Everyone expected SpaceX to come under pressure.
On Aug. 6, about 911.5 million insider shares became eligible for sale following the company’s first lock-up expiration. Traditionally, that kind of event creates meaningful selling pressure as early investors gain the ability to monetize their holdings.
Instead, the opposite happened.
After touching roughly US$108 on Aug. 5, SpaceX rallied to nearly US$138 over the following three trading sessions on exceptionally strong trading volume. Every initial public offering (IPO) is different, and it’s too early to draw sweeping conclusions from a single example. But the market’s reaction raises an important question: has the AI trade entered a new phase?
The first phase of the AI trade was driven by excitement around the technology itself. Increasingly, investors also need to understand how public markets value these companies once they begin trading.
We’ve become accustomed to comparing new IPOs with companies we already understand. Banks are compared with banks. Software companies with software companies. Semiconductor manufacturers with semiconductor manufacturers.
The next generation of AI companies won’t fit neatly into any of those categories.
OpenAI isn’t simply a software business. It’s a research organization, a platform, a computing company and, increasingly, an infrastructure company. Anthropic faces many of the same dynamics. Their products improve at extraordinary speed, but so do their competitors’. Their addressable markets continue to expand, yet so do the capital requirements needed to compete.
That makes traditional valuation metrics much less useful on their own.
Price-to-sales multiples can only tell part of the story when billions of dollars are being invested in computing infrastructure before revenue has fully matured. Discounted cash flow models become highly sensitive when no one can confidently predict what AI pricing, customer behaviour or competitive dynamics will look like five years from now.
That doesn’t mean these companies are overvalued. It means they’re unusually difficult to value.
The recent trading in SpaceX reinforced this point. Once a company enters the public markets, the discussion quickly shifts from what the business could become to how investors value it relative to expectations, peers, institutional ownership and its potential role in long-term portfolios. Those questions — not just the opening-day price — often determine how the story unfolds.
The difference is that AI companies face another challenge that SpaceX largely avoided.
A computing arms race
Every query has a cost. Every improvement in model performance requires more computing power. Every new generation of model demands more advanced chips and more electricity than the generation before it.
That changes the economics entirely.
For years, software investors became accustomed to businesses whose margins improved as they scaled. AI may not follow that path as neatly. The race to build larger, more capable models has become an arms race in computing infrastructure.
Which brings us to the most overlooked investment story in AI.
People often describe AI as a software revolution. In reality, it’s just as much a chips-and-energy revolution. Every advance in AI requires exponentially more computing power, and every increase in computing power requires reliable, affordable electricity. Those aren’t side stories to the AI trade. They are becoming some of its primary drivers.
Without advanced semiconductors, AI doesn’t exist. Without abundant, reliable and affordable electricity, those chips don’t run.
That’s one reason I’m increasingly interested in the businesses enabling AI rather than focusing exclusively on those building the models.
It also creates opportunities much closer to home.
North America — and Canada in particular — is unusually well positioned for this phase of the AI trade. Canada combines abundant energy resources, growing investment in data centres, access to critical minerals and a long history of AI research.
Silicon Valley may produce many of the headline companies, but the infrastructure supporting AI will increasingly be built across the continent.
For advisors, that broadens the conversation considerably. Clients will inevitably ask whether they should buy the next OpenAI IPO. A better question is whether they’re getting enough exposure to the broader AI ecosystem.
That includes semiconductor companies, networking businesses, energy producers, utilities, data centre operators, enterprise software providers and the infrastructure required to support exponential growth in computing demand.
History suggests those ecosystems often create as much value as the platforms themselves.
The internet produced Amazon, but it also produced Cisco, Equinix and thousands of companies that built the digital economy.
AI is unlikely to be any different.
That’s why the next AI IPO cycle isn’t simply about identifying tomorrow’s biggest public company. It’s about understanding how the AI trade itself is evolving: from software to infrastructure, from possibility to profitability and from private-market expectations to public-market discipline.
Successful investing has never been about chasing the most exciting story. It’s about recognizing when the market begins asking different questions. That’s exactly where the AI trade is today.
Jay Bala is founder, CEO and senior portfolio manager at AIP Asset Management.