The Economic Multiplier or Misery AI: Macroeconomic Impact

Artificial intelligence (AI) is increasingly hailed as a revolutionary force for the global economy, but experts fiercely disagree on whether it will act as an economic multiplier or a source of “misery.” On one hand, technologists and optimistic economists argue that AI could dramatically boost productivity and gross domestic product (GDP) growth, perhaps even to unprecedented levelsilluminem.com. On the other hand, many worry that AI’s gains might be unevenly distributed – magnifying inequality and concentrating wealth in the hands of a few tech giantsimf.orgimf.org. This debate has intensified as recent advances (such as generative AI systems like GPT-4) spark both excitement over efficiency gains and anxiety over potential job displacement. The ultimate macroeconomic effect of AI remains uncertain, especially in the near term, and the stakes are high: Will AI inaugurate a new era of broad-based prosperity, or will its benefits accrue to a narrow elite? Economists are examining early trends and historical parallels to determine whether AI will principally multiply economic output or spread misery through disruption – or some balance of both.

Lofty Promises: AI as a Productivity Multiplier

Enthusiasts often compare AI to past general-purpose technologies (like steam power or electricity) that unlocked waves of productivity and growth. They note that AI’s ability to automate routine tasks and augment human work could turbo-charge economic growth. Some projections are strikingly optimistic. Goldman Sachs researchers, for example, estimate that widespread AI adoption could raise labor productivity by about 1.5% per year (adding roughly 0.5 to 1.5 percentage points to annual GDP growth in the United States over the next decade)thirdway.org. Cumulatively, AI might contribute around $7 trillion to global GDP over 10 yearsthirdway.org. Consulting firm McKinsey goes further, suggesting AI could unlock $11–$17 trillion in economic value annually for businesses worldwide once fully adoptedthirdway.org. Even these figures are viewed as conservative by some futurists. In an extreme scenario, a breakthrough to artificial general intelligence (AGI) could, according to technologists, drive annual global GDP growth to 20–30% – a pace that would dwarf any historical industrial revolutionilluminem.comilluminem.com. (For context, 20th-century global growth averaged only about 2.8% per year, so such AGI-driven expansion would be virtually unprecedentedilluminem.com.)

Why might AI be such a powerful economic multiplier? Proponents highlight several factors. First, AI can automate a wide range of tasks, from administrative duties to aspects of professional “knowledge work,” allowing human workers to focus on more complex or creative activities. By streamlining communication, coding, document processing, and research, AI tools enable workers to accomplish more in less timethirdway.org. Early studies lend credence to this: in one field experiment at a call center, giving customer service agents an AI assistant boosted their productivity by an average of 14% – with entry-level workers improving output by 30%thirdway.orgthirdway.org. Such augmentation could make less-experienced employees as effective as their veteran counterparts, dramatically improving overall workforce efficiencyimf.orgimf.org. As tasks are completed faster and more accurately, businesses can produce more output with the same input, raising productivity, which is the fundamental driver of long-run economic growthimf.org.

Second, AI is expected to spur innovation. Machine learning systems can sift through data to discover new patterns, optimize designs, or even aid scientific R&D. This could accelerate technological progress in fields like medicine, clean energy, and materials science. For example, advanced AI might enable breakthroughs in drug discovery or climate modeling, yielding benefits beyond just the tech sectorilluminem.comilluminem.com. Such knock-on innovations can create entirely new industries and markets, further expanding GDP.

Third, AI may reduce production costs and improve quality. Automation tends to lower the cost of goods and services – a boon for consumers and a spur to demand. AI-driven efficiencies (from smarter supply chains to predictive maintenance in factories) could lower costs and prices, potentially boosting real incomes. Optimists even suggest AI might contribute to taming inflation over time by increasing the economy’s productive capacity (though central bankers caution against assuming AI will magically counteract inflation in the short run)americanbanker.com. In sectors like manufacturing, generative design algorithms and AI-driven robotics can increase output while using fewer resources, translating into productivity gains without commensurate wage pressures. For instance, the integration of generative AI with 3D printing allows cheaper customization of products and could cut manufacturing costs significantly over timethirdway.org.

