Marketdash
Back to News

Big Tech's AI Spending Spree Could Turn Them Into Low-Return Utilities, Analyst Warns

MarketDash Editorial Team
21 days ago
Wall Street's AI darlings are pouring hundreds of billions into infrastructure, but Sparkline Capital's Kai Wu sees a dangerous shift toward capital-heavy models that have historically crushed investor returns. The Magnificent Seven might be building their way into mediocrity.

Here's an uncomfortable question for anyone holding the Magnificent Seven: what if the biggest winners of the AI revolution aren't the companies spending hundreds of billions to build it?

That's the warning from Kai Wu, founder of Sparkline Capital, who's watching Wall Street's favorite tech giants transform themselves from nimble, high-margin software companies into something that looks uncomfortably like... utilities. And not in a good way.

The thesis is straightforward but unsettling: AI investment is exploding, capital expenditures are skyrocketing, and the asset-light business models that made Big Tech so profitable are disappearing fast. What's replacing them? A capital-intensive arms race that history suggests will end badly for shareholders.

The Numbers Are Getting Wild

U.S. tech companies are expected to pour nearly $400 billion into capital expenditures this year alone. McKinsey projects total AI-related spending will hit $5.2 trillion by 2030. That's trillion with a T.

Markets have been cheering this spending spree. Last month, Oracle Corp. (ORCL) jumped 36% in a single day after announcing an OpenAI data center deal. Wall Street loves a good AI story.

But Wu sees red flags everywhere. "While the market has rewarded this spending so far," he wrote in his report, "we find that historical capital expenditure booms have typically resulted in overinvestment, excess competition, and poor stock returns."

The math is sobering. Wu estimates current AI-related revenues sit around $20 billion. To justify the investment trajectory we're on, those revenues need to grow 100-fold by the end of the decade. That's not a typo—one hundred times current levels. And that assumes adoption accelerates dramatically from here.

Sound familiar? It should. Wu draws a direct line to the telecom fiber optic boom of the late 1990s, when companies overspent building infrastructure that took years to generate returns. "The parallels to past technology buildouts are hard to ignore," he notes.

From Software Giants to Infrastructure Utilities

Since ChatGPT launched, JPMorgan data shows AI stocks have contributed 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth. The Magnificent Seven—Apple Inc. (AAPL), Microsoft Corp. (MSFT), Amazon.com Inc. (AMZN), Meta Platforms Inc. (META), Alphabet Inc. (GOOGL), Nvidia Corp. (NVDA) and Tesla Inc. (TSLA)—now represent 35% of the entire S&P 500, a concentration that exceeds even the dot-com bubble peak.

For years, these companies were celebrated for their asset-light models. Software scales beautifully—you write code once and sell it forever with minimal marginal costs. High margins, high returns, happy shareholders.

That era is ending. Fast.

According to Wu, "the AI arms race is transforming Big Tech from asset-light to asset-heavy, a model we find associated with inferior returns." Capital expenditures by these companies have exploded from 4% of revenue in 2012 to 15% today. Meta, Microsoft and Google are expected to spend between 21% and 35% of revenue on infrastructure—levels you'd expect from electric utilities or railroads, not Silicon Valley growth machines.

"The Magnificent Seven's capital intensity is quickly approaching that of utilities," Wu warned.

All that spending comes with consequences. Depreciation costs alone could climb from $150 billion to $400 billion over the next five years, Wu estimates, and that's before accounting for how quickly tech infrastructure becomes obsolete. GPU clusters don't age gracefully.

There's also something slightly absurd happening with the funding. "Nvidia recently invested $100 billion in OpenAI, providing capital that OpenAI could use to buy Nvidia chips," Wu wrote. A week later, OpenAI signed a deal with AMD. It's starting to look like a circular money-shuffling exercise, with AI companies essentially funding each other's spending sprees.

The Prisoner's Dilemma Nobody Can Escape

Wu compares the current situation to a classic game theory trap: "While the optimal move is for firms to mutually agree to moderate their AI investments... each firm is incentivized to unilaterally ramp up investment."

Even if massive spending destroys margins industry-wide, no individual player can afford to pull back without risking irrelevance. It's a self-destructive arms race where everyone knows the ending but nobody can stop running.

Meta's Mark Zuckerberg put it bluntly: "If we end up misspending a couple of hundred billion dollars... that is going to be very unfortunate. But... the risk is higher on the other side."

