Tech firms plan to pour trillions of dollars into developing artificial intelligence (AI). Every incremental dollar of new investment carries an even higher value in the stock market. This investment splurge is unlikely to earn a positive return on capital. But that’s beside the point. Companies and investors are trapped inside the bubble, and for many of them there’s no escape.
The cloud computing giants’ multibillion-dollar data centres come with grandiose names: OpenAI is constructing Stargate, Meta Platforms has Prometheus, while Elon Musk’s xAI is building a supercomputer called Colossus. Icarus would appear more fitting. Big Tech’s business model used to be capital-light and cash flow-heavy. No longer. Aggregate capital spending at Amazon, Alphabet, Meta and Microsoft has risen from below US$100 billion (RM421 billion) five years ago to nearly US$300 billion this year. Around two-thirds of US venture capital deals involve AI or machine learning.
The surge in AI investment added a percentage point to US GDP growth in the first quarter, according to Apollo Global Management. The tech firms are just getting started. Consultancy Bain & Company’s latest Global Technology Report estimates that some US$500 billion a year will be spent in the US over the rest of the decade. That’s a relatively sober forecast. Morgan Stanley sees cumulative investment in US data centres reaching US$3 trillion by 2029. McKinsey & Co. anticipates the bill will exceed US$5 trillion by 2030.
Estimates of the level of AI-related revenue needed to justify this investment splurge are equally wide. A rapid payback is necessary because the expensive graphics processors housed in the data centres have a short lifespan. Bain reckons some US$2 trillion in additional AI revenue is required by 2030. Working off Morgan Stanley’s US$3 trillion estimate, Charles Carter of Marathon Asset Management calculates that those investments must generate a similar level of annual AI sales if they are to earn their cost of capital. Yet US$3 trillion is equivalent to a tenth of current US GDP and 70 times Citi’s estimate of the revenue that AI will generate this year.
Tech leaders claim that AI is going to usher in a new golden age, boosting productivity and profits. But there’s little evidence so far to justify the hype. A recent report from the Massachusetts Institute of Technology (MIT) found that 95% of businesses that had integrated AI into their operations had yet to see any return on their investment. Of the nine sectors examined in the study, only media and technology had experienced major structural changes. It’s not for want of effort, as the report states: “adoption is high, but disruption is low.”
The trouble, according to MIT, is that generative AI systems don’t retain feedback, adapt to context or improve over time. For mission-critical work, the vast majority of companies still prefer to use humans. Instead, employees are using personal chatbot accounts for relatively mundane tasks such as reading emails. Though loss-making OpenAI’s revenue has grown rapidly, it does not justify the investment spend. Fewer than 2% of the 800 million or so people using ChatGPT pay for the service, and a growing number of them live in low-income countries such as India.
So why do companies continue to sink trillions of dollars into highly speculative and possibly unprofitable investments? Carter suggests that AI creates a so-called “innovator’s dilemma” for Big Tech. There’s a danger that the new technology will, at some stage, drain the competitive moats that surround their hitherto extremely profitable business models. Cloud computing firms also face a classic prisoner’s dilemma: if one operator fails to invest, it risks losing business to competitors which forge ahead. European mobile phone companies faced the same quandary during the telecom boom of the late 1990s and ended up massively overpaying for 3G spectrum auctioned by the UK and other countries.
Several top tech executives have publicly acknowledged the dilemma. Last year, Alphabet CEO Sundar Pichai stated that “the risk of underinvesting is dramatically greater than the risk of overinvesting for us.” In a recent interview, Meta founder Mark Zuckerberg said much the same: “If we end up misspending a couple of hundred billion dollars, I think that is going to be very unfortunate obviously … I actually think the greater risk is on the other side.”
Investors also face a dilemma. They have good reasons to adopt a conservative stance. Valuations for AI-related businesses are elevated. Around 35% of the market capitalisation of the S&P 500 Index trades at more than 10 times sales, according to investment firm GQG. Big Tech is diverting its cash flow into speculative investments. The previously lucrative cartel of cloud providers is being disrupted by new entrants such as CoreWeave.
The current surge in profit generated by AI investment is unlikely to last. Strategist Gerard Minack writes, “as in the TMT cycle … a positive feedback loop (exists) between rising investment spending and rising profits: the firm selling capital goods immediately reports its profits in full, while the firm buying the capital good depreciates its cost over time.” After the TMT boom ended, the returns for internet hardware companies, such as Cisco collapsed.
On the other hand, AI is generating extraordinary stock market returns. Nvidia’s stock is up nearly 350 times over the past decade. Oracle’s recent announcement that cloud infrastructure revenue would reach US$144 billion by 2030 sent its shares up 36% in a single day, adding around US$250 billion to the software firm’s market value. The US market is more concentrated than ever in a handful of AI-related names. Minack’s list of the leading AI plays (Alphabet, Amazon, Broadcom, Meta, Microsoft, Nvidia, Oracle, and Palantir) has climbed nearly 30% since the start of the year, while the rest of the S&P 500 has delivered just 8%.
Professional investors are in a similar position to their predecessors during the TMT boom 25 years ago. If they don’t participate in the bubble they risk extreme underperformance and a possible loss of business. Yet if they run with the herd, they face potentially large losses at some uncertain future date. “At least for active investors it’s a dilemma,” says Marathon’s Carter, “for passive they’re just prisoners.”
This article is written by Edward Chancellor, a Reuters Breakingviews contributor.