Economists review nation’s fiscal health and the challenges facing AI expansion Article originally posted on CoStar on March 4, 2026 The National Association for Business Economics’ Economic Policy Conference in Washington, D.C., last week brought together economists, policymakers and executives to discuss a range of topics, including the impact of artificial intelligence and the challenge of an aging population on the nation’s fiscal future. Here are a few takeaways from those discussions. Demographic headwinds and higher interest rates will make deficit reduction more challenging Federal budget deficits show little sign of shrinking. Despite substantial shifts in U.S. revenue sources and expenditures, particularly from tariffs and the deregulation and tax breaks in the One Big Beautiful Bill Act, federal budget deficits have continued to grow. The federal deficit as a share of gross domestic product (GDP) is likely to hover just under 6% through at least 2030, according to the Congressional Budget Office’s most recent projections. Mandatory social spending on an aging population is likely to increase deficit spending from there, with the CBO projecting a 6.7% deficit-to-GDP ratio by fiscal year 2036. Higher interest rates and debt levels could complicate any attempts to rein in the national debt. The share of federal expenditures attributable to net interest outlays totaled 12.8% in 2025. Under current projections, interest obligations would reach $2.1 trillion, or 18.8% of budgeted outlays by 2036. “Deficit reduction would involve policies that affect the primary balance first and foremost,” said Congressional Budget Office Director Phillip Swagel at the conference, referring to that part of the deficit that excludes interest payments and reflects only revenues less expenditures. “This composition makes the challenge of deficit reduction harder because there’s more that’s not directly impacted as part of the primary balance.” Federal debt reached about 97% of U.S. GDP as of late 2025 and is set to continue rising to 120% of GDP by 2036, according to the CBO’s projections. Sustaining debt-to-GDP ratios at their current levels, Swagel said, would require debt to be reduced by the equivalent of 2% of GDP, or $8 trillion, over the next 10 years. Continued deficits and growing debt have raised the specter of a looming fiscal crisis if substantial budget changes aren’t made in the coming decades. Few forecasters, including the CBO, have projected a specific date for such a crisis. “We don’t know when we hit that brick wall,” Swagel said. “It’s not in our projections. I don’t anticipate it being in our projections anytime soon.” Other panelists in that session rated the probability of a fiscal crisis as remote in the near term, noting that political stability and the credibility and independence of institutions such as the Federal Reserve have taken on heightened importance, particularly for global investors, who currently hold U.S. federal debt in the form of securities promising a certain return equivalent to nearly 30% of GDP. “The issue is always going to come down to: whatever [securities] we have to issue, will global investors trust that we will pay it?” said Ellen Zentner, chief economic strategist at Morgan Stanley Wealth Management. AI promises a productivity boom, but also faces policy, process and physical constraints Artificial intelligence technologies and their increasing adoption across all aspects of the economy are advancing rapidly. Faster, in many cases, than infrastructure, organizational management processes and economic policies can keep up, panelists later in the day commented. Federal Reserve Gov. Lisa Cook was largely optimistic about AI’s potential to “democratize innovation” and enhance productivity by speeding up the process of idea creation previously available only to technical experts. Noting that 60% of today’s occupations did not exist in 1940, Cook highlighted the potential of artificial intelligence to create new occupations, even as it reduces demand for some existing ones. The challenge for policymakers is the timing of job destruction and creation, she continued. If a transition period of large-scale layoffs leads to temporarily elevated unemployment, the Federal Reserve’s traditional monetary policy response of lower interest rates may be less effective at lowering unemployment than targeted educational and workforce training policies, Cook suggested. That transition phase is likely to affect sectors, companies and functions at varying rates. And the reorganization of workflows to integrate AI into existing business processes is ongoing, even in the most advanced tech firms. The time to create a completely new product or “a zero-to-one prototype is far, far shorter,” said Ging Cee Ng, senior data science manager at Meta, in the panel discussion. “And yet, we have to integrate that into an existing code stack that has a lot of quirks. Changing the way that we work and thinking about designing completely different end-to-end processes for how we might go from the initial idea stage all the way to a prototype is going to be a continual change process that isn’t going to be solved today or tomorrow.” But the biggest check on AI’s ability to scale could come down to how quickly data centers and related electrical grid infrastructure can be built to supply the technology’s increasing energy needs. Data centers use 5% of electricity in the United States. They are on pace to double that share by 2030, with stronger concentrations in data center hubs such as Virginia, according to the Electric Power Research Institute. The sudden increase comes after several decades of effectively flat electricity demand, straining aging infrastructure and increasing prices. “We fundamentally right now have a supply-and-demand mismatch for electrons,” said Shane Londagin, senior policy adviser for energy and climate to Colorado Sen. John Hickenlooper, in a conference panel discussion on AI and the electric grid. “The [electricity load growth] numbers are huge, but they are also very recent. … A lot of the policies initially proposed are reactive policies. Things like moratoriums on data centers or freezes on electricity rates. Those are responses to a problem, not necessarily proactive planning for enhanced growth.” AI expansion faces other constraints as well. The utilities industry is subject to a maze of federal, state and local regulations on permitting and zoning, and community backlash to data center construction is growing in many areas. Construction timelines are increasing. What we’re watching … The military action in the Middle East has captured everyone’s attention, particularly market watchers closely monitoring oil prices, equity markets and Treasury yields. Last week brought evidence of faster-than-expected inflation in the producer price index, while this week’s February jobs report could show a marked slowdown in job growth compared to the prior month. But these domestic market indicators are likely to be overshadowed by the visible and potential economic implications of developments abroad. With the next Federal Open Market Committee meeting a few weeks away, policymakers appear somewhat divided between concerns about continued softness in the labor market and the persistence of sticky inflation. The current environment adds a new and complex dimension to their deliberations.