Latest / Minds, Bodies, and Terawatts / OpenAI Proposes 'Superintelligence New Deal'

Transcript

OpenAI Proposes 'Superintelligence New Deal' · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, where the machines write the policy memos now too. Axios reported on April 6th that OpenAI has proposed what it's calling a "Superintelligence New Deal," a package that includes a nationally managed Public Wealth Fund, robot taxes, and AI dividends paid directly to every American citizen.
• 0:15
  [Guest] And the timing is worth noting. OpenAI is about to go public. So the company that just closed the largest private funding round in history is now handing Congress a blueprint for redistributing AI profits. Sam Altman is simultaneously building the thing and proposing the tax code for the thing.
• 0:31
  [Host] Norway had fifty years and a sea floor full of oil to build its sovereign wealth fund. OpenAI has a chatbot and a roadshow. So what's actually in this proposal?
• 0:40
  [Guest] A Public Wealth Fund modeled on sovereign wealth funds. As AI-driven productivity generates enormous corporate profits, a slice gets captured into a national fund, and citizens receive dividends. Alaska does a version of this with the Permanent Fund, paying residents about $1,600 a year from oil revenue. OpenAI is proposing the AI equivalent, except built in years instead of decades, on revenues that don't exist yet.
• 1:00
  [Host] Okay, so a public wealth fund sounds great in a pitch deck. But Bloomberg reported OpenAI just raised $40 billion in private funding. Where does the public money actually come from?
• 1:10
  [Guest] Robot taxes. The proposal calls for shifting the tax base away from payroll taxes and toward levies on automation and robotic labor. A 2020 Brookings study by Acemoglu, Manera, and Restrepo showed the effective tax rate on capital invested in equipment and software is about 5 percent, while the effective labor tax rate sits above 28 percent.
• 1:27
  [Host] So the tax code is already picking sides.
• 1:29
  [Guest] Literally subsidizing replacing humans with machines. OpenAI is saying flip that.
• 1:33
  [Host] If you're a manufacturer in Ohio who just automated your line, you're hearing "we're going to make your investment more expensive."
• 1:39
  [Guest] Right, and the National Association of Manufacturers has fought every automation tax proposal for years. Their argument: tax automation, and you push production offshore to countries that don't. You don't stop the robots. You just lose the factories.
• 1:51
  [Host] Is that a real risk, or a lobbying talking point? Because every automation debate for the last decade has ended with someone saying "but the jobs will go overseas."
• 2:00
  [Guest] Both, but it depends on scope. If the U.S. acts alone with a flat robot tax, the offshoring math is obvious.
• 2:06
  [Host] So what's the smarter version? Because if you're a small business owner who just invested in automation to stay competitive, a flat robot tax feels like punishment for efficiency.
• 2:15
  [Guest] A January 2026 Brookings working paper titled "Future Tax Policy in the Age of AI" argues the better targets are economic rents — returns above what's necessary to induce the activity. Land under data centers, spectrum rights, monopoly profits from AI market concentration. Those rents can't easily flee jurisdiction. You tax the concentration, not the hardware.
• 2:32
  [Host] And payroll taxes fund Social Security and Medicare. If you're replacing payroll taxes with robot taxes, you're rewiring the entire social safety net.
• 2:39
  [Guest] Per CBO data, payroll taxes brought in roughly 36 percent of federal revenue in 2024, around $1.7 trillion. If structural unemployment climbs significantly over the next decade, that revenue base erodes whether you implement robot taxes or not. The current system isn't sustainable under AI-scale displacement. The question is only what replaces it.
• 2:55
  [Host] So OpenAI is saying, "We see the problem our technology creates, and here's our proposed fix." But I want to name what this actually resembles. This is a hush-money framework dressed up as social policy. The structure isn't "we'll share productive gains with you." It's "we'll pay you enough not to revolt while we capture the gains."
• 3:13
  [Guest] That's a sharp read, and it maps to how the proposal works financially. The dividends come from taxing a fraction of the surplus. The bulk of the value still accrues to the companies generating it.
• 3:24
  [Host] So it's not really sharing the wealth. It's skimming the wealth.
• 3:27
  [Guest] Right. A real wealth-sharing model would give citizens equity in the AI systems themselves, not a skim off corporate profits. What OpenAI is proposing is closer to a licensing fee paid to the public for the right to automate at scale.
• 3:40
  [Host] A licensing fee with the licensee writing the terms.
• 3:42
  [Guest] And that connects to the deeper problem. OpenAI is proposing policy for an industry where it's one of maybe five dominant players. Historically, industries that write their own regulation write regulation that favors incumbents.
• 3:53
  [Host] Broadband, pharma, banking — the playbook isn't exactly subtle.
• 3:55
  [Guest] Exactly. A robot tax designed by the companies that own the robots might end up with thresholds and exemptions that protect established players while creating barriers for smaller competitors.
• 4:04
  [Host] Let me push on something else the coverage didn't address. Who defines what counts as a "robot" for tax purposes? A chatbot that replaces three customer service agents doesn't look like a factory arm, but it displaces just as many workers.
• 4:17
  [Guest] That's the definitional nightmare. Bill Gates proposed a robot tax in 2017 and hit the same wall. The European Parliament considered and rejected one that year, partly because nobody could agree on what qualified.
• 4:27
  [Host] And nine years later the problem has only gotten harder.
• 4:30
  [Guest] Any workable version needs a definition broad enough to capture AI-driven displacement across software, not just physical automation. And the enforcement question is just as hard — how do you audit whether a company's productivity gains came from AI or from better management?
• 4:43
  [Host] You probably can't.
• 4:44
  [Guest] Which circles back to the rent-taxing approach. Tax the outsized profits from near-monopoly positions in AI. Tax the land, the compute resources, the market concentration. You sidestep the definitional problem entirely.
• 4:54
  [Host] So the proposal is simultaneously forward-thinking and self-serving.
• 4:56
  [Guest] Welcome to corporate policy advocacy. The revenue problem is real. The need for transition mechanisms is real. And the company proposing the solution stands to benefit from controlling the terms.
• 5:06
  [Host] What happens next? Does this go anywhere legislatively?
• 5:08
  [Guest] Not in its current form. It's a white paper, not a bill. But it shifts the Overton window.
• 5:13
  [Host] Because two years ago, robot taxes were still a fringe idea.
• 5:17
  [Guest] And now the most valuable AI company on Earth is endorsing the concept ahead of its IPO. That gives cover to legislators who want to introduce similar proposals. The vocabulary enters the mainstream even if this Congress doesn't act.
• 5:29
  [Host] And it creates a constituency. A few million retail shareholders holding OpenAI stock and receiving a government AI dividend have a personal financial interest in the company's success. That's voter capture through portfolio alignment.
• 5:39
  [Guest] Which is maybe the most sophisticated part of the whole play. You don't need to lobby Congress if you've turned the public into co-investors.
• 5:47
  [Host] Patrick Déglon maps exactly this transition challenge in his article "Where Does Universal High Income Money Come From?" on unscarcity dot ai — what he calls the bootstrap paradox: the gap between when the old revenue system breaks and when the new one starts working. OpenAI's proposal is one attempt to bridge that gap. Whether it bridges it for everyone or mostly for OpenAI is the question worth tracking.
• 6:08
  [Guest] For more on the funding mechanics behind AI-driven economic transitions, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.