Unscarcity
Sign in for free: Preamble (PDF, ebook & audiobook) + Forum access + Direct purchases Sign In By signing in, you get our daily podcast email. Unsubscribe any time.

Unscarcity Research

When AI Goes Public: Shareholders vs. Abundance

SpaceX listed at $1.75T in June, Anthropic targets $2T in November, OpenAI waits for 2027. How shareholder pressure re-imposes scarcity on abundance tech.

23 min read 5165 words Updated October 2026 /a/ipo-shareholder-primacy-vs-abundance

Note: This is a research note supplementing the book Unscarcity, now available for purchase. These notes expand on concepts from the main text. Start here or get the book.

When AI Goes Public: Shareholders vs. Abundance

In the first week of June 2026, three of the companies building the machines that are supposed to end scarcity lined up to sell pieces of themselves to the public.

The S&P 500 closed above 7,600 for the first time on June 1. That same day, Anthropic confidentially filed its IPO paperwork with the SEC, chasing a valuation near a trillion dollars on the back of a revenue run-rate that, by the company’s own account, went from about $9 billion at the end of 2025 to more than $65 billion by the end of July. OpenAI, which converted into a for-profit Public Benefit Corporation back in October 2025, had its own confidential S-1 in front of regulators, with a Nasdaq debut reportedly targeted for September. And on June 11, SpaceX priced the largest IPO in history: $135 a share, $75 billion raised, a valuation of about $1.75 trillion, with the stock opening on Nasdaq the next morning.

Three trillion-dollar listings lined up in a single year, the largest wave of financialization the technology sector has ever seen. Four months on, the wave has broken unevenly: SpaceX has been trading since June 12, Anthropic has picked Nasdaq for a listing that has slipped to November, after the midterms, at roughly $2 trillion, and OpenAI has said it will not go public in 2026 at all. Each of those moves is a data point for the question almost nobody is asking about the post-labor future: what happens to a technology whose entire promise is driving the cost of things toward zero, once its owners acquire a legal and financial obligation to keep prices up?

What an IPO Actually Changes

A startup answers to a handful of venture investors who are betting on a story years out. A public company answers to a rotating crowd of shareholders, index funds, and analysts who want to know what you did for them this quarter. The money is the same color. The gravity is not.

Going public installs a new master clock. Earnings land every 90 days. Guidance gets issued, then has to be beaten. A stock that “only” grows 20% in a year where the market expected 35% gets punished as if it shrank. The job quietly shifts from “build the thing” to “defend and grow the number,” because the number is now what your owners, your employees’ net worth, and your own compensation are denominated in.

People reach for the phrase “fiduciary duty to maximize shareholder value,” as if a statute somewhere commands every CEO to squeeze. The legal reality is softer and, for our purposes, scarier. As the legal scholar Lynn Stout spent a career arguing, no American law actually orders a company to maximize share price; Delaware’s business-judgment rule gives boards wide latitude to invest, wait, and weigh other interests. Shareholder primacy is enforced less by courts than by the market itself: the quarterly treadmill, the threat of an activist investor buying in and demanding cuts, the analyst downgrade, the simple fact that a falling stock makes it harder to raise money and keep talent. It is a norm with teeth, not a clause. Which is exactly why it reaches companies that swore they were built differently.

Four Months of Tape: What SpaceX Taught OpenAI

SpaceX is the control experiment, because it is the one that actually listed. It priced at $135 on June 11, opened at $150 on June 12, and closed its first day at $160.95, a market value around $2.2 trillion and enough to make Elon Musk, who holds about 42% of the shares and 82.4% of the votes, a trillionaire on paper. Within days the stock touched roughly $226 intraday. Then the gravity took over. By July 15 it had fallen back through its offer price, and in early August it bottomed at a close of $108.27, down more than half from that high, before recovering to about $150 by early September, still a third below the peak. It has gone nowhere since: on October 14 it closed at $152.16, a market value near $2.07 trillion. The slide was about share supply and spending plans, not about the rockets. 911 million insider shares unlocked on August 6, and smaller tranches have come free roughly every three weeks since, with the next on October 24. Up to 1.3 billion more shares can trade two days after the third-quarter report, which SpaceX had not dated by mid-October; the whole lock-up ends on December 8, and Musk’s own stake, nearly half the company, stays locked until June 2027. The market was pricing the sellers, not the product.

