Record-breaking issuance of new stock. Skyrocketing prices of semiconductor shares. Circular financing fuelling the boom. Established industry sectors reeling from the threat of new competition. And, hovering over it all, a lurking fear that a digital advance will mean the end of human civilization.
These are familiar headlines today from the artificial-intelligence supercycle, but all of this is straight out of 1999. WorldCom, Global Crossing, Lucent and Nortel, the hyperscalers of the day, stoked the entire internet ecosystem with investments and vendor financing of their products. Intel Corp. was the key player in producing chips, hitting a split-adjusted high of US$73 – a mark not to be exceeded for 26 years, until just this past May. The burgeoning internet took a wrecking ball to traditional retailers, as the market assumed bricks and mortar was dead. All that froth was mixed with a desperate fear that the Y2K bug would drop planes from the sky, disable power plants and blow up society. Remember the preppers?
What are we to make of the way the AI revolution is rhyming with that period of history? Aside from these outward similarities, we also observe aspects of behavioural market dynamics that represent the most interesting part of this recurring theme.
In the late 1990s, the bedrock assumption was that the demand for bandwidth was essentially infinite. Today, the core conjecture is that we can never have enough data centres and tokens. Once we fill the planet with them, apparently we will then have to build them in space, then on the moon. SpaceX is the quintessential company of our time, existing at the confluence of AI and orbital technology.
If you find this notion of infinite demand hard to swallow, you are not alone. But for the time being, the evidence of rapid growth is undeniable, requiring observers to perform some impressive mental gymnastics to reconcile what is happening.
Editorial: The shocking and predictable threat posed by AI
Somebody has to pay for all this digital plumbing. Consumers can pony up some, but they still have regular life to fund. In the 1990s, the favourite “victim” was traditional retail, which was to be devoured to pay for infrastructure. Today, it is the SaaSpocalypse: established software companies allegedly superseded by startups using AI agents. (The term borrows the initials for “software as a service.”) Their fat margins make them a particularly alluring target, though as yet, evidence of revenue implosion remains scant.
In the 1990s, we heard much agonizing over bottlenecks that ostensibly slowed the growth of the internet, much like today’s concerns over energy to power data centres. Many investors find it far more reassuring to worry about whether there are enough pumps available to cool server farms than to question the assumption of infinite demand. It is more comforting to believe we are limited by resources than by utility.
The thought of Tech Wreck 2.0 is so uncomfortable that investors would rather fret about AI reaching the point of singularity, where it grows so powerful, it renders humans superfluous. In an odd way, the looming Y2K apocalypse distracted from a more prosaic question: What if we didn’t actually need all that fibre? Today, the prospect of AI enslavement acts as a decoy from intractable commercial issues, such as calculating the return on massive compute investments. It is far better to worry about the existential consequences of ultimate success, such as mass unemployment, than the mundane possibility of failure.
So, as contrarian investors, do we fret about the bloom coming off the AI rose? Not a bit. The second-best year in Contra history was 2001, with a return of 64.8 per cent. We shot the lights out with stocks that had been deemed irrelevant by the internet revolution: coal, oil, pipelines and old-school retail. We also sidestepped the telecom crash with aerospace and medical technology.
One software stock we like today that is beat-up is Enghouse Systems Ltd. ENGH-T . It currently trades around $15.50, a fraction of its $78 high in 2020. Yet it sports a fortress balance sheet with a hefty cash reserve, no debt, a 7.3-per-cent dividend yield and a portfolio of sticky customers across transport, health care, municipal government and public transit.
Joël Blit: AI is disrupting education. Universities must change, and change fast
There are powerful reasons why hundreds of billions of lines of COBOL code written 60 years ago are still in active use – as impervious to new generations of computer-science tools as a geologic formation. For the majority of people, the experience of software change is on mobile phones, where apps can be swapped with a few taps. Enterprise software does not work that way. Changing core digital infrastructure is as complex as replumbing a nuclear plant: feasible, but a long, difficult and expensive process requiring meticulous execution.
Enghouse certainly faces some near-term challenges. In its contact centre division, the current market narrative assumes tech support will be replaced by bots, leading customers to delay commitments. While the company has integrated AI into its product platforms, the compute bill can be far higher than traditional licence seats, driving client friction. Furthermore, recent surveys indicate the majority of managers are not yet seeing tangible operational efficiencies or revenue growth to match those inflated bills.
Simultaneously, many of the firm’s competitors are in rough fiscal shape, leading them to slash prices in a desperate attempt to retain customers – which is squeezing margins across the entire sector. In a recent conference call, Enghouse chief executive Stephen Sadler issued a stark warning to analysts: watch cash flow, not EBITDA (earnings before interest, taxes, depreciation and amortization). He observed that industry players are capitalizing AI costs instead of expensing them, pushing that spending onto the balance sheet rather than the income statement. This is the ghost of WorldCom’s 1990s accounting tricks in action.
The view here is that Enghouse will be buffeted, but will not buckle, when the AI bubble bursts. Conservatively and patiently managed, without high levels of stock-based compensation that distort financial statements, it has the balance-sheet strength to emerge from the rubble of Tech Wreck 2.0 without diluting shareholders. While the herd chases data centres into space, superior returns will ultimately belong to the disciplined survivors who refused to bet the farm on a narrative.
Ben Stadelmann is vice-president emeritus of the Contra the Heard newsletter.