Finally, AI might help address demographic and labor challenges. In aging societies with workforce shortages, AI and robotics can supplement human labor, supporting growth even as the labor force declines. This is one reason countries like Japan and China are investing heavily in AI – to mitigate the economic drag of aging populations.

The Productivity Paradox: Delayed Gains and Modest Impact?

For all the lofty promises, there is also substantial skepticism about AI’s actual impact on macroeconomic productivity – especially in the near term. History offers a cautionary tale: past technological revolutions often took years or decades to translate into higher productivity statistics. The economist Robert Solow famously quipped in 1987 that “you can see the computer age everywhere but in the productivity statistics,” highlighting the lag between adopting computers and realizing efficiency gains. Some economists suspect we may see a similar “AI productivity paradox.”

Early evidence suggests that broad-based AI-driven growth has yet to materialize. As of 2024, adoption of AI in business remains limited. A U.S. Census survey found only 6% of American firms were using AI regularly in their operations, up slightly from 4% a year earliergoldmansachs.comgoldmansachs.com. Lack of knowledge, integration challenges, and privacy concerns – not just cost – are cited as the main barriers to adoptiongoldmansachs.comgoldmansachs.com. While many companies are experimenting with AI pilots, few have deployed it at scale in “core” production processes. By one estimate, only about 5% of tasks in the U.S. economy could be profitably automated with AI at present, once implementation costs and practical constraints are consideredmitsloan.mit.edumitsloan.mit.edu. This translates into a relatively small macroeconomic effect. MIT economist Daron Acemoglu projects a “nontrivial, but modest” contribution of AI to growth – on the order of only ~1% added GDP over an entire decade for the United Statesmitsloan.mit.edumitsloan.mit.edu. This sober forecast stands in stark contrast to the more optimistic multi-trillion dollar projections, and reflects a belief that current AI is not a magic bullet for growth. Acemoglu notes that so far generative AI excels mostly at “easy-to-learn” tasks with clear inputs and outputs; applying AI to more complex “hard” tasks (like medical diagnosis or creative problem-solving) will likely yield diminishing returns initiallymitsloan.mit.edu. Furthermore, many high-impact use cases of AI are in sectors dominated by small and medium enterprises that lack the capital or expertise to adopt these tools quicklymitsloan.mit.edu. These factors could limit AI’s productivity boost in the short run, keeping GDP growth closer to its pre-AI trend.

Additionally, capturing AI’s benefits often requires complementary investments – companies may need to reorganize workflows, train staff, and upgrade infrastructure to fully leverage AI. Such adjustment costs can postpone the productivity payoff. Indeed, there is active debate on when AI will noticeably lift productivity growth. Some analysts point out that we are already seeing glimmers of impact: certain workers are “doing far more with less” thanks to AI toolsweforum.org, and U.S. labor productivity statistics in late 2023 showed an uptick that coincided with heavy business investment in technology. Yet others caution that any aggregate boost might not emerge until a critical mass of firms effectively adopt AI – which could take several years. Goldman Sachs economists, while optimistic about the medium term, do not expect AI to significantly show up in productivity data until 2027 or later, with peak macro impact likely in the early 2030sgoldmansachs.comgoldmansachs.com. In their view, the next 1–3 years represent just the beginning of the AI investment cycle, where firms ramp up spending on AI (projected to reach ~2% of GDP in the U.S. at the peak of the investment surge) and experiment with applicationsgoldmansachs.comgoldmansachs.com. Only after this phase will the wider economy start reaping efficiency gains at scale.

Empirical research also underscores the time lag. A 2023 OECD study of AI’s macro impact over 10 years estimated a modest annual productivity growth boost of 0.3–0.6 percentage points for advanced economies – useful, but hardly explosive growthoecd.orgcepr.org. This suggests AI’s contribution might be incremental and gradual, rather than a sudden boom. In sum, while the long-run potential of AI to raise living standards is real, the short-run effects could disappoint if one is expecting an immediate economic revolution. The world may eventually get an AI-driven growth acceleration – but perhaps later and slower than the most bullish forecasts imply.

Disruptions and “Misery”: AI’s Labor Market Impact

If the optimistic scenario is AI as a rising tide lifting all boats, the pessimistic scenario is a tide that lifts only yachts while leaving many smaller boats stranded. Central to the concerns is the impact on jobs, wages, and inequality. Automation anxiety is not new – from textile machines to computers, workers have long feared being made obsolete – but AI’s breadth has reignited those worries across both blue-collar and white-collar occupations.

