AI and New York City’s Fiscal Future

A Message from the Comptroller

Dear fellow New Yorkers,

We are not helpless. Yes, it can feel that way, as the AI juggernaut gains speed. And as our federal government fails to act in any meaningful way to put guardrails on a technology that could reshape work, democracy, and daily life itself.

But this is not New York City’s first encounter with transformative technological change. Again and again, we have shown that we have the power to shape such change with democratic values and public purpose. We have proven we know how to build the institutions, infrastructure, and protections needed to expand opportunity instead of concentrating power.

Now the explosive advance of AI is putting that civic capacity to its greatest test yet.

What does this disruption mean for New York City’s economy, workforce, and tax base? Will it lead to rapid economic growth? Wide-scale unemployment? A collapse in the stock market?  As of today, there are no clear answers to these questions. Some have called this uncertainty the “AI fog.” But we know enough to at least identify likely scenarios for how AI will affect our city. Laying out those scenarios is the task of this report.

If predicting the impact of transformative AI is hard, designing the policies needed to respond is harder still. But we cannot let uncertainty paralyze us. We have to begin the work now—and in the months ahead, my office intends to lay out a broad agenda to help City government meet the challenges ahead.

One thing is already clear: whatever disruption lies ahead, New York City needs a stronger fiscal cushion to give our city the resources it will need to navigate the uncertain years ahead.

My hope is that this report sparks a serious citywide conversation about what the arrival of transformative AI will mean for New York, and what we should do about it. These are not questions we can leave to Silicon Valley, Washington, or the market alone. New Yorkers must help shape the future ourselves.

In the months and years ahead, we will need the full measure of this city’s creativity, urgency, toughness, and democratic spirit. So here is my ask: let’s meet this moment head-on, in the way only New Yorkers can.

Sincerely,

Mark Levine
New York City Comptroller

Introduction

– Prof. Ethan Mollick on the future of AI

There is no city in America—and perhaps none on earth—more exposed to both the promise and peril of artificial intelligence than New York City. And there are few places with more power to steer the transformation ahead.

The technology itself is being developed here, in an ecosystem of hundreds of firms competing to make New York the capital of applied AI. More than a million people go to work each day in Manhattan office towers, many in occupations now on the front lines of AI disruption. Our tax base depends heavily on a financial industry increasingly intertwined with AI. Our municipal government is larger than that of most nations, yet still built on a creaky foundation of 1980s and 1990s technology crying out for modernization. Our pension funds, approaching a third of a trillion dollars in assets, are invested across every level of the AI economic stack. In a city of extraordinary wealth where two million people live in poverty, AI could either help narrow inequality—or deepen it further.

New Yorkers, like all Americans, stand to benefit from AI-driven medical breakthroughs, more personalized and accessible education, and vast new possibilities for entrepreneurship.  We face profound risks as well, from job displacement to mass surveillance, cyberattacks to misinformation, and even the threat of AI itself going rogue.

These opportunities and risks demand national and global action. We need a robust safety framework for the most powerful AI systems, a tax on the companies amassing extraordinary wealth from AI so the gains are shared with working people, and ultimately a plan for democratic control of this world-changing technology.

But New York City has too much at stake to sit back and wait for action in Washington.  There is an enormous amount in our control that we must act on now—to prepare for potential job loss or economic turbulence, and to protect against new technology-driven stress on local government.  There are also significant upsides we can and should capitalize on: to improve how City agencies deliver for New Yorkers, to expand opportunity for workers and small businesses, to strengthen our fiscal resilience, and to make sure the gains from AI reach every neighborhood in the five boroughs.

Despite the impossibly high stakes for us as we stare down the radical transformation ahead, New York City—economic and cultural colossus—is doing almost nothing to prepare. We are sleepwalking into the age of AI.

The Office of the NYC Comptroller is beginning a long-term effort to change that.

We start by asking: what impact is AI already having on our economy? The answer is more complicated than either the optimists or pessimists often suggest. Some of the effects may already be visible in hiring patterns, productivity gains, and the changing structure of white-collar work. But, for now at least, it remains difficult to connect these trends to AI adoption with certainty.

The heart of this report looks to the future. We assess what accelerating AI could mean for New York City’s economy and for the City’s fisc in the years ahead. There is enormous uncertainty about how quickly this technology will advance, how deeply it will affect employment, and what it will mean for financial markets and the broader economy. We embrace this uncertainty, laying out five distinct scenarios with dramatically different implications for NYC. And we ascribe a percentage likelihood to each scenario.

Building such a forecast is inherently difficult. AI’s trajectory is impossible to predict with precision, and it is playing out against the backdrop of many other forces shaping the economy. So our goal is to map a range of plausible futures, grounded in the best available economic evidence. We begin with national AI scenarios developed by Moody’s Analytics, adapt them to reflect risks specific to New York City, and then estimate how each could affect local jobs, wages, key industries, and City tax revenues. We tweaked the range of Moody’s Analytics scenarios and their probability by adding a fifth set of assumptions that includes steeper job losses in high-pay white-collar industries.