OpenAI watched and blinked. The New York Times reported in June that it was weighing whether to hold off while SpaceX tumbled. On August 19 its CFO, Sarah Friar, told an all-hands that the company “will be a public company in 2027,” sooner only if “our business continues to inflect,” and gave the cleanest one-line definition of what an IPO is to a frontier lab: “not a finish line, it is a milestone, another fundraise.” On September 12 Sam Altman made it official, telling Fortune that “given everything happening with safety, right now would be an ill-advised moment to go public”, and, asked whether 2027 was the year instead: “I would say not 2026.”

What followed in the four weeks after that interview is this article’s argument in miniature. The safety Altman invoked took the form, on September 29, of the White House Accord on Super Intelligence, signed over lunch by Trump, Dario Amodei, Sundar Pichai, Mark Zuckerberg, OpenAI’s Greg Brockman, Jensen Huang and Elon Musk; it is voluntary, not legally binding, and sets no penalty for a company that ignores it. The next day a Senate subcommittee held the first hearing on rogue AI attacks; Altman did not respond to the invitation, and the committee heard that the evaluator METR had been allowed to investigate one incident and given six days on the premises. Around October 1, OpenAI fired three safety researchers, Jasmine Wang, Tomek Korbak and Mikita Balesni, for allegedly sharing confidential information with an outside safety organization; Korbak had been the company’s main technical point of contact with METR, and Balesni wrote that “we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation.” The fundraise, meanwhile, never paused: Bloomberg reported talks for at least $30 billion at about $1.4 trillion, a round still unsettled in mid-October. On October 9 Reuters reported that OpenAI had told investors its September annualized revenue was almost $50 billion, not the roughly $70 billion signaled two weeks earlier, a figure that had been “grossed up” to match Anthropic’s accounting.

Read the two explanations together. The CEO’s public reason is the mission. The CFO’s internal reason is the fundraise. Both run on the same clock. A company that raised $122 billion in March can afford to wait for a better tape; one that could not would be listing into it. Either way, the moment the public gets to own a piece of the machines is set by what the market will pay for the piece, which is the whole argument of this article in one delay.

The Collision With Abundance

The optimistic case for AI, robots, and cheap energy is the case for zero marginal cost. Once a model is trained, serving one more query costs a rounding error. Once a robot can build housing, the second house costs little more than materials. The natural resting state of a maturing digital technology is nearly free.

A publicly traded company cannot let its product drift toward free. Free has no margin, and margin is the whole game. So the incentive, the moment the IPO bell rings, is to take a technology that wants to be abundant and engineer scarcity back into it on purpose:

  • Meter it. Price by the token, the seat, the API call, the “credit.” The same model that could be ambient and nearly free instead arrives with a usage bar and an overage fee.
  • Tier it. Hold back the best capabilities for the enterprise plan. Ship the public a deliberately throttled version so the premium one has something to sell.
  • Moat it. Keep the weights closed, the training data proprietary, the integrations exclusive. Network effects and lock-in aren’t accidents of the technology; they’re the business model defending itself.

None of this requires villainy. It is what responsible management owes its shareholders once the technology could otherwise commoditize itself into a public utility. The cruelest irony of the abundance era is that the more genuinely abundant a technology becomes, the harder its owners must work to make it feel scarce, because scarcity is the only thing a market knows how to price.

This is the same fork the OpenAI $122 billion round already exposed, when the company opened its raise to retail investors and manufactured a political constituency that would howl at any regulator who threatened the share price. An IPO scales that maneuver to the entire investing public. A trillion-dollar float doesn’t just fund the technology; it conscripts millions of voters into protecting whatever pricing power produced the trillion.

