Can AI Growth Save the US Deficit? Why Economic Realities Tell Another Story
US officials claim AI-driven economic expansion will offset soaring debt, but massive infrastructure costs and fiscal data suggest serious headwinds.
The promise that rapid technological and economic expansion will eventually pay for federal deficits has returned to Washington, this time anchored in artificial intelligence. Treasury Secretary Scott Bessent has pointed to an AI-accelerated economy to support forecasts of 3% sustained annual growth—a benchmark rarely hit this century outside of the post-pandemic rebound. President Donald Trump has similarly argued that unprecedented growth will help resolve roughly $40 trillion in national debt. However, current financial indicators and economic models suggest that counting on AI to balance the federal ledger faces steep arithmetic hurdles.
Bond markets have responded with caution rather than optimism. Yields on the 10-year US Treasury bond recently climbed to their highest level in nearly twenty-five years, driven in part by inflation linked to military action against Iran and persistent fiscal imbalances. The federal deficit currently stands near 6% of GDP, with Congressional Budget Office estimates projecting it could approach 7% by 2033. Adding to this, legislative measures like the One Big Beautiful Bill Act are estimated to tack on another $4.7 trillion to the debt through 2035.
The steep arithmetic of productivity gains
To understand the scale of growth required to bridge the budget gap, the Committee for a Responsible Federal Budget (CRFB) outlined several projections. Assuming existing temporary tax provisions are extended and tariff revenues struck down by the Supreme Court remain uncollected, balancing the budget over the next decade would necessitate an annual growth rate of 7.2%. Simply lowering the deficit to 3% of GDP by 2036 would require 4.4% annual growth.
While some speculative economic models have imagined dramatic spikes in output as automated systems take over cognitive tasks, achieving sustainable debt stabilization requires total factor productivity to expand at roughly 2.5% annually. The United States has achieved that rate only once since 1959, despite historic technological waves such as electrification, national infrastructure buildouts, and early enterprise computing. Furthermore, because AI-driven gains tilt returns heavily toward capital rather than labor—and capital is taxed at roughly half the rate of labor—gross revenue collection for the government may not match overall output gains without substantial policy adjustments.
Big tech's borrowing race
At the same time the US Treasury is looking for buyers for its debt, it is actively competing for capital with the major tech companies building out AI infrastructure. With foreign central banks reducing their relative Treasury holdings, private markets must absorb both sovereign debt and the unprecedented debt issuances of technology superscalers.
Tech companies are channeling hundreds of billions into infrastructure. According to Stanford economist Hanno Lustig, datacenter investments are estimated to reach $1.43 trillion this year alone. Lustig calculates that for datacenter operators to break even on those outlays, their revenues would need to climb by an average of 45% each year over the next seven years, representing roughly 9.2% of GDP by 2032. Complementing this view, research by Stanford economist Ryan Cummings and Jared Bernstein, former chair of the Council of Economic Advisers, estimates that six dominant tech firms—Google, Meta, Microsoft, Oracle, SpaceX, and Amazon—would need between $13.1 trillion and $18.7 trillion in new revenue over the coming decade to justify their current AI infrastructure bets. That total matches what those same companies brought in collectively over the entire preceding decade.
What it means for developers
For engineering teams and software developers building on frontier models, the macroeconomic backdrop highlights an important shift: the era of subsidized, low-friction infrastructure will inevitably face commercial reality. As superscalers face intense pressure to turn massive capital investments into trillions of dollars in real revenue, pricing models, infrastructure access, and API costs across the industry are bound to tighten.
Teams building software can no longer design applications assuming compute will continuously become cheaper without trade-offs. To avoid vendor lock-in and keep budgets predictable, engineering teams are increasingly diversifying their model providers. Developers looking to stay flexible can try top AI models cheaply through one API at https://apixoai.online, allowing them to benchmark performance across different systems without managing multiple corporate contracts. As public debt and private datacenter spending collide, maintaining cost control and technical agility will be essential for developers navigating the next phase of the AI rollout.
Source: US’s Reagan-era economic promises return as Trump’s AI-fueled growth fantasy — The Guardian AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
