The artificial intelligence gold rush & bubbles past
Last week, The Wall Street Journal (WSJ) published a piece commenting on the artificial intelligence (AI) infrastructure frenzy. As an investor, one can’t help but be awestruck while also feeling concern and a more than a hint of déjà vu.
According to the WSJ, Ellendale, North Dakota, is a sleepy town of just 1,100 now playing host to a half-built AI data centre that’s will be larger than 10 Home Depots – the U.S. equivalent of a large format Bunnings Warehouse here in Australia.
With a price tag north of US$15 billion, that one data centre also represents a quarter of the North Dakota’s annual Gross Domestic Product (GDP).
We’ve commented before on whether the spend is justified by the revenue, but each news story outlining the dollar magnitude of the AI race forces one to pause again and reflect.
AI proponents are calling it the Fourth Industrial Revolution – the ongoing transformation of industries and societies driven by artificial intelligence, robotics, the Internet of Things (IoT), blockchain, quantum computing, and biotechnology, and building on the digital foundation of the Third Industrial Revolution, which was centred on computers and the internet.
Payback periods
Quite rightly, however, the article points out the big question: How on earth do investors plan on recouping these investments, and when?
The scale of what is the technological equivalent of an ‘on spec’ build in the office construction industry, is staggering.
Over the past three years, tech titans have poured commitments into data centres, chips, and energy that dwarf historical precedents. Microsoft’s Satya Nadella has said he hopes it doesn’t take 50 years like electricity’s slow adoption. Meta’s Mark Zuckerberg says they’re all investing as if it won’t – pointing to Meta’s potential US$600 billion U.S. spend through 2028.
But ultimately, this is a high-stakes gamble on AI evolving fast enough to reshape the economy and squeeze out profits.
The WSJ draws alarming parallels to the late ’90s dot-com mania, where telecoms blanketed the U.S. with over US$100 billion in fibre optics, only to crash spectacularly. Back then, giants like WorldCom and Global Crossing went belly-up amid massive overbuilding. Today, investors should be reminded that chips and data centres – now a near-trillion-dollar arena – were once boring.
The WSJ journal describes CoreWeave’s Ellendale project as a poster child for the gold rush. Just six years ago, it was a small crypto miner with fewer than two dozen staff members. Now, flush with Wall Street cash, it’s valued higher than General Motors or Target. Pivoting from crypto to AI cloud computing post-ChatGPT, and with a culture shouting “YOLO” (You Only Live Once) and “GSD,” (Get Stuff Done) the company has amassed over US$42 billion in contracts, including a beefed-up deal with OpenAI. But it’s all fueled by debt – US$15 billion in loans at rates starting above eight per cent, plus $US56 billion in long-term leases for data centres.
Somewhat frighteningly for anyone who understands the concept of ‘operating as a going concern’ CoreWeave’s tech deals are shorter-term (2-5 years). Matching short-term revenue contracts to long-term contingent liabilities leaves the company exposed if demand dries up. If overbuilding occurs or tenants bail or switch, CoreWeaves products and services become the dark fibre cables of the 2020s. A phoenix (AI version 2.0) may rise from the ashes, but one requires a fire to torch AI 1.0 first.
According to the report, Sequoia’s David Cahn estimates that 2023-2024 AI infrastructure investments alone will require US$800 billion in product sales over the chips’ 3-5 year lifespan for solid returns. Bain & Co. ups that to US$2 trillion in annual AI revenue by 2030 – more than the combined sales of Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia today, and five times the global subscription software market.
Where will all this money come from? Which customers have it to spare after they have also spent money with Amazon, Apple, Alphabet, Microsoft, and Meta?
Morgan Stanley estimated last year’s AI revenue at just US$45 billion, mostly from chatbot subscriptions and cloud access. While consumers love free AI tools, businesses are reportedly stingy beyond US$30/month per user for things like Microsoft’s Copilot. Investor Roger McNamee reckons this bubble is bigger than all prior tech bubbles combined. He says that even success won’t justify the current spend.