Our five scenarios, including probabilities, are:

1. AI-Empowered Economy (35 percent) — AI boosts productivity with limited economic disruption.
2. AI Falls Flat (25 percent) — AI investment boom fizzles and markets retreat.
3. Job Replacement (20 percent) — Rapid productivity advances cause a widespread replacement of jobs, and an increase in unemployment.
4. Productivity Boon (15 percent) — AI drives faster growth, wages, and prosperity.
5. AI Shockwave (5 percent) — Labor markets fail to absorb AI-induced displacement smoothly, with job losses concentrated among white-collar jobs.

Scenarios 2, 3, and 5 involve negative outcomes for employment, economic growth, or tax revenue, and collectively account for 50 percent of our forecasted probability. In those cases, the City’s first obligation would be to maintain vital services despite significant budget strain.

But preserving core services may not be enough. If large numbers of New Yorkers quickly lose their jobs, the City may need to provide supplemental unemployment assistance or other forms of emergency support. Ideally, we would also have the capacity to invest in retraining programs, or to offer targeted incentives to employers hiring young people entering the workforce. But under our current fiscal structure, we would simply be unable to do any of those things if a crisis hits.

A well-prepared City would be able to draw on robust reserves if the more damaging AI scenarios come to pass. Unfortunately, the City’s reserves—and especially its rainy day fund—are nowhere near adequate for that purpose. At the end of FY 2026, the Revenue Stabilization Fund and the Retiree Health Benefit Trust are expected to hold a combined balance of $7.2 billion, or 8.5 percent of budgeted tax revenues.  The City’s broader fiscal cushion (as defined below in this report) has been falling in absolute terms and as a percentage of operating revenues since the start of FY 2023, and is projected to shrink further in FY 2027. As measured by S&P Ratings, in FY 2025 the City had the lowest percentage of available reserves (relative to the size of its budget) among the 10 most populous cities, ahead only of Chicago.

We propose increasing the City’s rainy day fund to make it a true fiscal shock absorber—one capable of helping the City withstand the more negative AI scenarios, as well as other major economic shocks. That means establishing strict deposit rules so that when revenues surge—whether from Wall Street gains or broader economic growth—the City automatically sets aside a share of that upside. The goal should be to build the fund to 16 percent of annual tax revenue, enough to offset the revenue losses of a typical recession. Our full set of recommendations on the rainy day fund can be found in the Strengthening the City’s Rainy Day Fund report published last April.

Shoring up the rainy day fund should be an immediate priority for the  City.  But much more work beyond this looms ahead. We need to ensure our pension system is robust to potential AI-driven market turbulence.  We need to use our power as shareholders to ensure that corporations adopt responsible AI policies.  And we must prepare City government itself for the transformation ahead—modernizing outdated systems, making agencies more nimble, and hardening our public infrastructure against AI-powered threats, from cyberattacks to a flood of AI-generated incoming slop. We intend to put forward policy solutions on each of these issues in the months ahead.

This report is an opening salvo in what will be an on-going effort by the Comptroller’s Office to help New York City understand, prepare for, and shape the most dramatic technological revolution in our lifetimes.

Our ambitions are based on a core belief: the cities that thrive in the AI era will not be the ones that wait to see what happens. They will be the ones that seize the initiative—fiscally, operationally, and financially. New York has become the greatest city in the world in part because we have done just that during every previous technological revolution. Now it’s time for us to mobilize for the biggest transformation yet.

AI’s Impact on the Economy Today

Artificial Intelligence (AI) is no longer a speculative technology on the horizon. It is here, already reshaping how businesses operate, how workers perform their jobs, and how consumers interact with services. Across industries, AI tools are being used to automate routine tasks, accelerate data analysis, expand coding capabilities, support customer service, generate written and visual content, improve logistics, and augment professional work. These changes have created real opportunities for productivity growth, innovation, and new forms of economic activity. But they have also introduced significant uncertainty about the future of work, wages, inequality, public revenues, and the resilience of local economies. And this is just the beginning.

Like earlier major technological innovations, from the steam engine and the airplane to the PC and the internet, its effects on the global economy and the lives of people are likely to be transformative. Pessimists express concern about the potentially massive displacement of workers (jobs) and further increases in income and wealth inequality. Optimists foresee a boom in productivity, and a corresponding expansion in wages, economic opportunity, and the overall quality of life.

Low-Hire, Low-Fire Economy

Both nationally and in NYC, some unprecedented and seemingly quirky economic patterns have appeared over the past couple of years. Historically, over the business cycle, booms have reliably been characterized by strong labor markets, with extensive hiring and modest layoffs. Conversely, downturns and periods of sluggish growth have been characterized by a pullback in hiring and increased layoffs. These two indicators—layoffs and hiring—have reliably moved together (in opposite directions). Over the past couple years, however, this pattern has been broken: new hiring and net job creation have been moribund, while layoffs have remained quite low. As shown in Chart 1, while new layoffs (jobless claims) and job growth were highly negatively correlated throughout the 1975-2019 period, they have moved in the same direction over the past couple year—following the pandemic.

Chart 1

Sources: U.S. Bureau of Labor Statistics; NY Department of Labor, Moody’s Analytics
*Pandemic period is excluded because both series were literally off the charts.

One explanation, especially in knowledge-driven industries, is that entry-level jobs have been the most susceptible to replacement by AI technology. This is because technical skills—many supplanted by AI—have become less imperative, while soft skills and specialized company or industry-specific knowledge have become more so.