“But They’re Public Benefit Corporations”

Here is where a careful reader objects, correctly. Anthropic and OpenAI are not ordinary corporations. Both are Public Benefit Corporations, a structure that explicitly authorizes the board to weigh a stated mission, in their case the safe development of advanced AI, alongside financial return rather than beneath it. Anthropic goes further: a Long-Term Benefit Trust holds a special class of shares, Class T, that carries no economic value and exactly one power, appointing directors. The Trust elects four of the seven directors; its three sitting trustees, a global-health CEO, a think-tank director and a former Federal Reserve chair, own no equity, and shareholders can fire the whole Trust only with 85% of the voting power. OpenAI’s nonprofit parent, renamed the OpenAI Foundation, kept about a quarter of the equity and a governance role.

This is a real and serious attempt to inoculate the mission against exactly the gravity described above. It deserves credit, not a sneer. The founders of these labs clearly saw the trap.

The trouble is that the benefit-corporation shield has never been stress-tested at trillion-dollar public scale, and the IPO is the stress test. Benzinga, surveying Anthropic’s filing, flagged the company’s own mission as the “unique risk” it must disclose to investors. Sit with that. The thing that makes Anthropic worth buying, its discipline, is filed in the prospectus as a hazard to the people buying it.

A PBC board may weigh mission; it is not thereby immune from a stock that craters when it leaves money on the table, from securities lawsuits alleging it misled investors about its commercial aggression, or from an activist accumulating shares and demanding the safety brakes come off. And the standard workaround, a dual-class share structure that lets founders keep voting control with a minority of the economics (the Google playbook), solves shareholder capture by handing the company to a handful of insiders in perpetuity.

Anthropic is building exactly that. The Information and Bloomberg reported in August that the founders would receive super-voting shares before the listing, because years of dilution have left Dario Amodei with roughly 2% of the equity. The draft prospectus Reuters reviewed at the end of September spells out the mechanism: the seven founders will jointly direct a single Class F share, held through a Founder LLC, that carries 50.1% of the company’s total voting power. Those rights begin to sunset only when two or fewer founders or their successors remain. The Trust picks four directors; Class A and Class F holders elect the other three, with the founders holding the controlling vote. So the company heading to Nasdaq carries two locks at once, a founder share on one axis and an unelected trust on the other, and the public will hold Class A shares that can move neither. That is not a victory over concentration. It is the Iron Law of Oligarchy wearing a halo: you escape the tyranny of the quarterly crowd by installing the tyranny of the founder. Either the public can rein in the mission, or it can’t and a founder-king can. “Mission-locked and publicly traded at two trillion dollars” is a bet that has never paid out, because it has never been placed.

The public prospectus had still not reached EDGAR in mid-October, with the roadshow now expected from the week of November 9 and trading before Thanksgiving, while Nvidia weighed a $10 billion anchor order in a deal sized at up to $100 billion. What the draft already says is enough. 2025 revenue of nearly $4.6 billion against an operating loss above $8 billion, and a net loss of about $42 billion of which $34 billion is an accounting charge. At least $518 billion of compute commitments over a decade, about 80% of it non-cancelable. And a risk factor acknowledging that the governance structure “could lead to decisions that conflict with the short-, medium- or long-term financial interests of shareholders.” That is the Benzinga headline above, now in the company’s own words. On October 14, at an investor event in San Francisco, Anthropic said its third-quarter revenue reached $16.8 billion, up from $11.6 billion in the second quarter and $4.7 billion in the first, and that it was profitable on an adjusted basis.

Why This Is the Book’s Whole Argument

Unscarcity opens on a crossroad: the same technologies can deliver Star Trek (abundance as shared infrastructure) or Star Wars (a technological aristocracy with the rest of us on a stipend). The three scenarios turn on a single variable, who owns the means of abundance, and the IPO wave is that variable resolving in real time, in public, on a ticker.

