By Lina Irawan 
In a detailed note circulated on Saturday, Slok deliberately invoked the famous quip by Nobel laureate economist Robert Solow from the 1980s, a period when personal computers were rapidly infiltrating businesses and homes, yet their profound influence remained elusive in official statistics: "You can see the computer age everywhere but in the productivity statistics." Slok contends that an identical sentiment perfectly encapsulates the current state of AI’s integration into the broader economy. "AI is everywhere except in the incoming macroeconomic data," he asserted, drawing a stark line between market enthusiasm and empirical evidence.
Slok’s assessment points to several key macroeconomic indicators that, despite the widespread adoption and discussion of AI, show no discernible signs of its transformative power. He meticulously scrutinizes data related to employment, productivity, and inflation, finding them remarkably stable or trending along pre-AI trajectories rather than exhibiting the dramatic shifts often predicted by AI proponents. Furthermore, a deep dive into corporate financial health reveals a similar pattern. While the "Magnificent 7" — a cohort of tech giants comprising Apple, Microsoft, Amazon, Google (Alphabet), Meta Platforms, Nvidia, and Tesla — have seen their valuations soar, largely on the back of AI investments and future potential, the vast majority of S&P 500 companies show no significant AI-driven boosts to their profit margins or earnings forecasts. This dichotomy underscores a critical point: the perceived economic benefits of AI are currently heavily concentrated within a select few, rather than broadly disseminated across the economy.
The "Magnificent 7" stand as a notable exception, largely because they are either direct developers of foundational AI models, key suppliers of the infrastructure (like advanced semiconductors from Nvidia), or early adopters integrating AI at scale into their core business models. Their enormous capital expenditures on AI research, development, and deployment, coupled with their sheer market capitalization, naturally skew aggregated market data. However, Slok’s argument focuses on the broader economic fabric, where the impact is, for now, conspicuously absent from traditional metrics.
To be sure, the investment community is not waiting for AI to definitively appear in national accounts before adjusting its strategies. On the contrary, the mere anticipation of AI’s disruptive potential has already sent powerful ripples through the stock market. Investor fears regarding AI’s ability to upend established business models have led to significant market reallocations, often resulting in a "laid waste" scenario for sectors deemed vulnerable. Shares with substantial exposure to industries such as wealth management, insurance brokerages, tax preparation, accounting services, professional data analysis, legal research, trucking, and logistics have experienced sharp sell-offs. The logic behind these market movements is clear: these sectors are perceived as highly susceptible to automation, efficiency gains, and cost reductions driven by increasingly capable AI chatbots and specialized algorithms that can perform complex, data-intensive tasks with greater speed and accuracy than human counterparts. The perceived threat is not just to jobs but to the very business models that underpin these industries, potentially compressing margins and eroding market share for incumbents who fail to adapt.
This cautious macroeconomic perspective stands in stark contrast to the often-euphoric predictions from AI evangelists and industry leaders. At the recent World Economic Forum, Anthropic CEO Dario Amodei captivated audiences by suggesting that AI could dramatically boost global GDP growth by an astonishing 5%-10%. Such a surge would represent a monumental shift in economic potential, far surpassing typical annual growth rates. Even more audacious is the vision articulated by Elon Musk, cofounder of xAI, who has frequently predicted that AI will generate such unprecedented levels of wealth and productivity that human labor could become largely optional in the not-too-distant future, fundamentally altering the concept of work itself. These grand pronouncements, while inspiring for some, highlight the vast chasm between technological possibility and current economic reality.
Despite the fervent optimism from these prominent figures, Slok remains unconvinced by the immediate prospects of AI-driven macro-economic transformation. He posits the intriguing question of a "J-curve effect" for AI. In economics, a J-curve describes a phenomenon where a country’s trade balance initially worsens after a currency depreciation before eventually improving. More broadly, it refers to an initial period of decline or stagnation followed by a significant upturn. Applied to technology adoption, it suggests that there might be an initial period of investment, disruption, and learning where productivity gains are minimal or even negative, before the technology matures and its full benefits are realized, leading to exponential growth. "Maybe there is a J-curve effect for AI, where it takes time for AI to show up in the macro data. Maybe not," he wrote, acknowledging the possibility but maintaining a stance of cautious skepticism.