As highlighted in our February Spotlight, this pullback in entry-level jobs has hit recent college graduates particularly hard, especially locally. As shown in Chart 2, the unemployment rate among recent college graduates (aged 22-27) has risen, both in absolute terms and relative to both older college grads and young people without college degrees. Interestingly, the unemployment rate among young adults without a college degree, which has historically been relatively high, has receded substantially over the past two years. During the 12 months ending March 2026, for the first time on record, young adults with a college degree faced a slightly higher unemployment rate (7.3 percent) than those without (7.1 percent).

Chart 2

Sources: US Bureau of Labor Statistics, Current Population Survey

Despite the lackluster hiring trends and the potential risks to white-collar jobs explored below in this report, the Manhattan office market is going from strength to strength. In 2025 new leasing increased to nearly 31.0 million square feet, a 32.3 percent increase from the previous year, with leasing in 5-star buildings lifting market activity to multi-decade highs. Leasing accelerated further in the first quarter of 2026, fueled by Bank of America’s renewal and expansion at One Bryant Park and American Express anchoring of the Two World Trade Center development. Leasing was strong across submarkets and there are signs of an increase in AI/tech leasing, as exemplified by announcements at One Madison and One World Trade Center, and the expected Anthropic lease of 330 Hudson Street.

Adoption Trends

Direct measurement of AI adoption is now possible across a handful of complementary sources. The Census Business Trends and Outlook Survey (BTOS), the most comprehensive nationally representative source, reports that approximately 18 percent of U.S. establishments (and 32 percent on an employment-weighted basis) used AI in business functions in late 2025 and early 2026, with adoption expected to reach 22 percent within the next six months. Adoption is concentrated among larger firms: on an employment-weighted basis, AI adoption reaches 32 percent.[1]

New York State adoption, however, lags the national average. The BTOS state series places New York at 16.8 percent as of early April 2026 — below the national rate of 19.8 percent. In the BTOS AI supplement covering November 2025 to January 2026, New York lagged the nation in the use of AI across all reported business functions.

User statistics show a different result for New York. The March release of Anthropic’s Economic Index data registers New York at roughly 9 percent of national Claude usage, against a state share of total nonfarm employment near 6 percent, a higher rate than most states. Adoption is sharpest in finance, information, and professional services — the sectors that disproportionately drive New York City employment, wages, and tax receipts. This concentration is consistent with our Office’s prior analyses of AI’s likely first-order effect on the office-using economy (see the August 2025 and the February 2026 Spotlights).

The Atlanta Fed’s survey of approximately 750 CFOs provides a granular firm-level picture of AI adoption to date.[2] Over half of the companies surveyed had invested in AI by early 2026, though investment is heavily concentrated among large firms. The primary barriers to adoption are immature technology (cited by 42 percent of firms), an untrained workforce (36 percent), and privacy/data security concerns (36 percent). Notably, the top motivation for AI investment is to improve production efficiency, not to reduce labor costs.

Within adopting firms, deployment remains relatively narrow as 57 percent have integrated AI use in no more than three business functions. The most common applications are sales and marketing (52 percent of adopting firms), strategy and business development (45 percent), and information technology (41 percent). Generative AI in worker tasks is common— about 23 percent of all U.S. firms (41 percent employment-weighted) — but is overwhelmingly used for writing, document analysis, and information search rather than for end-to-end automation of work processes.

Productivity gains, where measurable, remain concentrated in high-skill services and finance, where firms report annualized labor productivity gains of roughly 0.8 percent (expected to exceed 2 percent over the course of 2026).

Aggregate AI-driven employment effects through 2026 remain small in the CFO data — under 0.4 percent — but the underlying composition is shifting: routine clerical work shrinks while skilled-technical roles expand. Furthermore, expected job losses in large firms are partially offset by expectations of employment growth among smaller firms. Recent news reports suggest that industries at the adoption frontier may be experiencing more acute job cuts.[3]

The results from the CFO survey appear broadly consistent with the productivity growth observed in 2025 among industries more exposed to AI, such as tech-heavy industries (data processing, computer manufacturing, etc.), finance, legal services, and others. Morgan Stanley estimates that output growth drove an increase in labor productivity in these industries, with little evidence of labor replacement.

A Growing Wealth & Income Gap

Related to the disproportionate effect on entry level workers is the broader issue of further increases in wealth and income inequality. While some relatively high paying occupations—e.g. coders, data analysts, legal assistants, and even stock & bond traders—are certainly susceptible to AI displacement, this is also an issue for lower-paying occupations, ranging from data entry clerks, telemarketers, etc. In effect, as noted in our August 2025 Spotlight, there should be a positive effect on jobs which can be enhanced (i.e. made more productive) by AI but a negative effect on jobs that can easily be replaced by AI.

Less predictable and measurable are whole new occupations that arise because of AI. Before computers, computer programming was not a thing; before airplanes, there were no pilots; before the internet, nobody had heard of web designers; and before automated telephone switching systems, switchboard repairers & network engineers were, at best, rare. Because this technology is likely to shrink employment in some occupations, create jobs in both new and existing occupations, and shift the mix of tasks within many occupations, the directional effect on income inequality is hard to predict.

