Candice Horbacz
As U.S. investment in artificial intelligence (AI) accelerates at an unprecedented pace, so too have a growing number of proposals to uniquely tax AI technologies and the burgeoning industry surrounding them. These proposals, ranging from broad levies on large corporations benefiting from AI to highly specific charges on "tokens" flowing through AI models or the vast "compute" resources used to train them, share an intriguing historical lineage with a concept born out of a similar era of technological revolution: the 1990s "bit tax." This parallel, often overlooked, offers critical lessons for contemporary policymakers grappling with the complexities of taxing a rapidly evolving digital frontier.
The Rise of AI and the Call for New Taxes
The current era is defined by the transformative potential of artificial intelligence. Global investment in AI is skyrocketing, with projections indicating hundreds of billions of dollars poured into research, development, and deployment annually. Major tech giants and countless startups are competing to develop more sophisticated models, leading to breakthroughs in fields from healthcare and finance to entertainment and logistics. This rapid advancement, however, is not without its anxieties. Concerns about job displacement, the concentration of wealth and power in the hands of a few AI pioneers, and the ethical implications of autonomous systems have fueled calls for new regulatory frameworks, including taxation.
Proponents of AI-specific taxes often articulate several key motivations. Some argue that such taxes could serve as a mechanism to fund social safety nets or universal basic income programs, mitigating the economic disruption anticipated from widespread AI adoption. Others view them as a means to capture a portion of the immense economic value generated by AI, which they believe might otherwise escape traditional tax structures. There’s also a sentiment that AI, as a powerful and potentially disruptive technology, should contribute specifically to public services or research, akin to how certain industries bear specific levies. Proposals vary widely, from those targeting the profits of AI companies or the capital gains of their investors, to highly granular taxes on the processing power (compute) consumed by AI training or the data units (tokens) exchanged by AI models.
Echoes from the Past: The 1990s "Bit Tax" Debate
The current debate over AI taxation bears a striking resemblance to discussions that emerged nearly three decades ago concerning the nascent World Wide Web. In the mid-1990s, as the internet began its rapid ascent from academic novelty to public utility, policymakers and economists grappled with its potential impacts on the economy, employment, and societal structures. It was within this context that the "bit tax" emerged – a proposal to levy a tax on the uploads and downloads of digital information over the internet.
The core idea of the bit tax was to capture value from the burgeoning digital economy, which many perceived as operating outside traditional tax nets and potentially disrupting established industries. Some likened it to a modern-day "email tax" or a toll on digital communication. However, despite initial interest in some circles, the proposal swiftly encountered bipartisan opposition in the United States, largely on grounds of its complexity, its non-neutral application to the economy, and its potential to stifle economic growth and innovation. The parallels to today’s AI discussions are stark: a revolutionary technology emerges, concerns about its impact surface, and proposals for unique, targeted taxes follow.
A Deeper Look at the Bit Tax Proposal
The original bit tax proposal was put forth by Canadian economist Arthur Cordell. He formally introduced the concept at a conference in 1995, delivering remarks that, in hindsight, eerily presage many of the conversations happening today about AI. Cordell envisioned a future where information would become the primary commodity, and traditional economic metrics would struggle to capture its value. He proposed a minuscule tax rate: 0.000001 cents per bit. To illustrate the scale, consider that a single gigabyte, a common unit of digital storage today, contains 8 billion bits. At Cordell’s proposed rate, a gigabyte would incur an 8-cent tax. While seemingly small, such a tax, if applied broadly to today’s data-intensive world, would quickly accumulate. For example, a high-definition movie stream, which can easily be several gigabytes, would incur multiple cents in tax for each viewing.
Cordell himself acknowledged the dynamic nature of technology, suggesting in a 1997 speech that "the bit tax rate will have to be adjusted for changing times" due to rapid technological advances. He also proposed the establishment of a "global fund" to manage the revenues generated by the bit tax, directing them towards initiatives like infrastructure development in poorer nations, a concept that finds echoes in some contemporary discussions about using AI tax revenues for social good or technological equity.
