The latest wave of AI-related IPOs has a familiar look from a distance: big names, big promises, and a market eager to turn “breakthrough” into “breakthrough company.” But when you get closer, the story starts to feel less like a straightforward celebration of innovation and more like an exercise in financial engineering—one that can quietly change who bears the uncertainty.
That’s the core tension highlighted by recent coverage: the “math” behind valuations doesn’t always reconcile with the underlying economics of the business models being priced. And when that mismatch exists, the mechanism that resolves it is often not a dramatic collapse or a sudden revelation. It’s something subtler. It’s risk being redistributed—sometimes away from private investors and venture backers, and toward public-market participants, including retail investors.
To understand why this matters, it helps to separate three things that markets often blend together: growth expectations, cash-flow reality, and the structure of uncertainty. In AI-heavy sectors, all three are moving targets. The technology may be advancing quickly, but monetisation timelines, customer adoption curves, unit economics, regulatory constraints, and competitive dynamics can all shift faster than investors can update their models. When those variables are hard to forecast, valuation becomes less about what a company will earn and more about what investors are willing to pay for optionality—the possibility that the company becomes the platform, the infrastructure layer, or the dominant application.
Optionality is not inherently irrational. It’s how markets price many early-stage businesses. The problem arises when optionality is priced as if it were certainty, or when the distribution of outcomes is treated as though it’s evenly shared. In practice, the distribution of outcomes is rarely shared equally. It depends on who bought in at what price, what protections exist, how lockups work, and how the company’s capital structure evolves after listing.
In other words, the “impossible maths” isn’t necessarily that AI companies can’t succeed. It’s that the public-market valuation can imply a level of predictability that the business itself does not yet demonstrate. When that gap exists, the IPO can function as a transfer of risk rather than a clean conversion of private optimism into public confidence.
A useful way to think about this is to ask: what exactly changes when a company goes public?
First, the investor base changes. Private rounds often involve sophisticated investors—venture funds, strategic investors, and institutional players—who typically have different time horizons, different tolerance for volatility, and different ways of underwriting risk. They may also have access to information that retail investors don’t, whether through diligence processes, industry networks, or simply the fact that they are closer to management and product roadmaps.
Second, liquidity changes. Public shares can be traded daily, which means the market can reprice expectations quickly. That repricing can be rational—new information arrives—or it can be reflexive—sentiment shifts, macro conditions tighten, or the market decides that “AI” is crowded. Either way, the ability to exit changes the payoff profile for different investors. Those who can sell earlier may effectively reduce their exposure to downside scenarios.
Third, the company’s capital structure and incentives can shift. IPO proceeds can fund growth, but they can also reduce the need for future fundraising at potentially worse terms. That can be positive for the company. Yet it can also mean that some of the uncertainty that would otherwise have been borne by future private investors is now borne by public shareholders who buy at the IPO price or shortly thereafter.
This is where the “risk transfer” idea becomes concrete. If the company’s future is uncertain, then the question is: who is positioned to absorb the consequences of that uncertainty?
In a typical private-to-public transition, early investors often benefit from a combination of upside participation and downside mitigation. They may sell part of their holdings after lockups expire, they may have negotiated terms that protect them in certain scenarios, and they may have already captured gains through secondary sales. Meanwhile, retail investors—often drawn in by media attention, social amplification, and the narrative of “the next big thing”—may be buying at a moment when the company’s story is at its most compelling and the risks are at their least visible.
None of this requires fraud or deception. It’s a structural feature of how capital markets work. The IPO is not just a fundraising event; it’s also a reallocation of who holds the uncertainty.
So what does “impossible maths” refer to in practical terms?
It often shows up in the relationship between valuation and fundamentals. AI companies can be valued on revenue growth, gross margins, customer retention, and the strength of their distribution. But many AI businesses are still in the phase where costs scale in ways that are difficult to forecast. Compute costs, data acquisition, model training and inference expenses, hiring, and infrastructure spending can all behave differently depending on product design and usage patterns. A company might appear to have strong demand while still burning cash at a rate that makes long-term profitability uncertain.
