The timeworn mantra that “this time it’s different” does not provide comfort to those who fear that investors are getting over their skis and setting up another boom-bust cycle in the stock market that bleeds into the overall economy. While those fears seem premature at this point, there are good reasons to remain watchful for a frenzied market’s impacts on states’ and municipalities’ revenues, retirement plans and economies.
The current boom is not yet a bubble: While AI stock valuations have enjoyed a nice run-up, most of the market’s gains have been driven by large companies already making profits. The goliath AI industry leaders are similar to the high-flying Nifty 50 stocks of the 1960s; they don’t much resemble the upstart vaporware companies that were wiped out when the dot-com bubble burst. Many speculators, especially younger ones, are fixated on other nontraditional investment venues — think cryptocurrency, sports gambling and now the prediction markets — where they seek to make a quick buck by playing hunches. But widespread speculative fever and telltale hubris have not infected the stock market at this point.
As long as AI technology continues to advance toward pragmatic applications that can yield real economic benefits, the risk of a replay of the 2000 stock market collapse is less fearsome. And if the advent of artificial “superintelligence” is pushed back from the near future to the 2030s, as more-cautious technologists now assert, that probably elongates the timeline for AI bubble formation by a few years — especially if market forces deliver some occasionally severe “cleansing corrections” to drive out the weaker hands without impairing deep-pocket long-term investors.
Nonetheless, the recent circular corporate financing that intertwines the fortunes of the AI industry goliaths feels eerily similar to the Japanese keiretsu buildup of cross-invested companies before those stocks crashed in 1990, which begot the “Lost Decade” in that country’s economy. The risk-fueling prerequisites for a worrisome market bubble are already in place in the U.S.: mounting laissez-faire deregulation of markets and the financial industry; the “gamification” of trading platforms; billion-dollar private equity unicorns and special purpose acquisition companies awaiting big IPO payoffs; meme stocks artificially boosted by social media hype; and now the completely unregulated prediction markets, which could easily become today’s equivalent of 1929’s infamous bucket shops that wiped out so many gullible small investors. (For a succinct 70-page summary of what history should teach us about all this, get yourself a copy of John Kenneth Galbraith’s A Short History of Financial Euphoria from 1994.)
It’s therefore instructive to remind ourselves of how much damage the dot-com bubble caused 25 years ago. The high-flyer stocks trading on the NASDAQ surged 86 percent in 1999 alone and then collapsed by 77 percent. Speculators were wiped out, never-profitable companies went bust and a recession ensued. Notably, the public pension funds that had deemed themselves overfunded in 1999 — and granted their current workers retroactive benefit calculation enhancements — quickly discovered that their paper portfolio profits had evaporated, setting the stage for a decade of “pension tsunami” angst and two decades of endlessly increasing employer payroll costs.
Today the average public pension fund is still less than 80 percent funded actuarially, with about 45 percent of their portfolios invested in equities, so it would take a further dot-com-sized surge of stock prices to get them back to the point of hubris where politicians would think they can raid that cookie jar to provide new benefits. That’s not a realistic risk at this point, but if these are ultimately the “Roaring 2020s,” then circa 2029 would bode ill.
In the aftermath of the dot-com collapse, state income and sales tax revenues declined by $38 billion, about 13 percent in a single year, even as consumers continue to spend. State personal income tax revenues dropped the most, by 22 percent. Taking everything into account, state budget shortfalls tallied up to $150 billion over the two fiscal years after the bubble burst, reflecting not just the revenue losses but many states’ prior budgetary dependence on bubble-based economics, including foolish “pension contribution holidays.”
Therein lies the most important lessons from this historical chronicle: Any state revenue surpluses derived from today’s booming stock markets should be socked away for the rainy days that eventually follow, and pension funds must not increase benefits or shortchange funding on the basis of paper profits. Local governments that depend on state-shared revenue must build these sober facts of life into their own multiyear financial plans and models. Those models should include an “alternate boom-bust” scenario that mirrors what happened to their own budgets and operations between 1999 and 2003.
Putting aside the revenue crunch from any potential AI bubble burst, there are two other admonitions that state and local policymakers and senior managers should keep in mind as this new technology takes root in their backyards and their daily operations. One involves land use and the other is a procurement precaution.
The land-use issue centers on data centers. If the demand for AI applications skyrockets as investors expect, it’s conceivable that there will be overbuilding amid surging power consumption. Eventually, just as the freewheeling railroad industry overextended itself in the late 19th century, leading to its own bankruptcy crisis, an AI bubble could easily result in excess spatial capacity and abandoned projects — especially if evolving technology transforms the computing process to require smaller equipment and less space. Municipalities that expect economic benefits from these mega-complexes could end up hosting white elephants and abandoned operations. Local officials need to incorporate that possible scenario into their approval process.
A final practical point, for procurement officials: It’s too early for this to be an immediate concern, but if the AI bonanza does become a bubble, it’s likely that some of the AI applications sold to public agencies will be supplied by undercapitalized companies that collapse in a market washout. Whether it’s robotic devices or software-as-a-service contracts, purchasing officers should insist on suppliers’ liability insurance policy riders and faithful performance bonds to protect their agencies from vendor business failures or product malfunctions that make the user liable. If nothing else, that’s a reputational risk to avoid.
Overall, the potential risks to state and local financial operations from the AI boom still seem far enough away to become too excited about right now, but the potential consequences of a speculative bubble are severe enough to warrant a sober annual review of “what could possibly go wrong?”
EDITOR’S NOTE: After contributing some 200 biweekly public finance columns to Governing, and writing monthly for a half-decade before that for his “Benefits Beat” series, Girard Miller is retiring from this space. He remains a valued friend of our staff and a stalwart of the government finance community, and expects to craft occasional guest commentaries for our publications.
Governing’s opinion columns reflect the views of their authors and not necessarily those of Governing’s editors or management. Nothing herein should be construed as investment advice.