Numerous studies indicate AI will transform work across a large swath of the economy. The International Monetary Fund estimates that almost 40% of all jobs worldwide could be affected in some way by AI and automationmitsloan.mit.edu. While many roles will be augmented by AI, certain tasks (and the jobs built on them) will be eliminated. Pessimistic projections highlight massive potential job displacement: for instance, a Pew Research analysis found one in five workers are in occupations with high exposure to AI-driven automation, many of them well-paid, highly educated rolesthirdway.org. Unlike past automation that mostly hit routine manufacturing jobs, this wave targets even college-educated, white-collar jobs – from paralegals to radiologists to financial analysts – due to AI’s proficiency with information workweforum.orgweforum.org. Some researchers warn that by 2040, AI could automate tasks so rapidly that tens of millions of workers may need to change occupations within a single decadethirdway.org. Already, there are early signs of this upheaval: companies have begun announcing hiring freezes or layoffs attributable to AI. In 2023, for example, IBM’s CEO said the company would pause hiring for thousands of back-office roles, anticipating AI could perform those tasksshrm.org. In the media and content industry, AI tools capable of generating text and imagery have led to job cuts and a re-evaluation of work – contributing to strikes by Hollywood writers and actors in 2023 who feared their creative labor would be undervalued or replacedimf.orgimf.org.

The crucial economic question is not whether AI will displace jobs – it certainly will eliminate some tasks – but whether it will also create new jobs and raise incomes, or instead leave a portion of the workforce worse off. Optimists point out that historical technological revolutions ultimately created more jobs than they destroyed, often in ways that are hard to foresee. They argue AI could generate new roles (for example, in AI system training, maintenance, or entirely new industries) and increase demand such that human labor is still needed, albeit in different capacities. In this hopeful view, AI makes workers more productive and valuable, and those productivity gains lead to higher wages and more employment opportunities in the long runthirdway.orgthirdway.org.

Many economists, however, are not convinced the transition will be smooth. There is a real risk of a mismatch – workers whose jobs are automated may not easily shift into the new roles that AI creates, especially if those require different skills or are in different locations. This could lead to a period of elevated unemployment or under-employment for certain groups. Moreover, if AI primarily enables automation (replacing workers) rather than augmentation, the benefits of higher productivity might flow mostly to business owners and shareholders, not to labor. Indeed, recent economic history already shows a divergence between productivity and pay. Since the 1970s, productivity in many advanced economies (notably the U.S.) has grown faster than median wages, contributing to a widening income gapweforum.orgweforum.org. Technological change – including automation and computerization – is believed to be one factor behind this decouplingweforum.orgweforum.org. In the late 20th century, as computers spread, productivity rose ~60% but real wages for typical workers rose only ~10%, with the lion’s share of gains accruing to higher executives and capital ownersthirdway.org. AI could exacerbate this trend. Half of chief economists surveyed by the World Economic Forum in 2025 said “increasing inequality” is one of the biggest risks from AI adoption, even as they acknowledge AI’s potential to boost growthweforum.org. The concern is that AI’s rewards – greater output, lower costs – may not be broadly shared. If, for instance, a company can produce the same output with fewer workers thanks to AI, the savings might primarily boost profits unless there are mechanisms to redistribute those gains (such as higher wages for remaining workers or lower prices for consumers).

A likely outcome, absent policy intervention, is greater labor market polarizationimf.org. AI can directly substitute for many middle-skill, middle-income jobs (data processing, routine coding, administrative support), while complementing high-skill jobs and leaving many low-skill service jobs untouched (it’s harder to automate caregivers, cleaners, food service, etc.). This means we could see more jobs at the top and bottom of the wage spectrum, but fewer in the middleimf.orgimf.org. As one scenario, imagine a future workforce composed of a relatively small elite of AI developers, engineers, and managers who design and oversee AI (earning very high salaries), a diminished middle of office and factory workers, and a large class of lower-paid workers in manual or service roles that still require a human touch (such as home health aides or hospitality staff). In this “hollowed-out” labor market, inequality would widen by defaultimf.orgimf.org. Early trends give some weight to this view: recent studies of AI deployments found the greatest productivity boosts often go to the least-skilled workers (by helping them catch up to more skilled colleagues)imf.org, which is promising for reducing skill gaps. Yet paradoxically, if entry-level workers become as productive as veterans, employers may feel less need to pay a premium for experience – potentially compressing wages or eroding some middle-tier positionsweforum.orgweforum.org. Indeed, the call-center study noted junior agents improved greatly with AI assistance, while experienced agents saw little improvementthirdway.orgthirdway.org; this suggests the extra value of experience diminished when AI leveled the playing field.