Most critiques of elite capture imagine it arriving through the robots, through some sci-fi moment when Optimus refuses to share. The 2026 IPO wave shows capture arriving through something far more boring and far more durable: the cap table. You don’t need the machines to revolt. You need only bind the machines’ output to a security that millions of people own and that is legally and culturally obligated to appreciate. Once abundance is the collateral behind the S&P 500 and the index fund in every 401(k), making it genuinely free becomes an attack on everyone’s retirement. The system will defend its scarcity the way an immune system defends a body.

This is why the book insists that you cannot redistribute your way out of the problem after the fact, and cannot finance your way into post-scarcity at all. A cash transfer like Universal High Income leaves the ownership structure untouched and simply routes a dividend from the owners to everyone else, at the owners’ pleasure. The IPO wave is that structure hardening into law before the transfer is even on the table.

The Democratization Mirage: Owning a Share Isn’t Having Access

There’s a hopeful reading of all this, and by June 2026 it had gone bipartisan. If the danger is that a few insiders own the abundance machines, then surely the fix is to let everyone own them: put OpenAI in every 401(k), hand the public a stake. Senator Bernie Sanders proposed exactly that. His American AI Sovereign Wealth Fund Act would have the federal government take a 50% equity stake in the major AI labs and route the dividends to citizens. The bill he introduced on June 18, S. 4825, spells out the mechanism: a one-time 50% tax, paid in stock, on any company with more than $200 million a year in AI revenue, the shares parked in a Treasury trust run by a seven-member commission, and 5% of the fund paid out every year, about $1,000 per American by his own estimate. Sam Altman took the meeting. President Trump, from the far end of the political spectrum, floated his own version of public equity. When Sanders and Trump start nodding at the same idea, it’s worth asking what they’ve both skipped past.

(Nodding is as far as it went: as of mid-October S. 4825 had no cosponsors, no hearing, and no action since its referral to the Finance Committee on June 18. The administration’s version got no further. In June Treasury Secretary Scott Bessent wanted AI equity to seed “Trump Accounts” while Commerce Secretary Howard Lutnick wanted a sovereign wealth fund, and in July OpenAI offered the government a 5% stake to defuse the pressure; nothing has been decided since.)

What they’ve skipped is the gap between owning a share of abundance and having access to it, and that gap is the whole game.

A share is a claim on the profit a company extracts. Access is using what the company makes. Those aren’t merely different; they pull in opposite directions. A dividend is only worth something if the product stays expensive enough to throw one off. The instant the technology does the single thing it was built to do, drive the cost of intelligence or energy or housing toward zero, the profit behind your share goes with it. So a nation of citizen-shareholders has been handed a financial stake in keeping abundance scarce. Owning a wider slice of the toll booth never made the road free; it just gives more people a reason to defend the toll.

It is the cap-table capture from the section above, now scaled to the entire electorate. Sanders’ fund would put abundance in every American’s portfolio by statute, the most efficient machine ever built for converting “make it cheaper for everyone” into “you’re raiding my retirement.” SpaceX is the live demo: a week before its trillion-dollar debut, it disclosed a $30 billion deal to rent Google 110,000 GPUs, and the market promptly repriced the rocket company as a compute landlord. In September its chief financial officer told a Goldman Sachs conference that a second hosting deal, with a customer it did not name, would bring in about $1.11 billion a month from December 1. Every GPU-hour it might point toward making AI cheaper now reads as revenue owed to shareholders. Wire that conflict into an indexed economy and the trap clicks shut.

Broad ownership doesn’t even deliver broad control. Give the public half the shares and the founders still hold the keys: the dual-class votes, the closed weights, the board, the roadmap. That is the Iron Law of Oligarchy running one floor up, in the capital markets. You can hand the upside to millions while the power stays welded to a few. A dividend is not a vote, and a vote is not the roadmap. Spreading ownership by headcount while control stays concentrated by design is how you manufacture the feeling of democratization without the fact of it.

