AI and US Debt: Why Growth May Not Be Enough
Can AI-driven growth ease America’s debt burden? Examine the growth, tax, and data-centre investment assumptions behind that promise—and what the numbers say.
Can AI Growth Really Help America Out of Debt?
The promise sounds straightforward: artificial intelligence boosts productivity, the economy grows faster, and higher tax receipts help the United States manage its debt. That's an attractive story - especially when it turns a messy fiscal debate into something that feels like it should have a clean ending. But the numbers don't cooperate that easily.
In an analysis for The Guardian, Eduardo Porter argues that the White House's hopes for AI-led growth echo an old political idea: that faster growth - whether driven by tax cuts or something else - will eventually pay for the bill. The real question isn't whether AI could make the economy more productive. It's whether the extra growth would be big enough, and taxed in a way that actually improves the federal budget.
Those are different questions. A technology can raise output without generating enough public revenue to close a deficit. And building AI capacity isn't costless - it can also create financing pressures that complicate the budget picture. This article walks through the growth targets, the tax mix, and the data-centre payback problem behind the debate, and then lays out a practical way to judge claims that AI can solve a fiscal challenge.
The growth target is much higher than the usual AI forecast
Porter's discussion starts with the idea that the economy needs to grow faster. Treasury Secretary Scott Bessent has pointed to the prospect of AI helping deliver annual growth of around 3%. That rate isn't impossible, but it's also not something the United States has managed to sustain for long stretches this century. A "great year" and a "new normal" aren't the same thing - especially when budgets are drawn up for many years at a time.
Even then, 3% growth isn't the same as balancing the federal budget. The Committee for a Responsible Federal Budget (CRFB) estimates that the economy would need to grow by about 4.4% per year to reduce the deficit to 3% of gross domestic product by 2036. Under the figures cited in Porter's analysis, balancing the budget would require annual growth of roughly 7.2%.
Those gaps matter. They show how far the growth goal being discussed is from the growth rates tied to much different budget outcomes. If AI helps growth, it would have to do it - consistently and on a large enough scale - to move those projections in a meaningful way.
The difference between the deficit and the debt
People use these terms as if they're interchangeable, but they aren't. The deficit is the annual gap between government spending and revenue. The debt is the total stack of money accumulated from past borrowing. Cutting the deficit means borrowing less going forward; it doesn't automatically mean the debt is shrinking.
Economic growth can affect the debt story through more than one route. A bigger economy can expand the tax base and make existing debt smaller relative to GDP. But how much these effects help depends on the growth rate, tax rules, spending commitments, and interest costs. A higher GDP number doesn't automatically turn into a balanced budget.
That distinction is worth keeping in mind when a political argument jumps from "AI could improve productivity" to "AI will fix the debt." The first claim is about economic output. The second is about federal finances. To connect them, you have to figure out how much extra output is expected, how long it lasts, and what share actually shows up as revenue for the Treasury.
Forecasts need a baseline and a time horizon
The Congressional Budget Office, as Porter reports, expects AI to add about 0.1 percentage points to annual growth. That's small compared with the growth rates associated with the CRFB's deficit and balance targets. And it's still a forecast, not a guarantee - because the long-term effect of a new technology depends on adoption, real productivity gains, and how fast businesses change what they do.
None of this is a claim that AI would have no economic impact. A modest annual boost could still matter over time. The key point is that "AI increases growth" and "AI increases growth enough to resolve a fiscal gap" are not the same statement. Good policy arguments have to show the bridge between them, not treat potential as a stand-in for budget arithmetic.
Economic growth does not automatically become tax revenue
Even if AI pushes growth beyond current forecasts, the government's share of that growth depends on who earns the money - and how that money is taxed. Porter's analysis points to a distributional problem: AI could shift income away from labour and toward owners of capital, including investors and businesses that hold productive assets.
That matters because, in the United States, capital income is taxed at about half the rate of labour income, according to the analysis. If a larger share of the gains ends up as returns to capital instead of wages, the same increase in total output could translate into less public revenue than a growth scenario where workers see bigger paychecks.
This isn't an argument that capital income "should" be taxed at any particular rate. It's just a reminder that the tax system determines how economic activity turns into government revenue. You can't answer the fiscal question with a GDP forecast alone.
Follow the income, not just the productivity claim
When you look at an AI growth estimate, it helps to ask where the benefits are expected to show up. They might look like:
- Higher output per worker, which could support higher wages or more production.
- Lower operating costs and higher profits for firms adopting AI.
- Returns to investors and owners of computing infrastructure, software, and other assets.
- New products and services that create revenue outside traditional technology sectors.
Those outcomes can overlap, and the split depends on how companies actually use AI. A firm might expand services and hire more people - or it might produce similar output with fewer labour hours. The fiscal consequences aren't identical. If the gain shows up more as profits or asset returns than as taxable wages, the government may collect less from each extra dollar of output under current rules.
That's why forecasts that focus only on productivity can miss the real question. Productivity tells you how efficiently the economy turns inputs into output. It doesn't tell you who captures the gains, whether they're taxed, or when the tax arrives.
A bigger economy can still have a difficult budget
A growing economy can make debt easier to manage relative to GDP, assuming everything else stays still. The trouble is that "everything else" never stays still. Government spending and revenue don't automatically adjust to keep a stable relationship with output. A bigger tax base also doesn't guarantee revenue will grow faster than spending and interest costs.
In practice, an AI-driven expansion could come alongside changes in wages, profits, investment, and public spending. With all those moving parts, it's risky to treat a technology forecast like a fiscal plan. A credible budget case needs to spell out the expected revenue gains and account for how the income tied to AI is taxed.
