to cry foul, lending credence to the notion that the emperor of AI’s immediate macroeconomic impact might not be fully clothed.

Dario Perkins, the insightful head of macroeconomics at the prominent consulting firm TS Lombard, unequivocally agrees with the assessment that AI’s effects on overall productivity remain largely nonexistent. This stance is particularly notable given that the workforce is simultaneously reeling from massive layoffs in the tech sector and beyond, often attributed, at least anecdotally, to automation. Perkins was recently quoted in the Financial Times, making a direct and unambiguous argument: "there is no evidence that AI deployment is either boosting productivity or damaging US employment." He further elaborated, asserting that "While US productivity has been strong and hiring weak, our analysis finds that cyclical forces – not automation – are to blame." This perspective suggests that current economic trends, such as the post-pandemic normalization, shifts in consumer demand, and the effects of monetary policy adjustments, are far more significant drivers of the observed economic landscape than the much-hyped integration of AI.

Adding further weight to this skeptical viewpoint, Brian Peters, a former bank regulator at the influential New York Fed, recently penned an analysis in his "Perspective on Risk" newsletter. He acknowledged that while AI’s "capabilities are extraordinary" and the "capital deployment is unprecedented" – echoing the massive investment figures – he concluded that the "near-term economic payoff is, at best, debatable." Peters’ perspective highlights the crucial distinction between technological potential and tangible, immediate economic returns, suggesting that the sheer volume of investment does not automatically guarantee proportionate short-term macroeconomic benefits.

Further substantiating these concerns, economists at the esteemed National Bureau of Economic Research (NBER) recently published a working paper delving into the effects of AI on productivity. Their research identified a distinct "productivity paradox," where "perceived productivity gains are larger than measured productivity gains, likely reflecting a delay in revenue realizations." This NBER finding is pivotal, as it provides an academic framework for understanding the discrepancy. It suggests that while companies and individuals might feel more productive due to AI, or anticipate future revenue growth, these benefits are not yet showing up in national economic statistics, either because they are difficult to measure with existing tools, or more likely, because there’s a significant lag between initial investment, full implementation, and the eventual realization of widespread economic value. New general-purpose technologies historically require considerable time for complementary innovations, infrastructure, and skills to develop before their full macroeconomic impact is felt.

The implications of this burgeoning consensus are stark and deeply concerning. An investment boom measured in the hundreds of billions of dollars has, by Goldman Sachs’ rigorous accounting, generated essentially no measurable economic return for the United States economy in the short term. The critical question now facing economists, policymakers, and investors in 2026 is whether an additional projected $660 billion – a sum almost doubling the previous year’s outlay – will produce anything substantially different, or if it will merely inflate an even larger and potentially more precarious AI bubble. The promise of AI remains immense, but the challenge lies in translating micro-level efficiencies and technological marvels into macro-economic prosperity, a journey that appears far more complex and protracted than the prevailing corporate narrative would suggest.