At the same time, the market may be pricing these companies as if they will achieve a particular trajectory: rapid scaling, improving unit economics, and durable differentiation. If those assumptions are even slightly off, the valuation can compress quickly. That compression is not necessarily a sign that the company is failing. It can be a sign that the market’s expectations were too optimistic relative to the evidence available at the time.
The “math” becomes “impossible” when the implied probability-weighted outcomes don’t match the observable risk profile. For example, a valuation might assume that the company will capture a large share of a market that is still forming, while also assuming that competitors will not undercut pricing or replicate capabilities. It might assume that regulatory and compliance hurdles will be manageable without materially slowing deployment. It might assume that customers will adopt the product at a pace that supports the projected revenue curve. Each assumption may be plausible on its own. The issue is the joint assumption set—how many things must go right simultaneously for the valuation to be justified.
When markets price that joint set as if it’s likely, the downside becomes asymmetric. If the company performs “well but not perfectly,” the stock can still fall because the valuation was built for a higher bar. That’s not unique to AI, but AI amplifies it because the product cycles and monetisation pathways can be less linear than in traditional software.
Now consider what happens when the company lists.
The IPO price is often set through a process that balances demand from institutional investors, underwriting considerations, and the desire to create a successful debut. Underwriters want the stock to trade well initially, because a weak debut can harm credibility and complicate future fundraising. Companies want to raise enough capital and establish a valuation that supports employee retention and future financing flexibility.
Retail investors, meanwhile, may not be part of the book-building process in the same way, but they can become part of the outcome once the shares begin trading. They may buy after the IPO, chase momentum, or enter through brokerage platforms that make access easy. Their participation can be driven by genuine interest in the technology, but also by the perception that IPOs represent a kind of validation—that the market has already done the hard work of pricing risk.
That perception is where the risk transfer can feel counterintuitive. An IPO can look like a milestone of certainty, when in reality it can be a milestone of redistribution. The uncertainty doesn’t disappear; it moves.
One reason this is particularly relevant in AI is that the sector’s narrative is powerful enough to attract capital even when the underlying economics are still evolving. AI is not just a product category; it’s a general-purpose technology in the public imagination. That makes it easier for investors to believe that “eventually” the economics will work out. But “eventually” is not a financial metric. It’s a timeline. And timelines matter because they determine discount rates, cash burn, and competitive windows.
If the market’s discount rate rises—because interest rates move, risk appetite declines, or macro conditions tighten—then the present value of future profits falls. In that environment, even companies with credible progress can see their valuations reset. Retail investors, who may have less ability to hedge or diversify, can experience that reset as a sudden loss of wealth rather than a recalibration of expected value.
There’s also a behavioural dimension. Retail investors often interpret volatility as information: if the stock drops, maybe the thesis is wrong; if it rises, maybe the thesis is confirmed. But in public markets, price movements can reflect many factors beyond company performance. Liquidity, index inclusion expectations, options positioning, and broader sentiment toward “AI” can all drive short-term moves. When the IPO brings in a larger retail audience, the stock can become more sensitive to sentiment swings, which can further amplify the perceived risk.
This is not an argument against IPOs. It’s an argument about what IPOs do.
An IPO can democratize access to growth in a literal sense: retail investors can buy shares in companies that previously were accessible only to institutions. That’s a real benefit. It can also improve transparency, because public companies must report financials, disclose material risks, and face ongoing scrutiny. Over time, that transparency can help investors make better decisions.
But democratization of access is not the same as democratization of risk. If the IPO price embeds optimistic assumptions, then retail investors can end up paying for uncertainty that private investors have already partially offloaded. The company benefits from raising capital and establishing a public currency for acquisitions and compensation. Early private investors may benefit from liquidity and valuation uplift. Retail investors may benefit from upside if the company exceeds expectations. But if the company meets expectations that are merely “good,” the downside can still be meaningful because the valuation was built for “great.”
This is why the “transfer of investment risk” framing resonates. It suggests that the IPO is not just a bridge between private and public markets; it’s also a bridge between different risk-bearing capacities.
Another angle worth considering is how AI companies’ competitive advantages are often described. Many AI firms claim differentiation through proprietary data, model architecture, distribution partnerships, or