Google co-founder Larry Page went even further: "I'm willing to go bankrupt rather than lose this race."

When company founders are openly discussing bankruptcy scenarios as preferable to slowing down, you know the competitive dynamics have gotten weird.

Who Actually Profits From This?

History offers an uncomfortable answer: usually not the infrastructure builders.

Wu points out that 85% of the fiber optic cables laid during the dot-com era went unused—at least initially. But that massive overbuild created cheap infrastructure that companies like Netflix and Facebook later exploited to build dominant businesses.

"Excess supply drives down prices, effectively resulting in a subsidy from the builders to their customers," Wu explained.

He's betting the same pattern will repeat with AI. The companies spending hundreds of billions on data centers and GPU clusters may be inadvertently subsidizing the next generation of AI-native businesses that will use that infrastructure without having paid to build it.

This dynamic creates promising AI infrastructure, but the economics remain deeply uncertain, and investors may be significantly overestimating the returns for companies doing the heavy lifting.

So Where Should the Smart Money Go?

Wu's recommendation is straightforward: look for "AI early adopters" rather than infrastructure builders. These are companies positioned to benefit from AI capabilities without shouldering massive capital expenditure burdens.

His list includes Walmart Inc. (WMT), Caterpillar Inc. (CAT), JPMorgan Chase & Co. (JPM), Sony Group Corp. (SONY), Siemens AG (SIEGY), and Roche Holding AG (RHHBY).

These companies, he argues, offer lower capital requirements, better valuations, and the ability to leverage cheap AI tools built by others. They get the benefits without the crushing infrastructure costs.

Meanwhile, AI infrastructure stocks led by megacaps like Nvidia, Amazon, and Microsoft have seen their valuation premiums balloon from 32% to 137% since 2015. That kind of expansion doesn't leave much room for disappointment.

"Investors should heed the lessons of history," Wu wrote. "Aggressive capital spending has generally led to poor stock returns."

The irony would be perfect: the companies spending the most to build the AI future might generate the lowest returns, while businesses that simply use AI as another tool quietly outperform. It's happened before with railroads, telecom, and fiber optics. The question is whether this time really is different—or whether Big Tech is about to learn an expensive lesson about the perils of capital-intensive business models.

Back to News

Big Tech's AI Spending Spree Could Turn Them Into Low-Return Utilities, Analyst Warns

MarketDash Editorial Team
21 days ago
Wall Street's AI darlings are pouring hundreds of billions into infrastructure, but Sparkline Capital's Kai Wu sees a dangerous shift toward capital-heavy models that have historically crushed investor returns. The Magnificent Seven might be building their way into mediocrity.

Here's an uncomfortable question for anyone holding the Magnificent Seven: what if the biggest winners of the AI revolution aren't the companies spending hundreds of billions to build it?

That's the warning from Kai Wu, founder of Sparkline Capital, who's watching Wall Street's favorite tech giants transform themselves from nimble, high-margin software companies into something that looks uncomfortably like... utilities. And not in a good way.

The thesis is straightforward but unsettling: AI investment is exploding, capital expenditures are skyrocketing, and the asset-light business models that made Big Tech so profitable are disappearing fast. What's replacing them? A capital-intensive arms race that history suggests will end badly for shareholders.

The Numbers Are Getting Wild

U.S. tech companies are expected to pour nearly $400 billion into capital expenditures this year alone. McKinsey projects total AI-related spending will hit $5.2 trillion by 2030. That's trillion with a T.

Markets have been cheering this spending spree. Last month, Oracle Corp. (ORCL) jumped 36% in a single day after announcing an OpenAI data center deal. Wall Street loves a good AI story.

But Wu sees red flags everywhere. "While the market has rewarded this spending so far," he wrote in his report, "we find that historical capital expenditure booms have typically resulted in overinvestment, excess competition, and poor stock returns."

The math is sobering. Wu estimates current AI-related revenues sit around $20 billion. To justify the investment trajectory we're on, those revenues need to grow 100-fold by the end of the decade. That's not a typo—one hundred times current levels. And that assumes adoption accelerates dramatically from here.

Sound familiar? It should. Wu draws a direct line to the telecom fiber optic boom of the late 1990s, when companies overspent building infrastructure that took years to generate returns. "The parallels to past technology buildouts are hard to ignore," he notes.