Same, same but different
Of course, as we’ve previously noted, this isn’t the dot-com bubble. Today’s hyperscalers (Alphabet, Microsoft, Amazon, Meta) generate cash, and much more than those ’90s fibre firms. Moreover, OpenAI’s ChatGPT boasts 700 million weekly users (up from 500 million in March), with revenue tripling in 2024 to US$13 billion – though that’s peanuts compared to the US$60 billion annual Oracle commitment and US$1 trillion+ data centre plans.
And here’s an interesting dichotomy. Many proponents note that if AI displaces white-collar jobs en masse, the savings could justify all the spend. But if all those white collar jobs are lost, who will have the money to buy AI-powered products and services? Henry Ford gave his factory workers a pay rise, because he realised they were his customers. If AI wipes out workers’ jobs, who will the customers be?
A web of debt and guarantors
Elsewhere, the financing web is intricate, with debt everywhere. The hyperscalers are reportedly eyeing US$400 billion in 2025 capex, which as an aside, is more than the Apollo program adjusted for inflation.
Some players are building their own centres; others lease from middlemen like CoreWeave, who then pack them with Nvidia chips. In Ellendale, CoreWeave leased from Applied Digital, another ex-crypto player turned AI hopeful.
History replete with new tech that torched investors
One reason for bubbles is that investors repeatedly make the mistake of believing technology that changes the course of human history must also be good for investors. From canals and railroads, to electricity, commercial air travel, the automobile and the television, history is replete with examples of technology that rewarded consumers more than investors.
The excitement of the technology’s potential encourages a race that results in overbuilding. That overbuild is then followed by losses. Ultimately the technology survives thanks to consumer demand, but not every investor in a provider of that tech makes money. Indeed, history reveals most investors lose.
The WSJ cites Andrew Odlyzko’s “collective hallucinations” in manias, where hype trumps risks – like the dot-com fibre rush, where traffic growth was overstated, leading to busts. Execs at firms like Level 3 spent wildly, only to see stocks tank 95 per cent and fibre idle for years until streaming came along and saved it.
Today, Applied Digital has broken ground on yet another tenant-less data centre.
Red flags
We recently chronicled an MIT report that revealed 95 per cent of firms generate zero return on investment (ROI) on their AI ventures. Elsewhere, a Chicago study found chatbots offer no earnings boost in Danish workplaces. Meanwhile, ChatGPT-5’s August release was middling rather than revolutionary, and that was despite escalating training costs of three to five times per model. Chips are also superseded and therefore depreciate fast, unlike enduring fibre.
Of course, a frog can’t feel the water temperature gently rising, so the AI race continues, fuelled by an apparently endless flow of money. As Ellendale’s Mayor Don Flaherty is quoted as saying, “we’re on the wave right now, and we’ve just got to keep riding it.”
Roger Montgomery is the Founder and Chairman of Montgomery Investment Management. Roger has over three decades of experience in funds management and related activities, including equities analysis, equity and derivatives strategy, trading and stockbroking. Prior to establishing Montgomery, Roger held positions at Ord Minnett Jardine Fleming, BT (Australia) Limited and Merrill Lynch.
He is also author of best-selling investment guide-book for the stock market, Value.able – how to value the best stocks and buy them for less than they are worth.
Roger appears regularly on television and radio, and in the press, including ABC radio and TV, The Australian and Ausbiz. View upcoming media appearances.
This post was contributed by a representative of Montgomery Investment Management Pty Limited (AFSL No. 354564). The principal purpose of this post is to provide factual information and not provide financial product advice. Additionally, the information provided is not intended to provide any recommendation or opinion about any financial product. Any commentary and statements of opinion however may contain general advice only that is prepared without taking into account your personal objectives, financial circumstances or needs. Because of this, before acting on any of the information provided, you should always consider its appropriateness in light of your personal objectives, financial circumstances and needs and should consider seeking independent advice from a financial advisor if necessary before making any decisions. This post specifically excludes personal advice.
