
The AI build out: The tortoise and the hare
By Avery Shenfeld, Chief Economist, CIBC Capital Markets
In the classic children’s tale, the slow but steady tortoise ends up winning the race against a quick-footed but overconfident hare. For various reasons, the moral of that story, that a faster start is not always better, might hold lessons for those involved in the race to build out the infrastructure needed to bring artificial intelligence (AI) to full fruition.
Markets have been making an argument against both extremes. Equity sell-offs of shares in the AI space have been triggered in some cases by fears that the rapid run-up in capital spending, and the related issuance of debt and equity, would overly dilute earnings for established firms, keep newer players in the red, and even imperil debt servicing. The hare could run out of financial market fuel before the finish line. There’s a related concern that in the race to be the first, the aggregate level of spending by all the players could exceed the returns available.
Some in the industry have advocated for regulatory restrictions that would “pace” the rollout of new models as a safety measure, lest we unleash dangerous agents on the world. Investor returns could end up benefiting if slowing down the hare when it comes to new product launches reduces the financial risks associated with those that do more harm than good.
True, at other times, the concerns have appeared when the revenue numbers aren’t growing fast enough. Suppliers of chips and computing power have been punished when the growth rate for their sales to AI players doesn’t measure up to ultra-lofty expectations. The post-election sell-off in the space was tied to chatter that Democratic victories, including in state capitals, could heighten a not-in-my-backyard resistance that would slow approvals for data centers and their power sources.
Such a slowdown could take some of the wind out of U.S. economic growth. Data from the first three quarters of the year suggested that even after netting out imports, AI capital spending could account for nearly a third of the economy’s gains this year. Remember, however, that the Fed is now tapping the brakes to cool inflationary pressures, so a more moderate pace of AI capex could just substitute for a slowing we would see with a more extensive run of rate hikes. AI spending is fueling some of the inflation the Fed is taking aim at, by bidding up market prices for memory chips, trucking rates and, in some areas, electricity prices.
Moreover, AI debt issuance, along with government deficits, has bid up bond yields and thereby crowded out other capital spending in the economy. Aside from data centers and power plants, real construction spending is down year to date in virtually all categories. So a pared-back economic contribution from AI capital spending could be partially offset by a reduced drag on housing and other business investment spending.
Some argue that America can’t afford to fall behind China in the race to be first out with the best models. But even in the U.S., when one AI lab comes out with a version that is touted as the best and scoops up market share, that advantage seems to last only a few months. Large language models don’t seem to be the kind of technology that can’t essentially be mimicked or surpassed by an up-and-comer.
In the tech world, once a sector reaches a mature phase, you often see entrenched players that gain a foothold that is difficult to overcome for later entrants. Witness the long-lived dominance of firms like Microsoft, Google and, to a lesser extent, Apple and Samsung in cell phones. But those at the top of the heap in the early stages of these technologies did not gain a sustained edge, as Excel surpassed Lotus 123 and VisiCalc in spreadsheets, cell phone leaders like Blackberry and Nokia faded out, as did search tools like Yahoo and Altavista. If that’s the way AI plays out, even if the U.S. builds out its AI platforms at a more measured pace than China, there would be time to catch up.
True, the overall impact of recent concerns over how quickly AI data centers and power plants can be delivered has been to take the steam out of equity markets. In the here and now, investors seem to prefer the hare over the tortoise. But even in that fable, the hare did get off to a faster start, and held the lead. How AI contributes to investor returns, and to the broader economy, won’t be visible until we get closer to the finish line. For the reasons outlined above, stakeholders across the economy might end up being better off with a slower paced approach, even if that’s an inadvertent result of a political swing that holds back the industry in the near term.

Related insights and perspectives
Economic and Financial Markets Outlook: Fourth Quarter 2026



