Billionaire investor and Dallas Mavericks owner Mark Cuban proposed a federal tax on artificial intelligence (AI) tokens at less than 50 cents per million tokens. He argues the levy would push large model operators toward efficiency gains while generating billions in annual revenue to reduce federal debt or fund programs addressing AI disruptions. The tax would apply at the provider level, excluding open-source models and local inference. Cuban compared the proposal to the regulatory path of cryptocurrency, noting that early crypto supporters rejected any rules but later embraced structured advocacy as the industry grew. Critics, including Anduril founder Palmer Luckey, warn the tax would make American AI firms less competitive against foreign models and require government tracking of AI usage. The proposal faces slim congressional appetite, but the debate may mirror how crypto firms eventually lobbied for clearer rules.

Key facts

• Mark Cuban proposes a federal tax on AI tokens at less than $0.50 per million tokens.
• Tax aims to drive efficiency in AI models and cut energy use, raising billions annually.
• Open-source models and local inference would be exempt from the tax.
• Critics argue the tax would hurt US AI firms and require government tracking.
• Cuban draws parallel to crypto's evolution from opposing regulation to accepting it.