What happens when large language models become commodities? Competitive advantage moves from the model itself to the balance sheet behind it. AI is now reversing 20 years of technology economics, turning what was once a software business into a capital-intensive industry.
For much of the past two decades, investors rewarded asset-light software companies that needed little capital and generated fat margins. Today, however, those same companies are spending at a scale the tech sector has never seen.
Since the AI boom began in 2023, Amazon, Microsoft, Alphabet and Meta have together poured $1.1 trillion into AI infrastructure. The four “hyperscalers” plan to invest another $745 billion this year alone. Capital intensity is, clearly, no longer something Big Tech can avoid. It has in fact become the cost of competing in the AI race.
But there is a far bigger shift under way: as AI models become increasingly interchangeable, competitive advantage will depend less on the models themselves than on who can finance, build and run the infrastructure behind them the most cheaply.
Which also helps to explain why Microsoft boss Satya Nadella said recently that “every model is substitutable” and Amazon chief Andy Jassy predicted that there will soon be “at least half a dozen” comparably good AI models.
That changes the basis of competition itself. Rather than betting the house on a single winning model, the hyperscalers are building more of the infrastructure capable of supporting many of them. And as that happens, financing and scale begin to matter more than owning the frontier model itself.
Yet one question still hangs over the investment cycle: whether the models themselves ultimately generate enough value to justify the trillions still being committed. The answer remains uncertain. Both OpenAI and Anthropic remain lossmaking today. Yet the flow of capital has anything but slowed.
Chipmaker Nvidia for instance, is now working with Apollo, Blackstone, Goldman Sachs and other Wall Street giants to mobilize more than $500 billion of additional capital for AI infrastructure.
And Google has gone even further. Rather than simply writing cheques, it has assembled a $200 billion financing structure with Broadcom, Apollo, Blackstone and Morgan Stanley to fund Anthropic’s chips and data centers. Which just underlines how the locus of competition has expanded into finance itself.
The tech giants already hold the strongest hand. Microsoft, Amazon and Google have the balance sheets, the cheapest capital and are generating huge revenues from the same data centers they use to train AI models. Those advantages should endure even if AI models themselves become interchangeable.
Big Tech is unlikely to have the field to itself, however. SpaceX could emerge as a serious competitor, while sovereign wealth funds such as Saudi Arabia’s PIF and Abu Dhabi’s MGX combine cheap capital, abundant power and the flexibility to work with both western and Chinese AI companies.
Regardless of who ultimately wins, the money is already moving. Cloud providers are capturing the first commercial returns, with chipmakers selling the picks and shovels and Wall Street financing the build-out. Not everyone benefits, of course. Enterprise software companies are no longer just competing with one another, but with AI infrastructure for the same corporate budgets.
IBM’s second-quarter results brought that shift into sharp relief. Customers postponed software purchases as they rushed to secure AI infrastructure ahead of expected price hikes. The result: a 25% one-day collapse in IBM’s share price in mid-July.
That shift has understandably unsettled investors. For much of the past year, Big Tech shares have been whipsawed by a key question: will the AI spending boom ever generate a decent return? The sheer scale of that splurge has already weighed heavily on free cash flow. Alphabet’s spending has pushed free cash flow into negative territory for the first time since its IPO. Meta’s free cash flow also fell sharply in the latest quarter.
However, the latest Big Tech earnings showed the investment cycle is now delivering a return in cloud computing, even as capital spending continues to soar. Overcapacity in AI infrastructure may eventually appear, but the results season shows that moment is still a long way off.
Indeed, Microsoft’s cloud business grew 32% to $39.3 billion in the latest quarter, helping drive an 18% increase in revenues. Amazon Web Services grew 37% to $42.2 billion, its fastest growth in more than four years. And Google’s cloud business surged 82% to $24.8 billion. Investors duly rewarded Microsoft’s results by adding a record $450 billion to its market value in a single day.
Yet the market has not reached a verdict, and Apple is the exception that proves the rule. While its rivals have poured hundreds of billions into AI infrastructure, it held back, and investors briefly rewarded that restraint with a $5 trillion valuation last month. The iPhone maker has therefore become the market’s control group.
Investors are now placing two very different bets: one backs companies willing to spend whatever it takes to build AI infrastructure; the other backs those that refuse to sacrifice financial discipline in the process. One side will be proved right, and one wrong.
But whichever side wins, the rules of competition have already changed. As AI models converge in capability, the companies with the strongest balance sheets, the cheapest capital and the highest utilization of their infrastructure will have the edge. AI is, in effect, becoming a financial engineering business.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
This story was originally featured on Fortune.com