The realization of such a J-curve, Slok explains, hinges entirely on the value creation derived from AI, and crucially, how that value is captured and distributed throughout the economy. Here, he identifies a significant divergence from previous technological revolutions, such as the computer age of the 1980s. Historically, early innovators in groundbreaking technologies often enjoyed a period of significant monopoly pricing power, allowing them to capture substantial economic rents before competition intensified and eroded those leads. This initial period of high profits and market dominance contributed to measurable economic gains within specific sectors. However, in the current landscape of large language model (LLM) development, Slok observes a different dynamic. Fierce competition among a growing number of well-funded developers, coupled with the open-source movement and the rapid commoditization of certain AI capabilities, has already driven the prices of many end-user AI applications and services toward zero. While this is beneficial for consumers and businesses adopting AI tools, it presents a challenge for traditional economic measurement. If powerful AI tools are essentially free or extremely low-cost for end-users, their immense value in terms of efficiency and productivity might not be adequately captured in GDP calculations, which typically measure the monetary value of goods and services produced.
From a macro perspective, Slok emphasizes that the true economic value of AI will ultimately be derived not from the specific AI products themselves, but from how these products are integrated and utilized across the economy to enhance existing processes, create new ones, and improve overall efficiency. And currently, most economists, after careful study, do not foresee a dramatically immediate impact on aggregate productivity.
Several authoritative studies support this conservative outlook. The Penn Wharton Budget Model, for instance, projects a modest annual gain in total factor productivity (TFP) from AI, amounting to just 0.1-0.2 percentage points. TFP is a critical measure of economic efficiency, reflecting how effectively labor and capital inputs are used to produce output. A cumulative boost of merely 1.5% by 2035 suggests that while AI will contribute, it won’t be the seismic economic shift some evangelists envision.
Similarly, the Congressional Budget Office (CBO) has adopted a relatively conservative stance in its latest projections. The CBO estimates that AI will add only 0.1 percentage point per year to total factor productivity growth, ultimately boosting overall output by a mere 1 percentage point by 2036. These figures, while positive, are far from the revolutionary increases predicted by optimists. This cautious assessment comes against a backdrop of recent revisions from the Labor Department, which adjusted its reading on 2025 job gains significantly downwards to just 181,000, a stark contrast to the initial print of 584,000 and the 2024 gain of 1.46 million. This revision presents a peculiar puzzle: if the economy continued to expand at a healthy pace while adding comparatively few workers last year, standard economic theory would suggest a surge in productivity. This scenario raises crucial questions about what, if any, direct effect AI had on these figures, or if other factors are at play, further complicating the attribution of productivity gains.
"After three years with ChatGPT and still no signs of AI in the incoming data, it looks like AI will likely be labor enhancing in some sectors rather than labor replacing in all sectors," Slok concluded. This distinction is vital. "Labor enhancing" implies that AI tools augment human capabilities, making existing workers more productive and efficient, potentially leading to higher wages or better quality output without necessarily displacing large numbers of workers. "Labor replacing," on the other hand, suggests widespread job displacement, which would have much more dramatic and measurable effects on employment statistics and potentially on overall economic structure. The CBO’s perspective aligns with Slok’s "labor enhancing" view: "The widespread adoption of the generative AI applications currently in production is expected to improve business efficiency and the organization of work and thus to lift TFP growth modestly over the next decade," as stated in their latest projections.
The challenge of measuring AI’s macroeconomic impact is multifaceted. Beyond the "free" nature of some AI tools, there are inherent difficulties in capturing quality improvements versus mere quantity increases. AI might make services better, faster, or more personalized without necessarily increasing the measurable volume of transactions in a way that directly inflates GDP. Furthermore, attributing productivity gains specifically to AI, as opposed to other ongoing technological advancements or organizational changes, is incredibly complex. There’s also a significant lag between technological adoption and its full manifestation in official statistics, compounded by the fact that many of AI’s benefits might initially be intangible or difficult to quantify through existing economic frameworks.
Ultimately, Slok’s analysis serves as a crucial reality check amidst the fervent AI narrative. While the technology’s potential remains immense, and its microeconomic impacts are already being felt in specific industries and companies, its broad macroeconomic footprint is, for now, surprisingly faint. The debate continues, with economists and policymakers closely monitoring whether AI will eventually emerge from the shadows of macro data, or if its transformative power will be diffused and absorbed in ways that defy easy measurement, mirroring the paradoxes of past technological revolutions. The jury is still out, but the initial signals suggest a more gradual, perhaps less immediately disruptive, integration than many had initially hoped for.