The geographic dimension of AI’s labor impact is also worth noting. Advanced economies could see certain jobs reshored (for example, if AI reduces the cost advantage of outsourcing tasks to low-wage countries). Emerging markets that have relied on labor-cost competitiveness may find some of their industries under threat if AI allows richer countries to automate rather than offshore. This raises concerns that inequality could widen not only within countries but also between them, with early AI-adopting nations gaining an edge in productivity. However, developing countries might also benefit from AI in areas like agriculture or healthcare if the technology is made accessible – the net effect is uncertain and likely uneven.

In summary, AI brings the promise of higher productivity but also the peril that those gains might bypass much of the workforce. Whether AI becomes an engine of broadly shared prosperity or fuels a new era of inequality and worker displacement will depend on how businesses implement the technology and how policymakers respond (for instance, through education, retraining programs, or social safety nets). The social “misery” scenario is not predetermined – as even AI boosters concede, realizing the best outcomes “will require good policy” and proactive efforts to “help shape the future of technology and the economy”imf.orgimf.org.

Big Tech and Market Concentration: Who Wins the AI Race?

Another facet of the AI debate centers on industrial concentration and the distribution of gains between firms. Will AI democratize innovation, allowing startups and small businesses to thrive? Or will it super-charge the dominance of today’s tech giants, effectively creating an AI oligopoly that captures most of the value?

Skeptics fear a scenario of increasing concentration, where the largest firms – equipped with superior data, computing power, and talent – pull further ahead of everyone elseimf.orgimf.org. Developing cutting-edge AI models is enormously expensive and resource-intensive. For example, training the advanced GPT-4 model reportedly cost over $100 million, and running it incurs about $700,000 per day in cloud computing costsimf.org. Only a handful of firms (think of trillion-dollar corporations like Alphabet/Google, Microsoft (through OpenAI), Amazon, or well-funded Chinese titans like Baidu and Tencent) can easily afford such investments. These companies also hold unique advantages: vast reservoirs of user data to train AI, and integrated platforms where they can deploy AI at scale. As executives in the AI industry acknowledge, scaling laws in AI research mean that bigger computing budgets yield better-performing models – giving an edge to those with the deepest pocketsimf.org. In effect, AI could erect higher barriers to entry: if state-of-the-art models require billions of dollars to develop, smaller competitors might be locked out. This concentration dynamic is already evident – by the mid-2020s, a few firms in the U.S. and China dominate frontier AI research and cloud infrastructure. It raises the prospect of “winner-takes-most” markets where the top AI platform providers reap enormous profits (and market share) by selling AI services or using AI to optimize their own operationsimf.orgimf.org.

The flip side is that these large firms become dramatically more efficient and profitable by leveraging AI internally, further widening the productivity gap between big and small companies. A multinational corporation can use AI to coordinate its complex global supply chain or personalize services for millions of customers – advantages smaller firms might struggle to attain. As one analysis noted, an AI-empowered “visible hand” of top management could make large enterprises even more efficient at resource allocation, eroding the nimble advantages that small firms traditionally claimimf.orgimf.org. If that happens, we might see consolidation in various industries, with AI-augmented giants absorbing market share from less automated rivals.