Global Interest and Mounting Opposition
The idea of the bit tax was not confined to academic discussions. It gained significant traction among policymakers globally in the late 1990s. A 1999 United Nations report, for instance, mentioned a tax of $0.01 per megabyte as a potential mechanism to "fund the global communications revolution," envisioning revenues that could support digital inclusion and development worldwide. The European Commission also explored the concept, indicating a broader international curiosity about how to tax the burgeoning digital economy. This global interest highlighted a shared challenge among nations: how to adapt fiscal policies to a world increasingly shaped by intangible digital assets and services.
However, despite this international intrigue, the bit tax faced formidable opposition, particularly in the United States, and never came close to enactment. President Bill Clinton, in a pivotal statement in 1997, declared his intention to keep the internet "free of new discriminatory taxes," signaling a strong executive branch stance against targeted levies on digital activity. This sentiment was codified in early, bipartisan versions of the Internet Tax Freedom Act (ITFA), which explicitly prohibited states from enacting "bit taxes" and other discriminatory internet taxes. The ITFA, first passed in 1998, effectively created a protected space for the internet to grow without the burden of novel, potentially stifling taxes.
The bipartisan Advisory Commission on Electronic Commerce (ACEC), established by the final version of the ITFA, further cemented this opposition. In its 2000 report to Congress, the ACEC noted that the bit tax was "met with little support by government officials," underscoring the broad consensus against it across political lines and branches of government. This unified front against the bit tax proved insurmountable.
Why the Bit Tax Never Saw the Light of Day
The demise of the bit tax can be attributed to several critical factors, each offering valuable insights for today’s AI tax debates:
-
Technological Volatility and Obsolescence: The very nature of digital technology meant that any tax based on specific units like "bits" would quickly become outdated or economically impractical. The exponential growth in data transmission speeds and storage capacities made a fixed per-bit tax either negligible or, if adjusted, potentially crippling. What seemed like a small tax in 1995 would become an enormous burden as data usage exploded, making the tax non-sensical and disproportionate.
-
Complexity of Implementation: Measuring and collecting a bit tax presented immense administrative challenges. The sheer volume of data flowing across the internet, the diversity of its uses (from email to streaming video), and the decentralized nature of its infrastructure made accurate and equitable taxation of individual bits an administrative nightmare. This complexity would have imposed significant compliance costs on businesses and made enforcement difficult for governments.
-
Non-Neutrality and Anti-Growth Concerns: Perhaps the most significant criticism was that the bit tax was non-neutral and anti-growth. It singled out digital information for taxation, potentially disadvantaging internet-based businesses and activities compared to traditional ones. Critics argued it would act as a "toll on internet usage," raising costs for consumers and businesses, and thus hindering the very growth and innovation the internet promised. This concern for economic neutrality – ensuring that taxes do not unduly favor or disfavor specific industries or technologies – was a cornerstone of the opposition.
Lessons for the AI Era: Avoiding Past Mistakes
Although the concept of a bit tax was last seriously considered nearly three decades ago, its rise and fall offer profoundly relevant lessons for policymakers currently contemplating AI taxes.
The Peril of Bespoke Taxation in Rapidly Evolving Tech
One of the clearest takeaways is that bespoke taxes, designed to target specific technological units or activities, are inherently vulnerable to rapid technological change. The median U.S. household today consumes approximately 532 gigabytes of data per month. If Cordell’s original bit tax rate were applied, this household would face an extraordinarily high tax bill of over $42, far beyond his original intent and a substantial toll on daily internet use. Few in the 1990s envisioned seamless global video calling, on-demand streaming of thousands of movies, or the vast amounts of data routinely consumed by smart devices.
Similarly, a "compute" tax measured by teraflops or a "token" tax applied to large language models today could become economically nonsensical in a mere five, or even two, years. AI hardware is becoming exponentially more efficient, and models are constantly evolving in how they process information. A tax structure fixed on current metrics risks either becoming laughably insignificant or oppressively burdensome as the technology shifts, undermining its purpose and creating unforeseen economic distortions. Policies must be designed with the understanding that the underlying technology will not remain static.
