The AI trade still has plenty of buyers, but the next test is whether those buyers can absorb a rush of giant share sales without losing discipline on price.
Wall Street is not short of enthusiasm for artificial intelligence. It is short of a clean answer to a harder question: how many mega deals can the same pool of institutional capital swallow before demand starts to thin?
That question has moved from theory to calendar. SpaceX is marketing what would be the largest IPO on record, seeking $75 billion at $135 a share, with Reuters reporting investor demand of about $150 billion. Quantinuum has just closed an upsized Nasdaq IPO, selling 28 million shares at $60 each and raising $1.68 billion after demand reportedly ran more than 20 times the available stock. Anthropic confidentially filed for an IPO on June 1. OpenAI is widely expected to follow.
On the surface, this looks like a market begging for more supply. The better reading is more cautious. One deal being oversubscribed tells you buyers want access. It does not tell you that buyers can absorb every AI-linked private company, secondary sale and pre-IPO block at the valuations now circulating around the market.
According to Bloomberg, the wave of AI-related equity deals has raised a pointed concern among capital-markets bankers and investors: there may be more shares coming than natural long-term buyers willing to own them at current prices. That is not the same as saying the AI market is broken. It means the market is entering the stage where price discovery matters again.
SpaceX is the cleanest stress test because of its size. A $75 billion raise is larger than full-year IPO proceeds in many recent years, and it arrives before the market has even dealt with potential offerings from Anthropic, OpenAI, Databricks and other companies selling themselves as core infrastructure for the AI cycle. Even if the SpaceX book is covered, the order book is not infinite. Institutions still have risk limits, sector limits and portfolio managers who need to explain why they bought a company at a valuation measured in trillions.
Interest rates make that discipline more important. When money was nearly free, investors could justify paying far ahead for future dominance. With rates still elevated, future cash flows are worth less today, and companies burning heavily to build compute capacity face tougher questions. AI companies are asking public investors to fund enormous data-center, chip and power commitments before the returns are fully visible.
Quantinuum shows both sides of the market. The company priced above its earlier range and raised more than expected, a clear sign that frontier technology can still command serious demand. But its 2025 revenue was only $30.9 million, while its net loss was $192.6 million, according to its IPO disclosures. Investors are not buying current earnings. They are buying a claim on a future market that may take years to mature.
Circular deals complicate the signal
The more complicated issue is that AI valuations are not being set in a clean market. Some of the biggest numbers are tied to circular arrangements, where one company invests in another company that then uses the money to buy cloud services, chips or infrastructure from the investor or its partners.
Microsoft has invested more than $13 billion in OpenAI over the years. Amazon has committed billions to Anthropic, while Anthropic uses Amazon Web Services and has also worked with Google Cloud. Nvidia has said it could invest up to $100 billion in OpenAI as OpenAI plans massive data-center purchases using Nvidia chips. These relationships may be commercially rational. Training and running frontier models is expensive, and access to chips and cloud capacity is a strategic asset.
But circularity changes what investors should read into the headline valuation. If capital comes bundled with future spending commitments, the valuation is not a pure market vote in the way a broad public offering is. It can make revenue growth look stronger, support higher private marks and give founders across the ecosystem a tempting benchmark for their own fundraising conversations.
That matters for startups that are nowhere near OpenAI or Anthropic in scale. A founder raising a Series B may point to frontier-model multiples and argue that the whole AI market deserves a premium. Investors should push back. The premium belongs to companies with distribution, proprietary data, compute access and a credible path to durable revenue. It does not automatically transfer to every company with AI in the pitch deck.
The next few months will show how much of this demand is durable and how much is scarcity. If SpaceX trades well after its debut and Anthropic can price cleanly, the IPO window opens wider. If the first giant deals absorb too much capital or stumble after listing, private valuations will have to reset before many founders are ready to admit it.
AI is still attracting money because the opportunity is real. The risk is that the financing machinery has started to move faster than the evidence. Investors should watch not only which deals are oversubscribed, but who is buying, how long they hold and whether revenue growth is coming from customers or from capital moving in a circle.
Also read: Trump’s AI memo redraws the defense market • Britain is becoming the first buyer for homegrown AI chips • Microsoft is turning enterprise AI agents into platform plumbing