However, an alternative outcome is possible. Some experts argue that AI could be a great leveler if its tools become widely accessible. The emergence of open-source AI models – such as Meta’s LLaMA or other freely available systems – might enable smaller companies and even individuals to use advanced AI without needing to build it all in-houseimf.org. In mid-2023, a leaked memo from a Google researcher observed that “open-source models are faster, more customizable, more private, and pound-for-pound more capable” than the leading proprietary models in some tasksimf.org. If open-source AI continues to improve and spread, it could lower the cost barrier and allow a broad community of developers to innovate on top of shared models, much as open-source software did in earlier IT wavesimf.org. This would undermine the dominance of any single company. Moreover, AI might spur entrepreneurship: with powerful AI tools at their fingertips, more innovators could start their own niche businesses rather than joining a tech behemothimf.org. In a hopeful scenario, AI could reverse the trend of rising market concentration seen in recent decades, enabling agile startups to compete with incumbents by quickly adopting AI solutions developed elsewhereimf.orgimf.org. For instance, a small ecommerce company could implement an open-source recommendation algorithm nearly as good as Amazon’s, or a local manufacturer could deploy AI-driven robotics via affordable cloud services – thus narrowing the efficiency gap.

As of the mid-2020s, the balance between these two trajectories (concentration vs. diffusion) is still playing out. Early evidence leans toward big firms benefiting first: they have been the primary adopters of AI so farimf.org, and Big Tech companies have seen surging valuations partly due to investor enthusiasm that AI will boost their future earnings. The stock market reflects this; in 2023, companies strongly associated with AI (from chip-maker NVIDIA to cloud providers and platform giants) saw dramatic rises in stock prices, creating enormous wealth for their shareholders. Meanwhile, smaller firms often cite difficulties in accessing AI talent and integrating AI systems. Policy choices could influence this balance – for example, antitrust enforcement, support for open research, or public investment in disseminating AI technology to smaller businesses could help counteract concentration. The outcome will determine not just who gets rich from AI, but also whether AI-driven growth, if it materializes, is accompanied by vibrant competition or an entrenched monopoly-like landscape.

Global Competition: AI’s Impact in the US vs. China (and Beyond)

AI has also become a cornerstone of geopolitical economic competition, most notably between the United States and China. Both nations view leadership in AI as critical to future economic and strategic prowess. However, their approaches and the potential economic impacts of AI differ in interesting ways.

The United States, home to many of the pioneering AI companies and researchers, currently leads in cutting-edge AI development. U.S. firms dominate in producing state-of-the-art AI models – in 2024, U.S.-based institutions produced 40 notable new AI models compared to 15 from Chinahai.stanford.edu. Private investment in AI in the U.S. has been enormous: in 2024, U.S. private AI investment reached $109 billion, almost 12 times the size of China’s private AI investment (~$9 billion) that yearhai.stanford.eduhai.stanford.edu. This reflects not only American tech giants’ deep pockets but also a thriving startup ecosystem that attracted venture capital. As a result, the U.S. arguably stands to benefit quickly from AI innovations, with its companies at the forefront of commercializing AI. Analysts expect the U.S. to see productivity gains sooner than other countries, with adoption of generative AI beginning to register in economic output by the later 2020sgoldmansachs.comgoldmansachs.com. Goldman Sachs forecasts the U.S. will adopt AI slightly faster and could unlock roughly $4.5 trillion in annual GDP through AI-enabled productivity by around 2030goldmansachs.com. However, the U.S. also faces challenges: its gains may be uneven (coastal tech hubs vs. the rest) and there are political debates over regulation, data privacy, and how to retrain workers disrupted by AI. So while the U.S. might reap substantial economic rewards, it must also grapple with internal inequality and adjustment costs.

China, for its part, has made AI a national priority with aggressive investment and industrial policy support. The Chinese government’s strategic plan calls for China to be the world leader in AI by 2030. Chinese companies have strengths in certain areas – for example, massive troves of data (given China’s large population and relatively lower data privacy restrictions) and rapid deployment of AI applications (like facial recognition, fintech, and e-commerce personalization). Some Chinese AI firms have recently achieved notable breakthroughs at lower costs than their U.S. counterpartsgoldmansachs.comgoldmansachs.com. In 2025, a Chinese company called DeepSeek was reported to have developed a sophisticated generative AI model at a fraction of the cost of Western modelsgoldmansachs.com. If true, this cost innovation could spur faster and cheaper adoption of AI across emerging markets, not just in Chinagoldmansachs.comgoldmansachs.com. Economically, forecasts suggest China might see an even larger boost from AI (as a share of GDP) than the U.S.. One analysis projected that by 2030, AI could raise China’s GDP by 26% above baseline, versus a 14% increase in North America’s GDPthirdway.orgthirdway.org. This is partly because China’s economy still has vast room for productivity improvements – AI could turbocharge manufacturing, logistics, and services in China, yielding huge efficiency gains. Moreover, AI might help offset China’s demographic headwinds (an aging workforce) by enabling higher output per worker. If China realizes these gains, it could potentially narrow or even surpass the U.S. in economic size in the long term, fueled by an AI dividend.