The apparent effect on wealth inequality seems clearer. Strong corporate profits and booming financial markets, driven largely by AI-driven productivity gains and investment in AI technology, have boosted the wealth of the most affluent. To the extent that AI has enabled corporations to extract more profits from consumers, some of the same forces driving up profits may have also crimped affordability—due to innovations such as algorithmic price discrimination. These two trends together would tend to lead to further increases in wealth inequality. However, there is also evidence that expanding AI usage has strongly boosted consumer surplus (economist lingo for enabling consumers to get more value for the money they spend). It remains to be seen how much of the increased benefits created by AI accrue to businesses versus consumers … not to mention which businesses and consumers benefit most.

Contribution to GDP Growth

While the distributional effects of AI are complex and difficult to measure, let alone predict, its overall contribution to economic growth to date can be estimated. A recent study by the Federal Reserve Bank of St. Louis found that nearly 40 percent of real GDP growth during the first 3 quarters of 2025 was directly attributable to AI investment in equipment, software, R&D, and data centers. This does not even count instances where AI usage contributed to GDP.

The contribution to GDP growth appears to be accelerating. In the first quarter of 2026 (first estimate) private nonresidential fixed investment grew 10.4 percent from the preceding year, pushed by information processing equipment (growth of 43.4 percent) and software (growth of 22.6 percent). These two components alone accounted for approximately 2/3 of GDP growth in the quarter.

Based on quarterly earnings reports, the four largest companies investing in AI capacity (“hyperscalers” Alphabet, Amazon, Meta, and Microsoft) have committed around $700 billion in capital expenditures, mostly on infrastructure, over 2026. This would represent a growth rate of around 70 percent from 2025, and nearly triple the amount invested by these firms in 2024. Investment might approach one trillion in 2027. If achieved, the 2026 investment target by only four firms would represent nearly 13 percent of all gross private domestic investment in 2025.

Two physical and financial constraints could affect the widespread adoption of AI. The first is energy. U.S. data centers consumed roughly 224 terawatt-hours of electricity in 2025[4] (more than 5 percent of total estimated consumption)[5] and are projected by the International Energy Agency to nearly double their consumption by 2030, accounting for an outsized share of incremental U.S. electricity demand.

The second constraint is financial. Leading AI companies are experiencing exponential growth, as exemplified by Anthropic’s recent meteoric trajectory. But the growth in token consumption also highlights the potential to run into compute constraints,[6] which in turn accelerates the AI infrastructure investment wave. The financial constraint would be hit if realized revenues were not sufficient to generate the expected rate of return. A BlackRock analysis suggests that this may be plausible (even if not the base case).[7]

In looking ahead, the path of AI adoption and its impact on the U.S. and NYC economies could take various forms. In short, pessimists worry about  a scenario in which AI adoption replaces huge numbers of jobs, causing massive unemployment, while optimists foresee its positive effects on economic growth overpowering, or at least offsetting, negative effects on overall employment (job replacement). Lowered expectations for AI revenues could spell the end of the current investment cycle and drive a correction in the financial markets. The next section explores a range of potential scenarios.

AI Scenarios for the US Economy

While AI’s impact on the economy is already being felt, how it will evolve in the future remains the subject of great debate and at times divergent views. Moody’s Analytics has modeled four distinct macroeconomic scenarios based on varying sets of assumptions about AI and its path of adoption into the economy.[8] We added a fifth scenario that is meant to capture more adverse negative impacts on white-collar jobs. In doing so, we resized Moody’s Analytics’ probabilities assigned to its four scenarios, as explained below. Using the projections of the U.S. economy under each of the five scenarios, we have estimated parallel projections for the New York City economy. The scenarios were drawn before the start of the war in Iran and its repercussions, as well as the (so far) increase in asset values, are not factored in the scenarios.

The scenarios are as follows:

• The “AI-Empowered Economy” (baseline) scenario (35 percent probability), assumes that AI both replaces some jobs and enhances others at a gradual rate, boosting productivity growth modestly with little disruptive effect on the economy or employment.
• The “AI Falls Flat” scenario (25 percent probability) assumes that AI investment is soon deemed excessive—that AI adoption rates, productivity gains, and profitability fall short of expectations, and a market selloff follows.
• The “Job Replacement” scenario (20 percent probability),[9] in contrast, assumes faster-than-expected AI adoption, leading to rapid productivity growth, widespread replacement of jobs, and an increase in unemployment.
• The “Productivity Boon” scenario (15 percent probability), the most optimistic of the five, assumes an outcome similar to the late-1990s Internet boom—one in which productivity growth doesn’t displace job growth but rather complements it, boosting compensation.
• The “AI Shockwave” scenario (5 percent probability) combines a rapid AI capability path with labor markets that fail to absorb the displacement smoothly. Capabilities advance fast but routine cognitive work in finance, law, customer service, and administrative support is replaced faster than displaced workers can be reabsorbed. In this scenario labor’s share of income contracts more sharply than in the Job Replacement scenario.

We assigned the AI Shockwave scenario a probability of five percent and reduced Moody’s Analytics’ probability for the baseline scenario accordingly (from 40 percent to 35 percent). One reason supporting the shift of probability toward the AI Shockwave is the rapid growth of projected AI-related investment and the exponential increase in compute usage since the beginning of the year. In our view, these trends make a rapid diffusion of AI and a deeper economic disruption slightly more likely, and therefore we gave it a 5 percent probability.