By June 2026 the idea had jumped from politicians to the people building the machines. In an essay titled “Policy on the AI Exponential”, Anthropic’s Dario Amodei conceded that AI-driven job loss “may be an intrinsic property of the technology” rather than a passing disruption, and floated long-term income support funded by taxing AI firms, raising the capital-gains tax, or routing money through “universal capital accounts.” It is a remarkable thing to hear from a founder whose own company is filing to go public on that same automation. But a capital account hits the identical wall as a sovereign-wealth share: it only pays out if whatever it holds stays profitable, which means it only pays out if abundance stays priced. A balance in a citizen’s account and a check in a citizen’s mailbox are the same bet in different clothes, that you can solve a scarcity problem by handing out better claims on a system whose entire logic is to keep the thing you need expensive.

So “let everyone own a piece,” whether through a retail float, a federal wealth fund, or a personal capital account, is a scarcity-era answer to a post-scarcity question. It routes a dividend and leaves the meter running. The book’s answer runs the other way.

The Unscarcity Answer: Design the Ownership, Don’t Bolt It On

If financialization is the mechanism of capture, then the response cannot be a better tax bolted onto the same corporate form. It has to be a different form.

That is what the book’s EXIT Protocol is for. Instead of asking a public company to behave against the financial gravity acting on it, the Enterprise EXIT converts productive organizations into Mission Guilds: entities whose purpose is the output, not the share price, and where stewardship replaces ownership. A company that has taken the EXIT has no quarterly number to defend, so it has no reason to meter a technology that wants to be free. Founder Status and legacy mechanisms let the people who built the thing convert their stake into durable recognition rather than a perpetual claim on everyone else’s access.

Underneath that sits the Foundation: the essentials, housing, food, energy, healthcare, connectivity, provided as infrastructure rather than sold as products. You cannot be squeezed on the price of a thing that has no price. The Foundation is the part of the economy deliberately removed from the cap table, so that no IPO, no activist, no quarterly miss can re-tariff your survival.

The contrast with the IPO wave is the contrast between two theories of how scarcity ends. One says: let the most capital-rich companies on Earth build the abundance machines, take them public, and trust that owners with a fiduciary obligation to extract will somehow choose to release. The other says: the obligation to extract is the problem, so build the machines inside structures that have no such obligation. The first is the road the market is paving this week. The second is the one the book draws a map for.

What to Watch

The IPO wave is not a villain origin story; it is a measurement. As these listings price and trade, watch for the tells that the financial gravity is winning:

  • Capability tiering by ability to pay — the gap between the free model and the paid one widening rather than shrinking as the technology matures. OpenAI’s GPT-6 launch on October 7 is the current reading: paying tiers get GPT-6 Sol, the free and Go tiers get GPT-6 Luna.
  • The benefit mission getting “clarified” in later filings, investor letters, or post-IPO governance changes, the slow conversion of the founding promise into a branding asset, exactly as happened to OpenAI’s original nonprofit charter. The Financial Times reported in August that OpenAI had shut its Preparedness team, the group that assessed catastrophic model risk, at the end of July and parceled its work out to other groups; the October firings above are the sequel. Watch for the same shape at Anthropic once there is a share price to defend.
  • Activist arrival — the first time a large fund buys in and publicly demands the safety or pricing discipline be loosened “to unlock shareholder value.”
  • Lobbying that protects margin, not the public — the retail-investor shield deployed against any policy that would push prices toward cost.

Each one is a small confirmation that a technology capable of ending scarcity is being domesticated into one more thing that has to be paid for, forever.

The robots really are coming, and they really could make almost everything nearly free. Whether they do is being decided right now, not in a lab, but in a prospectus. The labor cliff will arrive on schedule regardless of who owns the upside. The only open question, the one the ticker has been answering since June, is whether the abundance lands in the commons or on a balance sheet.

Unscarcity was written to argue that it doesn’t have to be the balance sheet. Read the blueprint.

Sources

Share this article:

Share Share Share