Vinod Khosla, the venture capitalist behind Sun Microsystems and Khosla Ventures, has floated a provocative idea: that artificial intelligence could grow the economy so dramatically that the United States could afford to eliminate income taxes for roughly 125 million lower-earning Americans. The proposal treats AI not as a mere efficiency tool but as a force capable of restructuring who pays for government and how much. Testing that ambition against the current tax system reveals both the scale of the promise and the gap between vision and operational reality.
165 Million Returns and the Scale Problem
Any serious conversation about erasing taxes for a large share of the population has to start with how big the current system actually is. The IRS processed more than 165 million individual returns in 2025, collecting revenue that funds everything from defense spending to Social Security. Khosla’s target of 125 million people represents roughly three-quarters of that filing population, a group whose collective tax contributions, while individually modest, add up to hundreds of billions of dollars annually. Replacing that revenue through AI-driven economic growth would require sustained GDP expansion well beyond historical norms and a political consensus that the gains should be captured primarily at the top of the income distribution.
The IRS Data Book, which covers fiscal years 2023 through 2024, offers a window into the composition of federal revenue. Individual income taxes make up the largest share of gross collections, and electronic filing rates have climbed steadily, according to the agency’s official statistics. That structure means removing the bottom tier of filers from the rolls would not simply trim a small line item. It would require either higher collections from top earners and corporations or an entirely new revenue mechanism, something Khosla’s AI thesis implicitly assumes but does not detail. Any such shift would also have to account for the cyclical nature of tax receipts, which tend to fall during recessions, precisely when safety-net spending for lower earners typically rises.
Strong Service in 2025, Strain Ahead in 2026
Even without a radical overhaul, the IRS is already struggling to keep pace with its existing obligations. The National Taxpayer Advocate’s latest annual report found that taxpayer service was strong in 2025 but warned of real challenges for taxpayers who encounter problems in 2026. That finding matters because it exposes a tension at the heart of Khosla’s argument: the agency responsible for implementing any AI-powered tax shift is simultaneously bracing for service degradation. Automating compliance for 125 million filers presupposes an infrastructure that does not yet exist and that current budget pressures make harder to build, especially as call volumes, paper backlogs, and identity verification demands remain persistent pain points.
The Advocate’s companion recommendations, outlined in the 2026 policy compendium, focus on practical fixes like improving case resolution timelines and expanding digital tools. Current taxpayer-facing systems, including the agency’s online account portal, handle basic tasks such as payment tracking and transcript requests. But those tools were designed for incremental improvement, not wholesale reinvention. Layering AI on top of a system that already anticipates bottlenecks in 2026 would demand significant new investment in secure data infrastructure, staff training, and algorithmic oversight, precisely the kind of spending that large-scale tax elimination would make harder to fund in the short term.
Where AI Meets Filing Infrastructure
Khosla’s broader thesis rests on the idea that AI will generate so much new economic value that traditional tax structures become obsolete. In theory, if AI-driven productivity gains expand corporate profits and high-end incomes fast enough, the government could collect more revenue from fewer payers while letting lower earners keep everything they earn. This is not entirely without precedent: the earned income tax credit and standard deduction already create an effective zero-tax rate for millions of filers, particularly those with very low wages or qualifying dependents. Khosla’s version simply scales that logic to a much larger population and ties it to a technology bet rather than a carefully scored legislative choice.
The practical challenge is that the IRS infrastructure needed to administer such a shift barely exists in prototype form. The agency’s business online tools and tax professional portal serve narrow functions for specific user groups. Expanding AI across the full filing chain would require not just software upgrades but fundamental changes to how the agency verifies income, detects fraud, and processes refunds at scale. The 165 million returns processed in 2025 each carry different income sources, deduction profiles, and compliance risks. Automating the simple cases is achievable; automating the judgment calls that keep the system honest (such as distinguishing legitimate credits from abusive schemes) poses a far more complex challenge that cannot be solved with generic AI tools alone.
The Revenue Gap No One Has Modeled
The most significant weakness in Khosla’s proposal is the absence of any published model showing how AI-generated growth would offset the revenue lost by exempting 125 million taxpayers. No government agency, independent research organization, or academic institution has produced a study quantifying this tradeoff with the specificity the claim demands. Without that analysis, the idea functions more as a thought experiment than a policy blueprint. It assumes a future state of the economy that may or may not materialize, and it asks policymakers to accept a revenue risk with no empirical safety net. In a budget environment already marked by structural deficits, that kind of leap would represent a dramatic departure from the cautious scoring that typically accompanies tax legislation.
This gap matters because the federal budget already operates under significant fiscal pressure. Individual income taxes remain the single largest source of federal revenue, and the filers Khosla wants to exempt, while individually small contributors, collectively represent a meaningful share of total collections. The IRS Data Book tables make clear that even modest shifts in filing patterns or collection rates ripple through the entire fiscal picture, influencing everything from deficit projections to debt-service costs. A proposal to zero out taxes for three-quarters of filers without a concrete replacement revenue stream is, at minimum, incomplete. At worst, it risks normalizing a vision of AI-powered abundance that distracts from the harder work of making the current system function well for the people already inside it, including low-income taxpayers who struggle with audits, notices, and access to representation.
Ambition Without a Blueprint
Khosla’s AI-driven tax holiday for 125 million Americans captures the imagination because it promises something rare in economic policy: a painless trade in which technology does the heavy lifting while ordinary workers simply keep more of their paychecks. Yet the current state of the tax system, as described by the National Taxpayer Advocate and reflected in IRS operational data, points to a more constrained reality. The agency is still working to stabilize basic services, modernize legacy systems, and expand digital access, not to mention respond to emerging threats like identity theft and sophisticated refund fraud. Against that backdrop, the notion that AI will soon generate enough surplus to underwrite sweeping tax exemptions looks less like an imminent reform agenda and more like a speculative horizon.
That does not mean AI has no role to play in tax administration or economic growth. It could help the IRS target enforcement more precisely, streamline correspondence, and reduce errors, while also boosting productivity in the private sector. But those incremental gains are a long way from the kind of transformative windfall Khosla envisions, and they will require steady investment, clear governance, and realistic expectations. Until someone produces a rigorous model that connects AI-driven growth to specific, sustainable revenue alternatives, eliminating income taxes for 125 million Americans will remain an intriguing slogan rather than a governing plan. Ambition without a blueprint in a system that can ill afford guesswork.
More From The Daily Overview
*This article was researched with the help of AI, with human editors creating the final content.
Julian Harrow specializes in taxation, IRS rules, and compliance strategy. His work helps readers navigate complex tax codes, deadlines, and reporting requirements while identifying opportunities for efficiency and risk reduction. At The Daily Overview, Julian breaks down tax-related topics with precision and clarity, making a traditionally dense subject easier to understand.