The investment trends do not affect our views on the probability of the “AI Falls Flat” scenario because observing faster investment (and higher stock prices) does not yet fundamentally alter the physical and financial constraints that motivate the scenario. If anything, the constraints may now be more acute.

The other scenarios’ probabilities are the same as those assigned by Moody’s Analytics.

National Economic Assumptions

Under the baseline AI-Empowered Economy scenario, near-term U.S. economic growth faces some headwinds from the disruption associated with AI integration, but productivity gains materialize over the medium term. Real Gross Domestic Product (GDP) grows at an average annual rate of 2.3 percent. In this scenario, private sector job gains average about 700,000 (0.5 percent) per year, and jobs in office-using industries (Finance, Insurance, Real Estate, Information, and Professional and Business Services) grow twice as fast.

In the AI Falls Flat scenario, investors and corporations realize that they may have over-invested in data centers, R&D, and other AI-development resources—possibly because revenues and profits from these investments fall short of expectations or because of energy constraints. A stock market decline of nearly 35 percent occurs and the negative shocks to consumer spending and capital investment lead to a brief recession. Nearly 3.3 million private sector jobs are lost by the middle of 2027 but recover by the end of 2028. Output growth also recovers after one year, but the level of Real GDP remains below the baseline forecast permanently, a result of lowered projected productivity.

In the Job Replacement scenario, real GDP initially grows slightly faster than in the baseline in 2026 but slows significantly in 2027 as the weight of rising unemployment begins to affect wage growth and consumer spending. Private sector job losses accumulate to 1.4 million before employment begins to rise slowly in late 2028. The stock market shows initial strength as corporate profit margins benefit from cost-saving substitutions for labor, but the negative macroeconomic feedback of higher unemployment and slower wage growth ultimately proves to be the stronger effect after a couple years.  In this scenario, output does not keep pace with the baseline within the five-year horizon examined.

The Productivity Boon scenario results in accelerating economic growth at an average annual rate of 2.9 percent, six-tenths of a percentage point above the baseline growth rate. Corporate profits outpace those projected for the baseline and the stock market is expected to grow by an average annual rate of 9 percent through 2030, versus 5 percent expected in the baseline. Employment levels are not dramatically altered in this scenario versus the baseline—an initial slight decline in the job growth rate versus baseline reflects stronger short-term adjustments, but high productivity leads to an acceleration in employment gains in the later years.

The AI Shockwave scenario assumes a more severe labor-market disruption, with private sector job losses of 5.4 million, but significantly protracted through mid-2028 and concentrated in high-paying office-using industries (nearly 2.0 million in this scenario versus a loss of 600,000 by the start of 2027 in the Job Replacement projections). Real GDP grows slowly at an annual average rate of 1.5 percent through 2030, driven by labor market dynamics and reduced demand resulting from the loss of high-paying jobs. Job displacement initially boosts profit margins but losses in aggregate consumption eventually drag earnings below baseline. The labor share of national income drops by 1.5 percentage points and average annual S&P growth is the slowest in this scenario (2.2 percent).

To be clear, the AI Shockwave scenario is not the catastrophic “global intelligence crisis” described by Citrini Research in a note that garnered wide attention.[10] It is more akin to a recent “fast diffusion” scenario modeled by Morgan Stanley and broadly consistent with the range of adverse economic outcomes of the “rapid AI capability” scenario in a survey of forecasters conducted by the Forecasting Research Institute and the Federal Reserve Bank of Chicago.[11]

Table 1 compares the projected path of key U.S. variables across scenarios.

Table 1. U.S. Scenario 2025-2030 Annual Growth Rates

Source: Moody’s Analytics and Office of the NYC Comptroller

What the Scenarios Mean for NYC

What do these scenarios mean for New York City’s economy? What story do they tell for job creation, incomes, and tax revenues? To answer these questions, we have run each U.S. scenario through our forecast models. The taxes included in the analysis are the more economically sensitive revenues from personal and business income, sales, and real estate transactions.[12]

These scenarios are not firm predictions of the future and should be understood as a planning exercise — plausible paths the economy could take, and that government should prepare to face.  The dates and timelines should be interpreted as an indication of how quickly these scenarios could unfold, rather than a strict prediction of outcomes on specific dates.

The AI-Empowered Economy (Baseline)

The forecasts show a wide range of outcomes for the New York City economy and the City’s tax base. The AI-Empowered Economy Baseline shows a moderate growth path with private sector employment expanding at an average of about 52,000 jobs (average annual growth rate of 1.2 percent). Office-using industries expand at an average of 21,000 per year (average annual growth rate of 1.3 percent) and drive the economic expansion. Wage income grows at a moderate pace between 3.8 percent and 4.6 percent. Wall Street profits gradually ease from the heights reached in 2025 to $51.6 billion in 2026 and $47.2 billion in 2027 before resuming growth. Tax revenues grow at an average growth rate of 3.1 percent from their level in FY 2026. Table 2 summarizes the forecast variables.

