OpenAI’s Invisible Shareholders
OpenAI's Invisible Shareholders
When OpenAI files its IPO prospectus in the coming days — as WSJ reports Goldman Sachs and Morgan Stanley are preparing to list it at $852 billion, the largest private company in Silicon Valley history — public investors will be handed a simple proposition: provide the capital, accept the risk, and hand over the controls.
The fine print will make this explicit. The OpenAI Foundation, a nonprofit, holds 26 percent of the company's equity, along with warrants and special voting rights that give it disproportionate influence over board decisions, according to Investing.com's analysis of the October 2025 recapitalization. Microsoft takes 20 cents of every dollar OpenAI earns through 2032 — a revenue-sharing arrangement renegotiated last October that has no ceiling. Public shareholders — the people whose money is supposed to own the company — are left with economic exposure and no governance.
This is the story no one is covering. The financial press has focused on the filing date, the valuation, and the burn rate. All of that is real and all of it matters. But the structural question underneath is simpler and stranger: who runs OpenAI when the stock market does?
The answer, under the current structure, is: not public shareholders.
The pressure that forced this
OpenAI did not choose this moment. Anthropic did.
By early 2026, OpenAI's leadership had watched a competitor close enough to threaten the gap they spent three years building. Anthropic's annualized revenue reached approximately $30 billion, with internal projections showing its burn rate dropping to roughly 9 percent of revenue by 2027, per Investing.com. OpenAI's burn rate was running at 57 percent, with no path to profitability before 2029 at the earliest. When Anthropic agreed to a funding round valuing it at $900 billion — surpassing OpenAI's $852 billion valuation — the signal to every institutional investor paying attention was clear: there was now a comparable AI company with better unit economics and a cleaner path to public markets, as Forbes reported.
OpenAI's response was to move first. CEO Sam Altman had been pushing for a September 2026 listing. CFO Sarah Friar had argued privately for 2027, cautioning colleagues that the company was not yet ready to meet the rigorous reporting standards public companies face, according to Investing.com. The board sided with Altman. The filing now sits days away — possibly by the end of this week, as CNBC and Reuters confirmed.
The numbers underneath
The financial picture the S-1 will force into the open is not one a CFO presents with confidence. OpenAI publicly disclosed annualized revenue exceeding $20 billion for 2025, with spending running at approximately $22 billion, resulting in a net loss of roughly $9 billion, according to Investing.com. Internal projections show the burn rate holding at 57 percent through 2026 and 2027, with operating losses projected to reach three-quarters of revenue by 2028. HSBC analysts estimate the company may require more than $207 billion in additional capital through 2030 to fund its commitments, the same analysis found.
Those commitments are themselves the more startling figure. OpenAI has publicly announced infrastructure spending totaling $1.4 trillion over eight years — a number larger than its entire private-market valuation, and more than 100 times its current annual revenue, per Investing.com. Whether these are contractual obligations, estimated build-out costs, or power purchase agreements will be among the first questions analysts ask when the prospectus opens.
The company will say this is the cost of building the infrastructure that defines the next era of computing. It will not say who pays if it is wrong.
OpenAI also missed internal revenue and user growth targets, with Friar warning colleagues that sluggish revenue growth could limit the company's ability to honor its data center commitments, as Quartz reported.
What public investors actually get
Under the October 2025 renegotiation, Microsoft holds approximately 27 percent of OpenAI's equity and a 20 percent revenue-sharing entitlement running through 2032, per Investing.com. The OpenAI Foundation holds 26 percent plus warrants and special voting rights. Public investors, if the offering prices successfully, absorb the economic consequences of those arrangements without the structural power to challenge them.
This is not a hidden disclosure. It will appear in the S-1. But the implications have received almost no attention in the press coverage of the filing. A company that burns $14 billion a year, needs $207 billion more in capital, and has committed to $1.4 trillion in infrastructure spending is asking public markets to fund it — while two parties with no fiduciary obligation to public shareholders hold the governance levers.
The closest historical parallel is not a technology company at all. It is a structure sometimes found in defense contracting and old-media companies, where a founding family or foundation retains super-voting stock while the public owns the economics. Those arrangements are controversial at any valuation. At $852 billion, they are unprecedented.
The governance transfer no one is naming
The IPO is being described as a milestone: the world's most watched AI company taking its next step toward public markets. What it actually represents is a governance transfer with no precedent in the technology industry.
Before listing, OpenAI's most consequential decisions — what to publish, what to withhold, what capabilities to develop and which to limit — were made by a board whose composition and deliberations were not subject to public disclosure. The safety council, the nonprofit structure, the special obligations embedded in the founding charter: none of it was accountable to anyone outside the room.
After listing, those decisions will be subject to a different kind of pressure. A board that must answer to quarterly earnings can no longer weigh safety considerations without modeling their cost. A company that must disclose material developments to the SEC cannot maintain the opacity that frontier AI labs have treated as a competitive advantage. A shareholder base that bought at a $852 billion valuation will apply pressure in proportion to what it paid.
Nobody has quantified what this transfer does to the safety research agenda, the publication cadence, or the internal balance of power between product and mission. That is precisely why it matters.
The Musk lawsuit, dismissed on May 18 on statute-of-limitations grounds, was the last existential legal overhang. The filing is no longer blocked. Goldman Sachs and Morgan Stanley are ready, as WSJ, Reuters, and CNBC all confirmed.
The question left for public markets is not whether OpenAI is worth $852 billion. It is whether the people who will actually run it — the nonprofit foundation, the Microsoft revenue arrangement, the board that answers to neither shareholders nor regulators — will be answerable to anyone at all.