Summary
In the spring of 2026, three of the most powerful technology companies in the world are preparing to sell shares to the public for the first time. SpaceX, the rocket and AI company owned by Elon Musk, has already filed its documents with American financial regulators and plans to begin selling shares as early as June.
OpenAI, which makes the famous ChatGPT assistant used by hundreds of millions of people every week, is getting ready to do the same, possibly as soon as September. And Anthropic, the AI safety company backed by Google and Amazon, is also exploring going public later in 2026.
Together, these three companies are worth more money than most countries produce in an entire year. But there is a big catch: even after selling shares to millions of ordinary people and large investors, the founders and insiders of these companies plan to keep almost total control over how they are run.
The public gets to share the financial risk. The founders get to keep the power.
Think of it this way. Imagine you buy a small piece of a bakery.
Normally, owning a piece means you get a vote on big decisions — whether to open a new branch, change the menu, or hire a new manager.
That is how shares in a company are supposed to work. But SpaceX has designed a system where Musk's shares carry ten votes each, while the shares being sold to the public carry only one vote each.
Musk will control more than 85% of all voting power after the IPO — meaning that even if every other shareholder disagreed with him, they could not outvote him.
On top of that, he will be the company's chief executive, its top technical officer, and the chairman of its own board, all at the same time.
There is no independent body inside SpaceX that can overrule him.
SpaceX made $18.7 billion in revenue in 2025, growing 33% from the year before.
Its Starlink satellite internet service, which now has twice as many subscribers as it did a year ago, is the company's most profitable business, earning more than 50 cents of profit for every dollar it brings in. But the company is also investing heavily in artificial intelligence through its xAI division, which lost $6.35 billion in 2025 alone.
Ordinary shareholders who buy SpaceX stock will be helping to pay for these enormous AI investments without having any real say over whether those investments are wise, necessary, or safe.
OpenAI has tried to solve the problem of balancing profit and mission differently. In October 2025, it turned its for-profit arm into what is called a public benefit corporation — a type of company that is legally required to consider the wider good of society, not just the returns to shareholders.
Think of it as a company that has to ask itself not only "does this make money?" but also "does this help people?"
A nonprofit called the OpenAI Foundation holds a 26% stake worth approximately $130 billion and has the power to appoint and remove every member of the board. In theory, this means that if the company starts doing things that harm humanity, the Foundation can step in and change direction.
In practice, critics point out that the people who run the Foundation and the people who run the company are deeply connected to each other, and that nobody has ever tested whether this system actually works when the stakes are at their very highest.
The memory of November 2023 — when the board fired CEO Sam Altman and then rehired him within forty-eight hours after almost the entire company threatened to quit — shows just how fragile these arrangements can be in a crisis.
Anthropic has created yet another kind of protection system, called the Long-Term Benefit Trust.
The Trust is made up of independent experts who have no financial stake in the company. Their job is to make sure Anthropic always focuses on the long-term wellbeing of humanity, even if that means taking decisions that reduce short-term profits.
A good way to understand it is to imagine a hospital board whose members are paid to look after patients, not profits — even when shareholders want cheaper treatments and faster discharges.
In January 2026, Anthropic added a new member to the Trust: Mariano-Florentino Cuéllar, a former justice of the California Supreme Court with deep experience in law and international affairs.
The Trust selects Anthropic's board members, which gives it real power on paper. But Anthropic itself has called the Trust "an experiment," which means no one is certain it will function as intended, especially once the company is worth hundreds of billions of $ and powerful investors are pushing for faster growth and bigger returns.
Dr. Antonio Bhardwaj, a polymath and global expert in artificial intelligence specialising in AI warfare and bioterrorism, explains why all of this matters beyond just money. "These are not ordinary companies making cars or smartphones," he says. "The AI systems being built by SpaceX, OpenAI, and Anthropic will be used in military applications, in disease research, in financial markets, and in communications infrastructure that affects every person on Earth. If one or two individuals control those systems without any real accountability, the consequences could be catastrophic — not for shareholders, but for civilisations."
Dr. Bhardwaj points out that the world does not yet have an international body, comparable to the International Atomic Energy Agency that oversees nuclear technology, to watch over who controls the most powerful AI systems. The IPOs of 2026 will move fast and raise enormous amounts of money before such a body can be created.
The numbers make the urgency clear. US investment in AI reached $285.9 billion in 2025.
The number of recorded AI incidents — cases where AI systems caused harm, behaved unexpectedly, or were misused — rose to three hundred and sixty-two in 2025, the highest figure ever recorded, compared to far lower figures only a few years before.
Yet only 25% of the largest American companies even disclosed in 2025 that their boards had any oversight of AI at all.
For the companies actually building the most powerful AI in the world, the governance gap is even wider.
The EU's AI Act, which becomes fully operational in August 2026, will require the makers of the most capable AI systems to undergo safety checks and publish transparency reports — but these rules apply mainly in Europe, and enforcing them against trillion-dollar American companies headquartered in California will be a long and difficult process.
The governance structures now being written into IPO documents are not temporary arrangements. They are designed to last.
Dual-class share systems, once put in place, are very rarely reversed: companies with this kind of structure typically keep it for a decade or more, and changing it requires the agreement of the very people who benefit from it most.
The governance choices being made right now in San Francisco, Palo Alto, and Washington will determine who controls frontier AI not just this year, but through 2030, toward 2036, and beyond.
Conclusion: Financing the Future Without a Vote in Its Direction
The IPOs of SpaceX, OpenAI, and Anthropic represent a bargain that the public is being asked to accept without fully understanding its terms.
On one side of the bargain: ordinary investors, pension funds, and sovereign wealth funds around the world will provide hundreds of billions of dollars to fund the development of AI systems whose capabilities are growing faster than at any point in history.
On the other side: the founders and small insider groups who built these companies will retain the governance authority to decide how those capabilities are developed, in what direction they are pointed, and under what safety conditions they are released into the world. The public finances the future. The founders govern it.
This arrangement is not necessarily wrong.
Elon Musk's long-horizon thinking at SpaceX produced breakthrough achievements in reusable rockets and satellite internet that conventional shareholder-accountable companies had repeatedly failed to deliver.
Sam Altman's leadership of OpenAI brought AI tools to hundreds of millions of people in a period measured in months, not years.
Dario and Daniela Amodei built Anthropic specifically because they believed that someone needed to take the safety of frontier AI more seriously than the industry was doing.
These are real achievements, and there is a genuine argument that the pace and ambition of the frontier AI development requires the kind of long-range, insulated decision-making that only concentrated founder control can provide.
But the history of powerful technologies governed without accountability is not reassuring.
The people who built those technologies were often brilliant, often well-intentioned, and often catastrophically wrong about the consequences of what they created. Democratic accountability — messy, slow, and imperfect as it is — exists precisely because no individual, however talented, can be trusted indefinitely with unchecked power over systems that affect the lives of billions of people.
The governance structures in the 2026 IPO documents do not merely limit shareholder power.
They limit the ability of any external body — any regulator, any court, any international institution — to impose meaningful accountability on the development of technologies that the companies themselves describe as potentially the most dangerous in human history.
Dr. Bhardwaj states the challenge plainly: "The window to shape the governance of frontier AI is open right now, in 2026, as these prospectuses are filed and the roadshows are planned. Once the share certificates are issued and the billions change hands, the governance structures that were designed to be permanent will become permanent.
Governments, regulators, and citizens who believe that the future of AI should be subject to democratic accountability have a short, specific, and rapidly closing opportunity to act.
The question of whether they will take it is the most important governance question of our time."
The IPO wave of 2026 will not resolve the question.