That said, China lags the U.S. in certain foundational areas: the most advanced semiconductor chips needed for AI are largely designed by U.S. firms and manufactured in allied countries like Taiwan; export controls on such chips to China could slow China’s progress. Also, a significant portion of China’s AI advancement has been in applying existing techniques (sometimes adapted from Western research) rather than inventing fundamentally new AI algorithms. But China’s strength lies in scaling and implementation. For example, AI-driven automation is being rapidly deployed in Chinese factories, and cities are integrating AI in traffic management and surveillance at scale. The economic impact of AI in China may thus be very tangible in infrastructure and industry within a few years, even if China hasn’t yet overtaken the U.S. in frontier research. It’s telling that Chinese public opinion is highly optimistic about AI – surveys show over 80% of people in China see AI as more beneficial than harmfulhai.stanford.eduhai.stanford.edu, a far higher optimism rate than in the West. This enthusiasm, combined with state support, suggests a relatively rapid embrace of AI across Chinese society.

Beyond the U.S. and China, other regions have varied prospects. Europe has considerable AI talent and industry, but the EU’s approach emphasizes regulation (e.g. the EU AI Act) which some fear might stifle AI innovation or slow deployment in the short term. Europe’s contribution to top AI models is smaller (only a few notable models in 2024)hai.stanford.edu. However, Europe could benefit from AI in areas like advanced manufacturing and has a large market to apply solutions once matured. Other Asian economies like Japan and South Korea are actively investing in AI, especially robotics, to cope with aging populations. Emerging economies in Africa, Latin America, and South/Southeast Asia could harness AI for leapfrogging in sectors like fintech, agriculture (through AI-driven yield improvements), and remote education or healthcare. But if they lack access to AI technology or skills, there’s a risk they fall further behind leading nations, widening the global digital divide.

Crucially, the competition is not zero-sum: an AI breakthrough in one country can spill over and raise productivity globally, and increased competition (say, between the U.S. and China) might actually accelerate progress and drive down costs, benefiting other countries. For instance, Goldman Sachs noted that the rise of a credible Chinese AI competitor could “provide an uplift to global adoption and productivity” by spurring more innovation and prompting other governments to support domestic AI effortsgoldmansachs.comgoldmansachs.com. In that sense, a more multipolar AI landscape could be a boon for the world economy – as long as it doesn’t trigger protectionism or an AI arms race that cuts off collaboration.

Near-Term Outlook (1–3 Years)

In the immediate future – roughly the next one to three years – the macroeconomic impact of AI is likely to be significant in certain sectors but subtle at the aggregate level. We are essentially in the early adoption and investment phase of the AI revolution. Businesses worldwide are ramping up spending on AI. By some estimates, global AI investment (public and private) is approaching $200 billion by 2025, with the United States accounting for about half of thatgoldmansachs.com. A McKinsey survey found an overwhelming 92% of firms expect to increase their AI investments in the coming yearsmckinsey.com. This surge of capital is funding everything from data center expansions (for cloud AI services) to AI startup ventures to workforce training in AI tools.

The immediate visible effects of this wave are likely to be: productivity improvements in specific tasks, growing efficiency in tech-savvy firms, and some labor displacement in easily automated roles. We can expect continued rapid development of AI applications, especially generative AI (text, image, and code generation) being integrated into office software, customer service, marketing, and software development. For employees in those areas, AI will increasingly act as a “copilot,” handling routine portions of their job. This should raise their output per hour in those tasks. Indeed, research already finds that workers using tools like coding assistants or AI customer-chat systems can save a significant portion of their time (on the order of 5–30% time saved, depending on the task)stlouisfed.orgweforum.org. If these productivity gains scale up, companies may be able to either produce more with the same staff or maintain output with fewer staff.

