Inside the AI IPO Wave That Could Redefine Wall Street’s Relationship With Technology Forever

The fact that the three most anticipated initial public offerings (IPOs) in a generation—SpaceX, OpenAI, and Anthropic—are getting closer to the public markets at the same time that Wall Street has started to discreetly, methodically, and somewhat sheepishly retract the most extravagant claims of the AI investment supercycle seems almost perfectly timed. Not a Lehman moment.

No collapse of the front page. Peter Oppenheimer of Goldman Sachs recently called this period of slow, controlled repricing “one of the worst periods of relative underperformance in the technology sector compared to the global market since the early 1970s.” In a way, the AI trade is over. In another sense, the AI opportunity might be just getting started. And which of those two statements is true will soon be put to the test by three companies whose combined private valuations are well over $700 billion.

The question of timing is crucial. These three initial public offerings (IPOs) have the potential to raise as much money as all venture-backed IPOs in the United States over the last ten years, according to research firm PitchBook. That is a structural observation about what happens to the larger IPO market when three entities of this size appear in close succession; it is not an exaggeration for effect.

Companies like OpenAI, which has a private valuation of about $300 billion, and SpaceX, which has already filed a confidential S-1, will compete with smaller businesses that might have gone public for a limited pool of institutional appetite. From the offices along Park Avenue to Sand Hill Road in Menlo Park, there is a feeling in venture capital circles that these three listings have the potential to effectively push out a whole generation of smaller technology listings simply by virtue of their existence.

Careful reading of the larger market context is warranted. Eighteen months ago, when the Magnificent Seven were the clear leaders of all major equity benchmarks, the forward P/E of the IT sector would have seemed genuinely unthinkable. Today, it sits below consumer discretionary, consumer staples, and industrials. The three-month realized correlation between the major AI hyperscalers has drastically decreased, indicating that the monolith has broken, according to Goldman’s analyst team. Oracle, Microsoft, Google, Amazon, and Meta are no longer following one another. In the second stage of a technology cycle, which comes after the initial surge of enthusiasm and before the real winners separate from the pretenders, the market has begun to demand differentiation.

Michael Wilson of Morgan Stanley contends that most of the correction has already been made. The forward P/E multiple of the S&P 500 has dropped 18% from its six-month high, a level seldom observed outside of recessions or aggressive Fed tightening, neither of which is the current baseline expectation.

He points out that the Magnificent Seven now carry more than three times the earnings growth of that defensive sector while trading at about 24 times forward earnings, which is almost the same multiple as consumer staples. The numbers appear differently when viewed from a relative value perspective than they did when everything was rising simultaneously. As all of this is happening, it seems like the market is in a challenging but potentially intriguing position: it’s not yet confident enough to feel cheap, but it’s repriced enough to feel less risky.

The most eagerly awaited listings in years enter this setting. It’s not easy to value OpenAI. Subscriptions, enterprise contracts, and API usage bring in a sizable amount of money for the company. However, its cost structure, which is dominated by compute, talent, and ongoing safety research, is very high, and its relationship with Microsoft, which has invested about $13 billion, raises structural concerns about the company’s actual independence.

Anthropic, which is supported by Google and Amazon and is renowned for its cautious approach to AI safety, raises various concerns regarding the rate of monetization. In terms of distinguishable revenue streams from Starlink and launch services, SpaceX is arguably the easiest of the three, but it carries Elon Musk’s unique brand of concentrated founder risk, which institutional investors have learned to price in multiple ways at once.

The historical similarities are instructive but not comforting. The more revolutionary a technology is, the bigger the accompanying bubble, according to Jeremy Grantham of GMO. The pattern has always followed the same arc after the railways, radio, and early internet: boom, overinvestment, collapse, and then a gradual maturity in which the technology eventually surpasses the wildest dreams of the original speculators. Before radio turned out to be the revolutionary medium that everyone had predicted, early investors in RCA who purchased at the company’s peak in 1929 saw the stock fall more than 97%. Those who purchased NVIDIA at low prices in 2015 and held onto it during the AI cycle saw remarkable profits. The price paid and the timing of entry were what separated those two results rather than the underlying technology’s quality.

It’s still unclear if the market is still recovering from the repricing or if it has already absorbed the worst of it when the upcoming wave of AI IPOs arrives. According to Goldman’s analysis of AI infrastructure investment, approximately 40% of all S&P 500 earnings growth this year can be attributed to AI capital expenditure. This indicates that the AI wager is essential to the health of the market rather than incidental.

The companies that supply the models at the top of the stack and the infrastructure could increase in value if that spending keeps up. The IPO window may close sooner than anyone anticipates if it doesn’t, or if return on invested capital turns out to be as elusive as it has been for the hyperscalers thus far. It’s a real moment. There is a genuine opportunity. The danger is also genuine.