Table 2. NYC Forecast in the AI-Empowered Economy Scenario

Source: Office of the NYC Comptroller

AI Falls Flat

The AI Falls Flat scenario paints a sharp reversal of the AI investment-driven economy. In this scenario the New York City economy and tax revenues are hit by a sudden shock, but the damage is temporary, as it would be for a typical recession. The private sector would lose 52,500 jobs over the course of the year, 5,000 of which are from office-using industries. Wage income stalls in 2027 and Wall Street profits drop to $33.1 billion ($18.5 billion below baseline), a result of the significant financial market decline. The economic shock would cause FY 2027 tax revenues to drop by $3.4 billion.

The economy starts healing in 2028 and the recovery is nearly complete by the end of the projection window. Table 3 summarizes this forecast.

Table 3. NYC Forecast in the AI Falls Flat Scenario

Source: Office of the NYC Comptroller

Chart 3 shows the difference between the baseline and the AI Falls Flat scenario. At the deepest point in 2027Q3, the economy has roughly 135,000 fewer private-sector jobs than in the baseline path. Private sector employment is again at baseline by 2030, but office-using industries continue to trail by about 12,000 jobs. Wage income (not shown) grows at a 3.7 percent compound annual rate from 2025 to 2030, but it remains about 2 percent below baseline. Wall Street profits never reach the levels forecasted in the baseline. By FY 2030, tax revenues are close to baseline levels but are cumulatively below the baseline by roughly $9 billion.

Chart 3

Source: Office of the NYC Comptroller

Job Replacement

In this scenario, faster-than-expected AI adoption produces rapid productivity growth and widespread replacement of routine cognitive work. The NYC private sector sheds 13,600 jobs over the course of the year. A rebound in office-using employment drives overall job gains in 2027, but its role in the recovery fades over time as job gains move to other sectors of the economy. Yearly private sector job gains average around 33,500, while office-using industries add 15,000 on average. On the strength of equity market gains in 2026 and declining interest rates through 2030, Wall Street profits exceed the baseline throughout the forecast window. Table 4 summarizes the data.

Table 4. NYC Forecast in the Job Replacement Scenario

Source: Office of the NYC Comptroller

Chart 4 shows the difference between the baseline and the Job Replacement scenario. The jobs shortfall against baseline does not narrow over the forecast horizon: by 2030 the New York City economy has roughly 96,000 fewer private-sector jobs, and a shortfall of 30,500 in office-using sectors. The structural hit to mid-skill office work weighs on the City’s wage base for the full forecast horizon. Yet, total nonfarm wages grow at a 3.6 percent compound annual rate over 2025 through 2030, finishing about 2½ percent below baseline by 2030, not that different from the AI Falls Flat scenario.

In the Job Replacement scenario, combined tax revenues trail the baseline by roughly $5.5 billion cumulatively through FY 2030. Personal income taxes bear the heaviest hit while business income taxes remain broadly resilient on stronger margins. The fiscal consequence is more dispersed than under AI Falls Flat: while total revenues avoid a year over year decline, they essentially stall for three fiscal years.

Chart 4

Source: Office of the NYC Comptroller

Productivity Boon

In this scenario, broad-based productivity gains lift output, wages, and corporate profits without large employment displacement. Private sector job gains remain close to the pace of the baseline scenario, although they lag initially as firms restructure to capture AI gains. Wall Street profits outperform — $57.5 billion in 2026 (nearly $6 billion above baseline) and rising toward $71 billion by 2030. Wage income grows at a 4.8 percent compound annual rate from 2025 to 2030 — the fastest of any scenario, leaving 2030 wage income roughly 3.4 percent above baseline. Table 5 summarizes the data.

Table 5. NYC Forecast in the Productivity Boon Scenario

Source: Office of the NYC Comptroller

Chart 5 shows the difference between the baseline and the Productivity Boon scenario. By the end of 2030 private sector employment is roughly 28,000 jobs above baseline and office-using industries are back at the baseline level. Combined tax revenues run roughly $8 billion above the baseline cumulatively through FY 2030, driven by income taxes. This is the only modeled scenario that leaves NYC unambiguously better off than the baseline on every tax category.

Chart 5

Source: Office of the NYC Comptroller

AI Shockwave

This scenario compounds the labor-displacement profile of Job Replacement with a Wall Street valuation correction and a sharper near-term contraction in office-using industries. The negative employment effects start accelerating toward the end of 2026. In 2027, the private sector loses more than 110,000 jobs. Nearly three out of five jobs lost are from office-using industries. Wall Street profits decline year-over-year for three consecutive years in 2026–2028 as some initial gains in corporate profitability are overpowered by a negative economic outlook stemming from the extent of job loss. Table 6 summarizes the data.

Table 6. NYC Forecast in the AI Shockwave Scenario

Source: Office of the NYC Comptroller

Chart 6 shows the difference between the baseline and the AI Shockwave scenario. Relative to baseline, the gap in private-sector jobs reaches 259,000 in the first quarter of 2029, with only a small recovery by the end of 2030. Office-using industries end 2030 with 145,000 jobs below baseline. Wage income grows at just a 2.6 percent compound annual rate over 2025 through 2030 — the slowest of any scenario — leaving it roughly 7 percent below the baseline level.

Combined tax revenues trail the baseline by roughly $14 billion cumulatively through FY 2030 — the largest shortfall of any scenario. The differential grows from $1.1 billion in FY 2027 to $5.0 billion in FY 2029, before narrowing slightly in FY 2030. Although the assigned probability is low (5 percent), the concentration of the shock in NYC’s strongest tax bases makes it the highest-impact risk for New York City’s economy and finances.