Upholding Neutrality in a Digital Economy
In the 1990s, a bipartisan coalition of U.S. executive and legislative branch officials championed the principle of neutrality: internet products and services should neither face discriminatory taxes nor gain an unfair advantage over brick-and-mortar goods and services. This principle was crucial in fostering the internet’s growth. The debate over remote sales taxes, which eventually led to the Supreme Court’s landmark Wayfair decision in 2018, largely centered on applying this same principle of neutrality to ensure online retailers collected sales taxes akin to their physical counterparts.
Many current AI tax proposals, while some are broad tax increases on wealth or capital, are narrowly targeted at specific AI companies or activities, making them inherently non-neutral. Such targeted taxes risk creating an uneven playing field, potentially penalizing innovation in AI while leaving other sectors untouched. Lawmakers today should adhere to the same principles of neutrality that guided the opposition to the bit tax, ensuring that AI is treated equitably within the broader economic and tax framework, rather than being singled out for special, potentially stifling, levies.
The Risk to Economic Growth and Innovation
Had the U.S. enacted a bit tax, it is highly probable that Americans would not enjoy the widespread benefits of today’s high-speed internet, including ubiquitous remote work, advanced telehealth services, and the profound social connectedness enabled by digital platforms. Even at a much lower rate than Cordell’s original proposal, a bit tax would have likely slowed activities that are crucial for job creation and economic expansion, such as cloud computing, data analytics, and digital content creation. The internet, far from causing widespread labor displacement as some feared, ultimately spurred a massive economic boom, bringing with it a temporary improvement in the U.S. economic and budgetary outlook in the late 1990s and early 2000s.
AI, similarly, holds immense promise for driving productivity gains, creating new industries, and solving complex societal challenges. Imposing targeted AI taxes risks stifling this potential, discouraging investment, and pushing innovation offshore. While concerns about labor displacement are valid and warrant policy responses, a direct tax on the underlying technology might inadvertently harm the very economic growth that could fund solutions to these challenges.
A Nuanced Future: The Broader Economic Context of AI
It is crucial to acknowledge that the 2020s are not merely a rerun of the 1990s. The economic landscape today presents different challenges. U.S. economic growth is slower than it was during the dot-com boom, and the national budget faces greater vulnerability due to a significantly higher national debt. Moreover, AI itself could prove to be an even more powerful and disruptive technology than the World Wide Web, with potential impacts that are still difficult to fully quantify.
These differences mean that while the lessons from the bit tax are invaluable, they cannot dictate every aspect of AI policy. Policymakers must consider the unique challenges and opportunities presented by AI, including its potential for unprecedented productivity gains, but also its societal risks. However, the bit tax episode serves as a powerful reminder of the timelessness of core tax principles: simplicity, neutrality, transparency, and stability.
Conclusion: A Cautionary Tale for Policymakers
The story of the bit tax offers a compelling cautionary tale. It demonstrates the dangers of crafting specific tax policies for rapidly evolving technologies without a deep understanding of their future trajectory or adherence to fundamental economic principles. Complex, non-neutral, or anti-growth tax proposals, even if well-intentioned, can stifle innovation and hinder economic progress.
As policymakers navigate the complex terrain of artificial intelligence, they should consider the 1990s bit tax a historical precedent. Instead of rushing to implement narrow, targeted AI taxes that may quickly become obsolete or detrimental, they should prioritize tax policies that are neutral across industries, simple to administer, transparent in their application, and stable enough to withstand technological shifts. Adhering to these core principles will better position the U.S. to harness the transformative potential of AI while avoiding the pitfalls of well-meaning but ultimately counterproductive taxation.
Help Us Improve TaxFoundation.org
Your feedback helps us make our research easier to find and use.
Share Your Feedback
About the Author
Andrew Lautz is Senior Director of Federal Policy with Tax Foundation’s Center for Federal Tax Policy. Before joining Tax Foundation, he was Director of Tax Policy at the Bipartisan Policy Center and Director of Federal Policy at the National Taxpayers Union. Andrew’s research and perspectives on federal tax policy have been featured in The Wall Street Journal, The New York Times, Bloomberg, and other major publications.