So far, though, these efficiencies have not translated into a macroeconomic productivity boom. Overall productivity statistics remain moderate, in part because adoption is uneven and gains in one area may be offset by lags elsewhere. Surveys indicate a rapidly rising share of companies experimenting with AI, but full integration is limited. Notably, according to the Stanford AI Index report, the percentage of organizations reporting use of AI in at least one function jumped from 55% in 2023 to 78% in 2024 (buoyed by the hype around new AI tools)hai.stanford.edu. Figure 1 below illustrates this sharp uptick in AI usage. However, using an AI tool in one department is a far cry from industry-wide transformation. Many of these firms may still be in trial stages.

Figure 1: Rapid increase in businesses reporting AI usage (at least in one function). Surveys show the share of organizations using AI rose to 78% by 2024 (up from 20% in 2017), with a particularly large jump following the advent of popular generative AI toolshai.stanford.edu. However, “using AI” often means limited pilots; only a small minority of firms have deployed AI at scale in core production as of 2025goldmansachs.com.

Over the next few years, we should watch for specific leading indicators of AI’s macro impact. One is business investment patterns: if companies are pouring money into AI-related capital (software, hardware) and intangibles (data, algorithms), that shows up in economic data as higher investment, even if output gains come later. Another indicator is labor market shifts: we may see continued strong demand for AI specialists and engineers, alongside hiring freezes or slower growth in roles deemed automatable (like certain administrative jobs). Unemployment rates might not spike, but job churn could increase as workers transition between sectors. Wage dynamics bear monitoring too – if AI truly boosts productivity, one would hope to see at least some uptick in wage growth for those leveraging the technology, though if most gains go to profits, the wage share of GDP could decline.

Importantly, many economists caution that policy responses in the next few years will be critical in shaping AI’s ultimate impact. Governments and firms that invest in reskilling programs, education, and complementary infrastructure (like digital infrastructure, broadband, cloud services) may enable a more seamless integration of AI, maximizing benefits and minimizing pain. For instance, three-quarters of chief economists polled by WEF said government spending on upskilling and redeployment of workers is a top priority in the AI eraweforum.org. If such measures are put in place now, it could mitigate the “misery” scenario by helping workers transition into new roles created by AI and ensuring the workforce can augment AI effectively. Conversely, a laissez-faire approach might lead to wider skill gaps – with those who know how to use AI prospering and those who don’t being left behind.

On the regulatory front, the near term will also show how different jurisdictions balance innovation and risk. The EU’s forthcoming AI regulations, U.S. initiatives on AI safety, and China’s government directives will all influence how freely AI can be deployed and commercialized. Too much restriction might slow economic benefits; too little oversight could allow harms that undermine public trust (and thus adoption).

In summary, the next 1–3 years are likely to set the trajectory for AI’s macroeconomic role. It is a period of experimentation, investment, and adjustment. We should not expect AI to add several percentage points to global GDP growth overnight – most experts see that as unlikely in the very short horizonmitsloan.mit.edugoldmansachs.com. Instead, we will see more micro-level success stories and incremental improvements. The hope for optimists is that these years lay the groundwork for a productivity surge later in the decade. The fear for pessimists is that even in the short term, AI could start creating a class of “losers” – workers and firms that find themselves outcompeted by early AI adoptersweforum.orgweforum.org. In effect, the coming few years will provide valuable information: they will reveal how feasible it is for AI to be adopted widely, how much friction there is in that process, and whether early adopters use their gains to further invest and hire (virtuous cycle) or simply to cut costs and reap profits.

Conclusion: A Balanced Perspective

In classic Economist fashion, the truth about AI’s macroeconomic impact likely lies between the extremes of euphoria and doom. AI is neither a guaranteed economic savior nor an unavoidable engine of misery – outcomes will depend on choices made by businesses, policymakers, and society at large. What is clear is that AI represents a new general-purpose technology with the potential to significantly influence productivity, employment, and the distribution of income. Such technologies historically have led to substantial gains, but only with time and often only with complementary innovations in how we work and govern the economy.

