To help you understand the trends surrounding business and technology and what we expect to happen in the future, our highly experienced Kiplinger Letter team will keep you abreast of the latest developments and forecasts. (Get a free issue of The Kiplinger Letter or subscribe.) You'll get all the latest news first by subscribing, but we will publish many (but not all) of the forecasts a few days afterward online. Here's the latest…
Artificial intelligence will be transformational, akin to the dawn of the internet, but unfolding at a faster pace. But overspending on AI is creating a bubble, in our judgment, like other major industrial projects in history, such as railroads, electrification or the web.
Here’s how we currently see the state of play.
AI outpaces previous building booms
AI is the largest infrastructure project ever. U.S. spending is double the dot-com peak, when the telecom and fiber build-out hit 1.2% of GDP in 2000, according to an analysis by the investment firm Apollo. AI is now at 2% of GDP and will soon be 3%.
Big Tech spending will hit $1 trillion next year in capital expenditures from Alphabet, Amazon, Meta and Microsoft, according to data compiled by Bloomberg. That’s up from $700 billion this year and $380 billion in 2025. Back in 2022, it was $150 billion. With big spenders Oracle and SpaceX, it’s even higher. Though the spending is expected to level off, there’s no end in sight for $1 trillion-plus yearly outlays.
Unlike previous infrastructure build-outs with up-front spending on things that lasted decades, AI spending resets every four to six years as high-priced AI chips in data centers need to be upgraded, along with other equipment.
And improving AI is getting more expensive, requiring more computation and electricity to develop leading-edge AI. New AI models can cost $1 billion. A huge, 1-gigawatt data center? $60 billion, according to Nvidia CEO Jensen Huang.
Some of the most significant risks
Among risks that bear watching: The drastic business shift of tech giants from high-margin software to capital-intensive infrastructure. Once fountains of cash, most of the top AI spenders will have negative free cash flow this year and in 2027. S&P Global doesn’t project a recovery of FCF for most major players until 2029, a timeline that would lead to mounting pressure and possible investor blowback.
Debt-fueled financing with increasingly complex structures, such as special-purpose vehicles, or SPVs, which get debt off Big Tech’s books. Recently, Nvidia spearheaded a plan with investment firms, such as BlackRock and Goldman Sachs, to unleash $500 billion in AI financing. The plan seeks to turn AI chips in a data center into an investable asset and highlights how the AI boom is being financed by unconventional means. BlackRock CEO Larry Fink compared it to the creation of mortgage-backed securities in the 1970s. Those MBS securities went on to play a leading role in the 2008 financial crisis.
The massive spending commitments. For example, Alphabet has $811 billion in obligations, mostly for AI. OpenAI has plans to spend $800 billion through 2030 to build some of the world’s biggest data centers in New Mexico, Texas, Wisconsin and Michigan. Flexibility may be hard to come by with multiyear deals for chips, power and land. Big Tech spenders are forecasting demand for AI years in advance, an inherently dicey proposition.
Circular deals, such as Nvidia financing the purchase of its own chips. For example, Nvidia is providing up to $105 billion in financing for an OpenAI data center in Ohio. Even with a torrent of cash from soaring chip sales, Nvidia can’t maintain the pace of its recent extensive dealmaking, according to a report by Morningstar’s PitchBook. Another example: Microsoft has a 25% stake in OpenAI, which accounts for 70% of Microsoft’s AI sales.
The collective action problem of tech executives, all willing to overspend in the pursuit of “superintelligence,” an ill-defined AI breakthrough. “The risk of underinvesting is dramatically greater than the risk of overinvesting,” said Sundar Pichai, the CEO of Alphabet, back in 2024. Four or five tech companies breaking the bank to build the most advanced AI is not viable over the long run. It would not be surprising if, eventually, the race results in two to three leading-edge providers.
Other risks include AI turning into a low-cost commodity, stemming from both efficiency gains and cheaper AI services, including Chinese models, flooding the market. That could lessen demand for the most advanced, premium-priced AI. A terror attack or other rogue event involving AI could spur regulations that slow down or restrict the technology in an onerous way. Or AI improvements could slow or stall, lessening the pressure to spend so heavily on developing new AI systems.
The arguments against a bubble
Many analysts take issue with bubble talk, since revenue is exploding at Anthropic, OpenAI and other firms. Tech giants such as Alphabet and Amazon have diversified global businesses that cater to both businesses and consumers, with massive amounts of revenue. Plus, demand for AI is vastly outpacing supply, even as businesses cope with high costs.
AI use has exploded among businesses and consumers as free chatbots from Meta, Alphabet and OpenAI have proliferated. OpenAI has 900 million active weekly users, plus over 50 million consumer subscribers, with plans starting at $8 per month. Anthropic has hundreds of thousands of paying business customers and is on track to make tens of billions of dollars in sales per year.
But the revenue earned from AI is still far from keeping up with the necessary computing’s soaring cost. For AI spending to be sustainable, the tech must expand beyond employee productivity software or even entire IT budgets. AI must disrupt and reshape entire industries, from manufacturing and financial services to health care and retail. Businesses would need to shift huge amounts of spending that they now devote to labor and funnel it to AI products and services.
Amid AI’s fast adoption, tech executives have acknowledged the risk of a bubble. For example, Amazon founder Jeff Bezos has called what’s happening an “industrial bubble,” where investor excitement leads to certain money-losing bets but also breakthrough ideas. Meta CEO Mark Zuckerberg has said that a bubble is “definitely a possibility,” like those stemming from major infrastructure projects of the past.
"I am worried about the world's compute build-out plans," said OpenAI CEO Sam Altman in an interview in early September. Altman sees the first signs of "unsustainable silliness" of certain AI cloud companies, known as neoclouds, that claim they will build “gigantic amounts of compute” without securing the needed financing or customers.
The bottlenecks to AI progress
There are likely bottlenecks that slow AI adoption by businesses and consumers, ranging from human behavior to regulatory issues. That friction is sure to delay the full potential of AI, including all the enhanced productivity it promises. For Big Tech spenders, such delays will be hard to endure as they seek higher revenue and profits.
Another issue not to be overlooked is the major roadblocks to AI deployment, especially accessing the needed power, along with shortages of chips, equipment and labor. Some analysts point to roadblocks as a built-in factor staving off a bubble, preventing too much development, too fast. But those roadblocks will be more evident over the long term rather than in the short run. The explosion of data center opposition is a real concern, but tech giants have ample data centers in the pipeline for now.
Bubbles fund breakthrough ideas
It’s not clear when a bubble would burst or deflate. A fallout could include AI start-ups going out of business and certain infrastructure providers facing lumbering debt loads. The Kiplinger Letter’s view is that a pullback in spending could spark a mild recession, as the tech bust did in 2001.
Still, overspending on AI will be useful, pouring money into valuable areas that lay the foundation for an economic leap. Promising areas include AI software that automates complex white-collar office tasks, AI models that accelerate drug discovery and AI coding tools that automate computer programming. The ultimate winners of the race may not be the heavy spenders of today, but future start-ups that launch new products and services enabled by AI.
But on the cusp of all this potential, there’s no doubt that uncertainty and financial risk are rising.
This forecast first appeared in The Kiplinger Letter, which has been running since 1923 and is a collection of concise weekly forecasts on business and economic trends, as well as what to expect from Washington, to help you understand what’s coming up to make the most of your investments and your money. Subscribe to The Kiplinger Letter.