The start of the new quarterly earnings season is prompting investors to focus less on the pace of artificial intelligence development and more on its cost. Google has raised its forecast for capital expenditures in 2026 from $190 billion to $205 billion. For a technology giant, an additional $15 billion may not seem like a critical sum in itself, but the market saw it as a worrying signal. The company now expects to spend significantly more than previously anticipated, while the path to monetising AI remains far from clear.
The issue goes beyond the sheer scale of spending. Technology companies are simultaneously increasing capital expenditures, expanding computing infrastructure, and keeping prices for AI services relatively low. As a result, the growth in demand for computing does not yet guarantee comparable growth in profits for end providers of AI services.
This situation is not unique to Google. Technology giants such as Meta, Amazon, and Microsoft are also actively expanding their data centres, and their spending on AI infrastructure is already measured in hundreds of billions of dollars. An increasing number of companies are using debt financing to build new data centres, turning the AI buildout into a vast financial ecosystem in which one participant’s spending becomes another’s revenue.
Source: Trading View
Nvidia’s example is particularly illustrative. The company remains one of the main beneficiaries of the AI boom, supplying a significant portion of the accelerators for new computing capacity. However, Nvidia is now gradually moving beyond its traditional role as a hardware manufacturer and is beginning to participate in financing the infrastructure where its products will ultimately be deployed.
The company has announced the creation of a fund that could potentially allocate more than $500 billion to the development of AI infrastructure. Institutional investors include Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. Nvidia itself is ready to provide up to $125 billion, but no more than 25% of the funding for specific projects.
At first glance, this looks like yet another confirmation of the huge demand for computing power. But the financial side of the issue is more complex. If a manufacturer simultaneously supplies accelerators, participates in financing the construction of data centres, and helps customers raise capital to purchase its own products, a potentially self-reinforcing financing loop emerges. Institutional investors’ money funds data centre construction; those data centres purchase Nvidia accelerators; and Nvidia generates revenue while simultaneously helping support demand for its own products.
The scale of this investment cycle is reflected in overall capital spending across the U.S. technology sector. In 2026, it is expected to reach about $765 billion, rising to approximately $1.2 trillion in 2027. Interestingly, the anticipated Anthropic IPO could value one of the major players in the AI race at at least $2 trillion — well above either figure.
Against this backdrop, the situation with Intel looks particularly interesting. The company has decided to raise $15 billion through a stock offering, with some of the proceeds going toward the development of artificial intelligence technologies, physical AI, and advanced chip packaging. At the same time, Intel plans to increase its own capital expenditures to approximately $20 billion this year, with spending expected to rise further in the coming year.
Thus, even semiconductor manufacturers are forced to seek additional capital to participate in the AI race. The market reacted negatively: Intel shares fell more than 5% after the announcement. For existing shareholders, the issuance could dilute their stakes, and the need to raise $15 billion shows how costly it is for the company to try to compete in the new technological reality.
Additional uncertainty is created by China. American investors are counting on the fact that the shortage of advanced accelerators will sustain strong demand for Nvidia’s accelerators for many years to come. However, Chinese developers are gradually demonstrating the ability to create competitive AI models even with limited access to modern hardware. If improvements in model efficiency significantly reduce the amount of computing power required to achieve comparable results, some of today’s investment plans could eventually prove excessive.
That is why the $15 billion increase in Google’s spending forecast may be much more significant than it initially appears. This is not just about an additional line item in one company’s budget, but a recognition that the cost of the AI race continues to rise while its ultimate economic returns remain uncertain.
As long as money continues to circulate within the system, growth appears to be self‑sustaining. Technology companies are building data centres, Nvidia is supplying accelerators, financial institutions are providing capital, and manufacturers like Intel are trying to capture their share of the market.
If AI spending continues to grow faster than revenue and profits, the market will sooner or later face the need to reassess the returns on these investments. Today’s winners of the AI boom may then split into two groups — those that have built sustainable new businesses and those that were primarily equipment suppliers to a massive investment cycle. Since some of these companies may carry outsized weight in major benchmarks, any such reassessment would likely extend well beyond individual stocks — affecting tech-concentrated Nasdaq 100 futures among others.
Leave a Comment