If AI is to be an economic multiplier in a positive sense, a few conditions may need to be met. Businesses must successfully integrate AI to genuinely augment human capabilities (not just cut costs), and workers will need the skills and tools to work alongside AI. The fruits of higher productivity must be shared – through market mechanisms or policy – such that a broad base of people see rising incomes and improved services. This could mean updating education systems, promoting competition (so that no single firm monopolizes AI gains), and perhaps even novel ideas like “AI dividends” or profit-sharing if automation drives huge profitsimf.orgimf.org. On the flip side, avoiding the misery scenario will require vigilance against widening inequality. Societies may consider stronger social safety nets or transition assistance for those displaced. There is also an ethical and political component: if large segments of the population feel left behind by the AI-driven economy, the social backlash could itself slow down technological progress (through resistance or political pressure).

Globally, AI could either widen divides or help close them. International cooperation – in setting standards, sharing research, and ensuring access to AI for developing countries – might tilt the balance toward more inclusive growth. A world where only the U.S. and China dominate AI and reap its rewards would be a poorer outcome than one where AI knowledge and benefits diffuse widely.

At this juncture (mid-2020s), economists remain intensely divided on AI’s ultimate macro impact because so much is unknown. Will AI soon achieve breakthroughs that dramatically raise its capabilities, or will progress be steady and incremental? Will firms use AI primarily to enhance their workforce or to cut it? Will the gains from AI-driven productivity flow into broad wage growth, lower prices, and new jobs, or chiefly into profits and valuations? These questions have different answers depending on one’s assumptions. It is telling that a respected economist like Daron Acemoglu projects only a 1% GDP boost from AI over 10 yearsmitsloan.mit.edu, while consultancies and tech visionaries project an order of magnitude moremitsloan.mit.eduthirdway.org. Such divergence highlights the deep uncertainty at play.

Perhaps the wisest stance is cautious optimism: AI can be a powerful force for prosperity if handled properly. The technology itself is a tool – it can be used to uplift society or to entrench imbalances. As one commentary noted, “it’s not about the technology… it’s about how we choose to apply it”weforum.org. In the next few years and beyond, the world will be running a real-time experiment with AI in the economy. The outcomes in the United States, China, and elsewhere will offer lessons on how to maximize the multiplier effect of AI while minimizing the misery. In all likelihood, AI will bring a mix of benefits and challenges: productivity gains alongside disruption, new wealth alongside difficult transitions. Navigating this complex impact is the central economic challenge of our times. Much like electricity or the internet, AI’s true macroeconomic legacy will take decades to fully unfold – but the choices made now will determine whether that legacy is one of broadly shared growth or one of heightened inequality. Economists will continue to watch the data closely, but for now, the verdict on whether AI becomes an unprecedented growth engine or a double-edged sword remains unsettled. The coming years will be critical in tipping the balance between multiplier and misery in the story of AI and the global economy.

References

• “What if AI made the world’s economic growth explode?” (Summary of The Economist, July 2025)
• Brynjolfsson, E. & Unger, G. (2023). “The Macroeconomics of Artificial Intelligence,” Finance & Development, IMFimf.orgimf.org.
• Brynjolfsson, E. & Unger, G. (2023). Ibid. Illustrating divergent scenarios: productivity, inequality, concentrationimf.orgimf.org.
• Letzing, J. (2025). “AI could make us more productive, can it also make us better paid?” World Economic Forumweforum.orgweforum.org.
• Third Way (2023). “AI and the US Economy: Optimism, Pessimism, or Realism?” – summarizing projections and perspectivesthirdway.orgthirdway.org.
• Third Way (2023). Ibid. Citing PwC projection for China vs North America GDP gainsthirdway.org.
• Goldman Sachs (2025). “China’s advances could boost AI’s impact on global GDP.” – Goldman Sachs Research briefinggoldmansachs.comgoldmansachs.com.
• Walsh, D. (2025). “A new look at the economics of AI,” MIT Sloan (discussing Acemoglu’s findings)mitsloan.mit.edumitsloan.mit.edu.
• Stanford HAI (2025). “AI Index Report 2025: Top Takeaways.” (AI investment and adoption statistics)hai.stanford.eduhai.stanford.edu.
• Letzing, J. (2025). WEF article, Ibid. Detailing chief economists’ views and labor market concernsweforum.orgweforum.org.