If you want a reference point, compare the assumptions in the Committee for a Responsible Federal Budget's fiscal analysis with the Congressional Budget Office's economic projections. They tackle different questions, but they're both more directly relevant to budget claims than broad talk about AI's potential.
AI infrastructure adds a financing and payback question
AI isn't just "software" running on low-cost hardware. Scaling current systems takes more computing power, data centres, chips, electricity, networking - and a lot of capital. Porter's analysis notes that major tech companies are investing heavily in that infrastructure, and that many are borrowing to help finance data-centre construction.
That investment can support future productivity and business growth. It also competes for capital in a market where the U.S. government is already borrowing at a high level. Porter notes that U.S. debt has passed 100% of GDP. The point isn't that private companies and the federal government borrow for the same reason. The point is that both are drawing from a wider capital market - and when financing needs are large, the expected return on investment becomes a big deal.
A data centre isn't "productive" just because it exists. The infrastructure has to be used enough - and it has to generate enough economic value - to justify what it costs. If demand, pricing, or utilization falls short, returns can disappoint. That matters both for the businesses paying for the buildout and for any public argument assuming the AI boom will reliably create a large fiscal payoff.
The revenue hurdle behind the AI buildout
Economists Jared Bernstein and Ryan Cummings estimate that the largest cloud firms would need between $13.1 trillion and $18.7 trillion in additional revenue over a decade to justify their spending, based on the figures cited in Porter's analysis.
That estimate describes the scale of commercial return needed to support this kind of investment - it isn't a prediction that the firms will definitely earn that revenue, and it isn't a direct calculation of federal tax receipts. The distinction matters. Company revenue is not the same as profit, and neither is the same as government revenue. Still, the estimate makes one practical issue hard to ignore: the investment story depends on substantial future demand.
The full chain runs like this:
- Businesses and consumers adopt AI products and services.
- That demand produces revenue for providers and their suppliers.
- Revenue exceeds the cost of infrastructure, energy, financing, and operations.
- Profits and productivity gains spread across companies, workers, and investors.
- Part of the resulting income becomes tax revenue under current rules.
Weakness anywhere along the chain can shrink the fiscal benefit. Strong adoption might still create value, but lower prices or high infrastructure costs could limit profits. Profitable providers might generate tax revenue, but the amount depends on where income is earned and how it's taxed. Those are separate assumptions - not one guaranteed outcome.
What happens if the payback disappoints?
If AI infrastructure spending produces less revenue than investors expect, companies may delay projects, cut spending, or push for better returns. That doesn't automatically trigger a U.S. debt crisis. But it would weaken the argument that AI investment offers a dependable path to faster growth and better public finances.
The broader risk is a mismatch between today's financing and tomorrow's returns. When a wave of investment is justified by future demand, the forecast is doing real work. If expectations are missed, the effects can go beyond one firm: capital may become more cautious, planned projects may get revisited, and the economic gains used to support optimistic growth assumptions may come later - or come in smaller amounts.
This is something to evaluate, not something to announce with confidence. AI could create substantial value. The question is whether the value generated across the economy will be large enough, persistent enough, and broadly taxable enough to affect the federal budget in a meaningful way.
How to assess the claim that AI can ease the debt burden
You don't have to predict exactly how AI will develop to evaluate the fiscal argument. You just need to separate the assumptions and check whether each one is supported. A useful checklist has four parts.
1. Check the growth assumption
Start with the annual growth rate the argument relies on, the time period, and how it compares with independent forecasts. A claim about one unusually strong year isn't the same as a claim about a decade of faster growth. Compare the proposed rate with the CBO's estimate mentioned in Porter's analysis, and with the growth rates the CRFB ties to specific deficit outcomes.
2. Check who receives the gains
Look for an explanation of how AI-related gains are divided among wages, profits, and returns to capital. If the case assumes higher output but says little about income distribution, it may leave out a key part of the tax calculation. The tax treatment of gains matters as much as the headline productivity number.
3. Check the revenue mechanism
Figure out how higher economic activity is supposed to increase federal receipts. Is the claim based on taxable wages, taxable profits, or something else? Does it reflect current tax rules? Does it distinguish between gross company revenue and taxable income? Without this step, "growth pays for the debt" stays a slogan instead of becoming a budget estimate.
4. Check the investment payback
Consider the spending required to build and run AI infrastructure, and whether expected demand can actually support that investment. The estimates for cloud-company revenue requirements are a reminder that the commercial case has its own hurdle. If the investment doesn't earn an adequate return, then the growth and tax benefits may be lower than assumed.
This framework doesn't require you to be bullish or skeptical about AI. It asks whether the full chain - technology adoption, productivity, income, taxes, and budget outcomes - has been spelled out. That's a better test than asking whether AI is "big enough" to transform the economy in the abstract.
Conclusion
AI could raise productivity and create valuable new businesses. Neither automatically fixes the U.S. budget. The fiscal case depends on a chain of assumptions: sustained growth, a favourable distribution of income, gains that are taxable, and investment returns strong enough to support the infrastructure being built.
The figures cited in Porter's analysis highlight the gap between AI's estimated contribution to growth and the rates tied to major deficit reduction or a balanced budget. They also point to a less obvious issue: growth doesn't turn into public revenue at a steady rate when income shifts between workers and owners of capital.
So treat "AI will grow America out of debt" as a claim to test, not a forecast to accept. Start with the growth estimate, follow who captures the gains, and ask how much reaches government revenue. For more context, read the Guardian analysis by Eduardo Porter alongside the CBO and CRFB projections. The technology could be important; the budget arithmetic still has to work.