Chart 6

Source: Office of the NYC Comptroller

How the Scenarios Fit in the Current Economic Context

While the scenarios used in this analysis make varying assumptions about the economic impact of AI adoption, there are other factors that may have a substantial impact on the forecasts for the U.S. and local economies. Most notably, the geopolitical situation, particularly in the Middle East, and energy costs comprise another significant dimension of uncertainty. These represent risks (mostly to the downside) across scenarios. It should be noted that oil price shocks have often been followed by recession and sometimes by stagflation.

Whether or not the economy slides into recession will depend on the interplay between the evolution of the AI investment and adoption cycle, the realized gains from AI technology, and the length and depth of the economic shock from the Iran war. So far, the U.S. economy has been fueled by expectations of conflict de-escalation and a temporary increase in inflation, coupled with a dramatic increase in business investment. Stock markets have been driven sharply higher by AI: as of mid-May, Goldman Sachs notes that technology firms accounted for 85 percent of the S&P 500 year-to-date return in 2026.[13] A strong stock market is supporting consumption through a wealth effect, which offsets the negative repercussions of high energy prices and modest job growth.

Building Reserves to Protect Services, Workers and New Yorkers

New York City cannot wait until the fiscal impacts of artificial intelligence are fully visible to begin preparing for them. By the time a shock appears in falling revenues, rising unemployment, or a weakened tax base, the City’s options will already be narrower and more painful. AI could strengthen our economy, create new industries, and expand productivity. But it could also move quickly through the sectors that matter most to the City’s finances, including finance, professional services, technology, real estate, and other high-wage office-using industries. Responsible fiscal management requires preparing now for both possibilities: capturing the upside of AI while building the capacity to withstand disruption.

The City already has a vital tool to protect against unexpected fiscal shocks and downturns: its rainy-day fund (known as the Revenue Stabilization Fund or RSF). The RSF should be used to set aside resources during stronger fiscal periods so the City can protect services, avoid sudden tax increases or layoffs, and stabilize the budget during an economic downturn or major shock. Yet, at nearly $2.0 billion, the fund remains far too small for the scale of risk New York faces.

The City’s broader fiscal cushion, which includes the RSF, Retiree Health Benefits Trust (RHBT), general and capital stabilization reserves, and budget stabilization (the prepayment of future expenses) gives another perspective on the state of the City’s finances. As shown in Chart 7, the cushion is projected to fall from a peak of $14.5 billion at the beginning of FY 2023 to $8.4 billion in FY 2027, based on the City’s assumptions in the Executive Budget. This amount is a marked improvement from the $6.6 billion projected in the Preliminary Budget but is at the lowest level since the start of FY 2021, in the middle of the pandemic emergency (and substantially lower as a ratio to total operating revenues).

Chart 7

Source: NYC Office of Management and Budget, Office of the NYC Comptroller. (E) denotes values are derived from the FY 2027Executive Budget. This chart was updated since the report release to revise the amount of the budget stabilization account in FY 2027.

Using a similar financial construct that includes RSF, RHBT and budget stabilization, S&P Ratings estimates that in FY 2025 the City had a fiscal cushion of 9.4 percent of total operating revenues. As shown in Chart 8, this was the second lowest fiscal cushion as a percentage of operating revenues among the 10 largest cities in the U.S.

Chart 8

Source: S&P Ratings. Data as of 2025, except for Chicago, San Diego, and Jacksonville where data are as of 2024.

A 16 Percent Rainy Day Fund Target

In April 2026, the Comptroller’s Office published Strengthening the City’s Rainy Day Fund, a proposal to establish a rainy-day fund target of 16 percent of tax revenues, with a 10 percent lower bound. The 16 percent target is not arbitrary. It is based on New York City’s historical experience in economic downturns, when cumulative revenue losses fell 16 percent below trend before fully recovering. Based on projected FY 2026 tax revenues of $84.4 billion, that means a full target of $13.5 billion. This amount is comparable to the impact of the AI Shockwave scenario on the City’s tax revenues estimated in this report. As mentioned above, RSF currently stands at $2.0 billion and RHBT (which has been used as a rainy day fund in the past) holds $5.2 billion. Even if combined, these amounts only represent 8.5 percent of projected FY 2026 tax revenues. New York City should take steps to adopt a policy and reach the 16 percent target as soon as possible.

Because an AI-driven negative shock would hit tax revenues, the City should build reserves to avoid forcing service cuts, layoffs, or tax increases. The reserves should be used first to protect vital City services that New Yorkers rely on every day, including education, housing, public health, sanitation, public safety, and social services. They could also help provide the City the fiscal flexibility to respond to job market disruption, including by supporting retraining and upskilling programs, strengthening job-placement pathways into growing sectors, and more for workers displaced by AI-driven restructuring. The goal is not simply to save money for a rainy day; it is to ensure that, when disruption comes, the City can protect core services, stabilize families, and help workers transition into the next economy.

Building reserves will not prevent AI disruption. But it can determine whether the City responds from a position of strength or crisis. A well-funded rainy day fund would help protect essential services, avoid sudden layoffs or tax increases, and give the City time to support workers, communities, and industries through a period of rapid economic change.

