AI’s forgotten booms hold real warnings for investors, says ACC CIO
The first AI boom began in the 1950s, according to ACC CIO David Iverson (Navy Medicine/Unsplash)
Most investors reaching for a historical analogue to the current artificial intelligence boom point to the wrong one, David Iverson, CIO of New Zealand’s NZ$50.8bn Accident Compensation Corporation, tells AOX. The railroads of the 1880s, the radio mania of the 1920s, and the dotcom bubble of the late 1990s are interesting examples of tech spending gone astray, he says, but the better comparison is AI itself.
“We’re on the fourth boom,” Iverson says. “We’ve had three prior AI booms. They are the best analogues.” Each, he says, ran into a reckoning, though not always the outright bust investors might expect.
The term AI was coined in the 1950s with the first, largely government-funded boom, which was built around symbolic reasoning and early neural networks. But that hit a wall when funding dried up in the mid-1970s, according to Iverson.
The second, the “expert systems” wave of the 1980s, promised AI for business and petered out by the early 1990s.
The third, the deep-learning and big-data surge of the 2010s, didn’t bust at all, but it was forced to change. On-premises big-data setups became notorious money pits, forcing the industry to migrate wholesale to the cloud to keep the economics viable.
Cautionary tales
Each boom, Iverson argues, has a lesson that many market participants are ignoring.
The first lesson concerns over-optimistic claims about what machines can do, he says. The 1950s boom promised thinking machines that would do everything humans could. “There’s an analogy right now with the idea of AI replacing humans,” Iverson says, adding that when such expectations outrun reality, disappointment follows.
The second relates to return on investment. Expert systems failed to deliver and cost too much. Today, Iverson notes, some companies are already finding AI more expensive than the human workers it replaced, and many companies still cannot clearly articulate how AI fits their processes and whether the ROI is there.
The third is that a boom can be entirely real and still ruin investors. Big data didn’t collapse; it became the foundation the current AI wave is built on, but the firms that overspent on the wrong infrastructure, and those backing them, paid for it anyway. “We’re going through a massive capex boom,” Iverson says. “I’m not saying we’re repeating all of these past mistakes, but we’re rhyming a lot with them.”
Iverson is careful to separate the technology from its financing. He rates AI as an excellent tool and notes that he even used it to research these past booms. He also sees it as a natural fit for coding, where a defined language plays to models’ strengths. But he is wary of any technology marketed as boundless. “That’s always a recipe for investor optimism that might not be justified,” he says.
The cracks are already showing, Iverson says, as LLMs seem to bump against their limits and capital continues to flood into the space. What worries him is not AI but the financing around it: circular deals, questionable GPU-depreciation accounting, Nvidia’s balance sheet underwriting much of the ecosystem, and acute concentration risks in the US market.
The danger, Iverson says, is when all of these things compound. “It’s not AI per se. It’s how this unfolds given what’s going on and the lessons we should have learned,” he says.