OpenAI Is Preparing for Its IPO. It's Also Proposing Robot Taxes.

On April 6, OpenAI published a policy document proposing that the US shift taxes from labor to capital, create a public wealth fund to give Americans a stake in AI companies, consider a robot tax, and explore a four-day workweek to absorb AI-driven job displacement. Four days later, CFO Sarah Friar confirmed the company is planning an IPO — potentially by Q4 2026 — at an $852 billion valuation following a $122 billion funding round that was so oversubscribed one bank’s system crashed opening the data room.

These two things are happening simultaneously, and the tension is worth sitting with.

What the policy paper actually says

OpenAI’s proposals are framed as a response to the economic disruption superintelligent machines will cause. The core argument: as AI expands corporate profits while reducing reliance on labor income and payroll taxes, the tax base that funds Social Security, Medicaid, and housing assistance will erode. Their proposed fixes blend left-leaning mechanisms — public wealth funds, capital gains taxes on AI-driven returns, a potential robot tax — with a fundamentally market-driven framework that stops well short of specifying rates or enforcement.

The document is careful to frame these as ideas for elected officials, not commitments. OpenAI describes it as “a public declaration that helps elected officials, investors, and the public understand how we see the world shifting.” The timing was not accidental — the paper landed while the Trump administration is building a national AI framework and with midterm elections approaching, giving the proposals maximum political surface area.

The contradiction that makes it interesting

Greg Brockman, OpenAI’s president, has donated millions to Donald Trump — the president who dropped the corporate tax rate to 21% and whose administration is pushing light-touch AI regulation. OpenAI’s policy proposals, if implemented, would meaningfully reverse that direction. The same week this paper dropped, the company was talking to banks about a public offering that would make its founders and early investors extraordinarily wealthy under the current tax structure.

The charitable read is that OpenAI genuinely believes AI will cause economic disruption that the current system cannot absorb, and is trying to start the conversation before it becomes a crisis. The less charitable read is that publishing left-leaning economic proposals is a cheap hedge for a company that has bet heavily on a right-leaning administration — a way to maintain credibility with workers, regulators, and international governments who might otherwise view them as another Silicon Valley wealth extraction machine.

Both can be true at once.

The week in numbers

The $122 billion round closed April 2, pushing OpenAI’s valuation to $852 billion. CFO Friar told CNBC that when banks opened the data room to retail investors, demand was so intense one bank’s system went down. OpenAI set out to raise $1 billion from individual investors through private placements with JP Morgan, Morgan Stanley, and Goldman Sachs — and ended up raising three times that, in what those banks described as the largest private placement they had ever done.

Enterprise is now 40% of revenue and on track to equal consumer revenue by end of 2026, according to CRO Denise Dresser. Codex has crossed 3 million users — from near-zero at the start of the quarter. Annualized revenue is $25 billion.

Also this week: OpenAI acquired TBPN, The Business Pod Network. The move looks like a distribution hedge — as AI commoditizes search and disrupts how people consume media, owning an audience directly becomes more valuable.

What to watch

The IPO timeline is the primary thread. Friar wouldn’t confirm Q4 but described it as “good hygiene” for a company at this scale to look and act like a public company — which is the kind of statement CFOs make when a listing is closer than they’ll confirm. Watch for S-1 registration activity in Q3.

The OpenClaw thread is worth tracking separately. Peter Steinberger, who built OpenClaw and now works at OpenAI, said he is “working on” OAuth support for third-party harnesses — which would let OpenClaw users access GPT models the same way they accessed Claude before Anthropic changed its pricing. That would meaningfully expand OpenAI’s share of the third-party agent ecosystem at a moment when developers are already evaluating their options.