Conclusions

Artificial intelligence is not a distant possibility. It is already reshaping the economy, changing how businesses operate, altering the skills workers need, influencing financial markets, and creating new risks and opportunities. For New York City, the stakes are especially high. Our economy is deeply tied to the industries most exposed to AI. Our tax base depends heavily on high-wage office work and financial-sector performance. Our municipal government has urgent modernization needs. And our communities could experience AI either as a force that expands opportunity or one that deepens inequality.

The path of AI economic impact remains uncertain. But in this report, we lay out a range of plausible scenarios and examine what each could mean for New York City’s jobs, wages, industries and tax revenues. The lesson is clear: uncertainty is not an excuse for inaction. It is the reason government must prepare.

That preparation must begin now. New York City needs stronger fiscal reserves to withstand potential economic shocks. We need to modernize City government so agencies can deliver faster, more reliable, and more accessible services in an AI-enabled world. We need procurement and technology systems that allow government to innovate responsibly, while protecting privacy, security and fairness. We must understand which workers and industries are most exposed to disruption. This report is the beginning of that work for the Office of the Comptroller.

New York City has navigated technological transformation before. Each time, our success has depended not on waiting passively for change, but on making deliberate choices about how that change should serve the public good. AI will test our economy, our government, and our civic institutions in new ways. But with clear-eyed planning, disciplined fiscal management, and a commitment to shared prosperity, New York City can meet this moment—and help shape an AI future that is more productive, more resilient, and more equitable.

Acknowledgements

This report was prepared by Jonathan Siegel, Chief Economist; Yaw Owusu-Ansah, Director of Tax Policy and Revenue Analysis; Aida Farmand, Senior Tax Policy Analyst; Jason Bram, Director of Economic Research; Aliyah Sahqani, Economic Research Analyst; Amber Born, Economic Development Analyst; Andre Vasilyev, Assistant Director for Economic Development; Krista Olson, Deputy Comptroller for Budget; and Francesco Brindisi and Chris Carroll, each an Executive Deputy Comptroller. Archer Hutchinson, Creative Director and Danbin Weng, Multimedia Designer led the design, with assistance from Angela Chen, Senior Website Developer, and Martina Carrington, Web Developer.

---

Endnotes

^[1] Bonney K., Breaux C., Dinlersoz E., Foster L., Haltiwanger J., Pande A. (2026) “The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks,” Center for Economic Studies Working Paper CES-26-25, Census Bureau

^[2] Baslandze S., Edwards Z., Graham J.R., McClure T., Sparks M., Meyer B., Ravindranath Waddell S., Waddell, and Weitz D. (2026) “Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives,” Federal Reserve Bank of Atlanta, Workin Paper 2026-4, March.

[3] Wall Street Journal 2026 Layoffs Tracker, accessed May 12, 2026.

[4] International Energy Agency (2026) Key Questions on Energy and AI, Table A.4.

[5] U.S. Energy Information Administration (2026) Short-Term Energy Outlook, April.

[6] See Atiyeh K. (2026) The Trillion-Dollar Blindspot You Are Missing, Ramp, April; Chiam I., Xie M., Wang R., Kundojjala S., Wong G., Patel D. (2026) The Great AI Silicon Shortage, SemiAnalysis, March.

[7] Boivin J., Fishwick E., Li W., Paul V., Purves G., Rieder R., Savi R. (2026) 2026 Global Outlook, BlackRock Investment Institute.

[8] Zandi M., deRitis C., DiNatale M., DeAntonio D., Colyar M., Whitcher S., Begley J., Hysa I., Semmens G. (2026) The Macroeconomic Consequences of AI, Moody’s Analytics, February.

[9] Moody’s refers to this as the “Job Market Upheaval” scenario.

[10] Citrini, Shah A. (2026) The 2028 Global Intelligence Crisis, February.

[11] Karger E., Kuusela O., Abaluck J., Bryan K.A., Halperin B., Jones T.R., Murphy C., Trammell P., Reynolds M., Mayland D., Viswanathan R., Mittal A., De Castro R.C.,, Rosenberg J., Tetlock P. (2026) “Forecasting the Economic Effects of AI,” NBER Working Paper 35046, April. The scenarios considered by the forecasters are defined by what AI can do, not necessarily deployment. In the “slow” scenario, AI is a capable assisting technology — handling perhaps half of routine software-engineering work and most customer-service complaints, drafting publishable literature reviews, and operating household robots in supervised settings. In the “moderate” scenario, AI is an effective collaborator across most professional domains, and AI manages multi-day operational projects under human direction. In the “rapid” scenario, AI surpasses the best human minds across most cognitive and physical tasks.

[12] Specifically, the forecasts include the Personal Income Tax and Pass-Through Entity Tax (PIT/PTET), the Business Corporation Tax, the General Corporation Tax, and the Unincorporated Business Tax (collectively Business Income Taxes – BIT), the Sales Tax, the Real Property Transfer Tax, and the Mortgage Recording Tax. The property tax is excluded because it is affected by economic conditions with a lag.

[13] Snider B., Hammond R., Ma J., Chavez D., Jaychandran K., Sung C. (2026) “AI momentum, and the One Big Trade,” Goldman Sachs, US Weekly Kickstart, May.