What the 1990s-Era “Bit Tax” Can Teach Us About AI Tax Proposals

As US investment in artificial intelligence (AI) accelerates, so too have proposals to uniquely tax AI. Some proposals are broadly aimed at large corporations and the wealthy. Others are narrowly targeted at AI companies and technologies, such as the tokens flowing in and out of AI models or the compute used to train them.

These proposals share DNA with a 1990s-era idea borne out of concern over the World Wide Web and its potential impacts on the economy: the bit taxA tax is a mandatory payment or charge collected by local, state, and national governments from individuals or businesses to cover the costs of general government services, goods, and activities.. The bit tax would have applied to uploads and downloads of information over the internet. Some observers likened the idea to an email tax, and the proposal drew bipartisan opposition in the US for being complex, non-neutral, and anti-growth.

Were the bit tax enacted and maintained as originally proposed, it would have either soaked consumers and businesses in higher taxes today—acting as a toll on internet usage—or prevented new and data-intensive technologies from being developed at all. While the economic, technological, and budgetary outlook now differs significantly from that of the 1990s, lawmakers can still draw valuable lessons from the rise and fall of the bit tax.

What Was the Bit Tax?

The original bit tax proposal came from Canadian economist Arthur Cordell. He presented the idea at a 1995 conference, in remarks that bear a striking resemblance to conversations happening today around AI:

• Cordell warned that the internet could “replace people in a great number of functions” and could do things “better, faster and cheaper than by people.”
• He worried that the Canadian tax baseThe tax base is the total amount of income, property, assets, consumption, transactions, or other economic activity subject to taxation by a tax authority. A narrow tax base is non-neutral and inefficient. A broad tax base reduces tax administration costs and allows more revenue to be raised at lower rates. was “threatened by the labour displacing capacity of new technologies,” and that Canada needed a new tax on the “new wealth of nations”: data flows.
• He argued the bit tax would be “easily administered,” “transparent,” and represent a growing tax base “at the heart of the new economy.”

Cordell suggested a tax of 0.000001 cents per bit. Table 1 below demonstrates how such a tax would apply to various digital storage amounts:

Table 1. A Crosswalk from Bits to (Giga)bytes

Source: Ashley Taylor, “Bits and Bytes,” Stanford University, 2018, https://web.stanford.edu/class/cs101/bits-bytes.html; author’s calculations.

Cordell acknowledged in a subsequent 1997 speech that, given technological advances, “the bit tax rate will have to be adjusted for changing times.” He also proposed:

• Limiting the tax to “interactive digital information,” such as emails or calls, rather than mass communications, such as digital broadcasts
• Requiring “telecom carriers, satellite networks and cable systems” to collect the tax
• Levying the tax not based on individual or household use but on local averages

The idea picked up steam in the late 1990s, with a 1999 UN report mentioning a $0.01 per megabyte tax as one way to “fund the global communications revolution.” The European Commission also explored the idea.

Why Did the Bit Tax Fail?

Despite global policymaker interest, the bit tax never came close to enactment. The idea drew swift and bipartisan opposition in the US. President Clinton stated in 1997 that he wanted to keep the internet “free of new discriminatory taxes.” An early and bipartisan version of the Internet Tax Freedom Act (ITFA) explicitly prohibited states from enacting “bit taxes.”

The bipartisan Advisory Commission on Electronic Commerce, established by the final version of ITFA, reported to Congress in 2000 that the bit tax was “met with little support by government officials.”

Three factors buried the bit tax:

1. Complexity: While Cordell argued the bit tax would be “easily administered,” others countered that cross-border information flows would be difficult to measure, tax, and enforce. Some worried internet providers would require new equipment to measure bits, like electricity meters. At the international level, there were also complexities with tax coordination among and across nations.
2. Non-Neutrality: The bipartisan ITFA commission, in response to bit tax proposals, stated that “there has been overwhelming support for the principles . . . [of] nondiscriminatory and neutral taxation of e-commerce.” Some experts were additionally concerned that taxing business-to-business digital interactions and transactions (e.g., emails, calls, and online purchases) at a different rate than in-person transactions would violate neutrality.
3. Anti-Growth: Above all, policymakers in the US and Europe were concerned a bit tax would slow innovation and economic growth attributable to the internet. A 1999 paper from the European Parliament noted that a bit tax would not discriminate between data used for economically productive activities versus non-productive ones. Others argued the tax would slow down investments.

What Can the Bit Tax Teach Us About AI Taxes?

Although bit taxes were last considered nearly 30 years ago, they offer some lessons for policymakers considering AI taxes today.

Technology Changes Quickly

A bespoke tax that makes some sense in one era may not make sense in the next. The median US household uses 532 GB of data per month today and would face extraordinarily high bit taxes. Cordell certainly did not intend that kind of impact.

Then again, few envisioned seamless video calling available across continents or thousands of movies available at the touch of a button. A “compute” or “token” tax measured today could end up looking nonsensical in 10, 5, or even 2 years. Table 2 below estimates bit tax liability for common internet uses in 2026.

Table 2. The Bit Tax Applied to Common Internet Uses Today

Source: AT&T, “Common reasons for high data use,” Apr. 29, 2025, https://www.att.com/support/article/wireless/KM1045105/; Victra, “What Uses the Most Data on Your Phone?” Jun. 28, 2026, https://victra.com/blog/what-uses-the-most-data-on-your-phone/; author’s calculations.

Neutrality Is Timeless

In the 1990s, a bipartisan group of executive and legislative branch officials in the US stood on 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.

The online advantage with respect to sales taxes was one that states and the federal government wrestled with in the 1990s and 2000s. The Supreme Court’s Wayfair decision ultimately settled the legal matter, in part by applying the principle of neutrality. 

While some AI tax proposals are broad tax increases on wealth or capital, others are narrowly targeted at AI companies or activities and are non-neutral. Lawmakers today should adhere to the same principles of neutrality as those arguing against bit taxes did in the 1990s.

AI Taxes May Adversely Affect Growth

If the US had enacted a bit tax, Americans today might not enjoy remote work, telehealth, or the broader social connectedness of high-speed internet today. Even at a much lower rate, say $0.01 per GB, a bit tax would have still slowed activities that create jobs and grow the economy.

The labor displacement Cordell warned of ultimately did not come to pass. The rise of the internet even brought with it a financial and tax revenue boom that—temporarily—improved the US economic and budgetary outlook in the late 1990s and early 2000s.

The bit tax episode can’t teach us everything about AI. The 2020s are not the 1990s. US economic growth is slower today than it was back then. The US budget is more vulnerable to major threats to its tax base. And AI could be a more powerful and disruptive technology than the World Wide Web.

However, the bit tax can teach us about the timelessness of core tax principles—simplicity, neutrality, transparency, and stability. It can also serve as a brake on narrow, targeted AI tax proposals that may not stand the test of time. Policymakers should consider the 1990s bit tax a cautionary tale when evaluating complex or non-neutral AI tax proposals in the 2020s.

Share this article