From Software Giants to Infrastructure Utilities

Since ChatGPT launched, JPMorgan data shows AI stocks have contributed 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth. The Magnificent Seven—Apple Inc. (AAPL), Microsoft Corp. (MSFT), Amazon.com Inc. (AMZN), Meta Platforms Inc. (META), Alphabet Inc. (GOOGL), Nvidia Corp. (NVDA) and Tesla Inc. (TSLA)—now represent 35% of the entire S&P 500, a concentration that exceeds even the dot-com bubble peak.

For years, these companies were celebrated for their asset-light models. Software scales beautifully—you write code once and sell it forever with minimal marginal costs. High margins, high returns, happy shareholders.

That era is ending. Fast.

According to Wu, "the AI arms race is transforming Big Tech from asset-light to asset-heavy, a model we find associated with inferior returns." Capital expenditures by these companies have exploded from 4% of revenue in 2012 to 15% today. Meta, Microsoft and Google are expected to spend between 21% and 35% of revenue on infrastructure—levels you'd expect from electric utilities or railroads, not Silicon Valley growth machines.

"The Magnificent Seven's capital intensity is quickly approaching that of utilities," Wu warned.

All that spending comes with consequences. Depreciation costs alone could climb from $150 billion to $400 billion over the next five years, Wu estimates, and that's before accounting for how quickly tech infrastructure becomes obsolete. GPU clusters don't age gracefully.

There's also something slightly absurd happening with the funding. "Nvidia recently invested $100 billion in OpenAI, providing capital that OpenAI could use to buy Nvidia chips," Wu wrote. A week later, OpenAI signed a deal with AMD. It's starting to look like a circular money-shuffling exercise, with AI companies essentially funding each other's spending sprees.

The Prisoner's Dilemma Nobody Can Escape

Wu compares the current situation to a classic game theory trap: "While the optimal move is for firms to mutually agree to moderate their AI investments... each firm is incentivized to unilaterally ramp up investment."

Even if massive spending destroys margins industry-wide, no individual player can afford to pull back without risking irrelevance. It's a self-destructive arms race where everyone knows the ending but nobody can stop running.

Meta's Mark Zuckerberg put it bluntly: "If we end up misspending a couple of hundred billion dollars... that is going to be very unfortunate. But... the risk is higher on the other side."

Google co-founder Larry Page went even further: "I'm willing to go bankrupt rather than lose this race."

When company founders are openly discussing bankruptcy scenarios as preferable to slowing down, you know the competitive dynamics have gotten weird.

Who Actually Profits From This?

History offers an uncomfortable answer: usually not the infrastructure builders.

Wu points out that 85% of the fiber optic cables laid during the dot-com era went unused—at least initially. But that massive overbuild created cheap infrastructure that companies like Netflix and Facebook later exploited to build dominant businesses.

"Excess supply drives down prices, effectively resulting in a subsidy from the builders to their customers," Wu explained.

He's betting the same pattern will repeat with AI. The companies spending hundreds of billions on data centers and GPU clusters may be inadvertently subsidizing the next generation of AI-native businesses that will use that infrastructure without having paid to build it.

This dynamic creates promising AI infrastructure, but the economics remain deeply uncertain, and investors may be significantly overestimating the returns for companies doing the heavy lifting.

So Where Should the Smart Money Go?

Wu's recommendation is straightforward: look for "AI early adopters" rather than infrastructure builders. These are companies positioned to benefit from AI capabilities without shouldering massive capital expenditure burdens.

His list includes Walmart Inc. (WMT), Caterpillar Inc. (CAT), JPMorgan Chase & Co. (JPM), Sony Group Corp. (SONY), Siemens AG (SIEGY), and Roche Holding AG (RHHBY).

These companies, he argues, offer lower capital requirements, better valuations, and the ability to leverage cheap AI tools built by others. They get the benefits without the crushing infrastructure costs.

Meanwhile, AI infrastructure stocks led by megacaps like Nvidia, Amazon, and Microsoft have seen their valuation premiums balloon from 32% to 137% since 2015. That kind of expansion doesn't leave much room for disappointment.

"Investors should heed the lessons of history," Wu wrote. "Aggressive capital spending has generally led to poor stock returns."

The irony would be perfect: the companies spending the most to build the AI future might generate the lowest returns, while businesses that simply use AI as another tool quietly outperform. It's happened before with railroads, telecom, and fiber optics. The question is whether this time really is different—or whether Big Tech is about to learn an expensive lesson about the perils of capital-intensive business models.