The Great AI Capital Expenditure Debate: What It Means for Investors

The rise of AI this decade has led to a spending boom driven by the world’s most valuable tech companies. Total AI spending is expected to climb from $1.76 trillion in 2025 to $2.6 trillion this year and $3.49 trillion next year, according to tech research firm Gartner. 

Tech giants such as Alphabet, Amazon, Meta, and Microsoft are boosting their spending on new plants and equipment from $125 billion in 2021 to an estimated $700 billion in 2026 and $1 trillion next year. In turn, AI accounted for a third of U.S. economic growth in 2026, said analysts from global bank ING, driven by a spending surge for building data centers and expanding power grids.

AI is now becoming the most well-funded infrastructure investment in recent history, surpassing what the U.S. spent on the Apollo space program in the 1960s and the multi-decade interstate highway system buildout. And observers expect even more spending in the years ahead. JPMorgan projected that AI data center spending could reach more than $5.5 trillion through 2030, with external estimates expecting as much as $10 trillion.

At the core of this capex explosion are the AI boom architects that invest in each other’s growth.

Today, the world’s biggest AI companies are interconnected in an investment web wherein they finance each other through equity, venture, and compute (processing power for AI models) deals. This is the assertion of the AI Circular Economy website, which tracks 29 AI companies’ transactions with one another. 

Proponents have hailed this arrangement as the growth driver keeping the global economy running amid crises in many other industries, while critics called it a circular financing scheme that could crash and burn if demand is not sustained. 

Chip making giant Nvidia is one of the focal points of the AI capex boom. In August, Nvidia agreed to provide up to $105 billion worth of credit and compute for an OpenAI data center in Ohio, while also securing deals with six large asset managers to line up more than $500 billion in capital for companies that will build data centers and buy Nvidia chips in the process.

Nvidia’s recent investments, while huge, have been normalized amid the rise of AI. Some of the other notable business arrangements among AI’s stewards, according to AI Circular Economy, include: 

• Microsoft’s 27% stake in OpenAI Group PBC, alongside OpenAI’s commitment to purchase an additional US$250 billion in Azure services.
• Amazon’s over-$83 billion into Anthropic and OpenAI combined. Both companies committed more than $238 billion of compute back to Amazon Web Services.
• Google’s $40 billion investment into Anthropic, which agreed to deploy a million Google tensor processing units (TPUs) used for boosting AI workloads.

In response to critics of Nvidia’s deals, CFO Colette Kress pointed out, “We are going through a major computing platform shift, the creation of one of the most important technologies in human history, and these are once-in-a-generation companies.”

The main hypothesis driving the AI capex boom is the assumption that as AI grows, compute will be a high-demand resource that, as OpenAI President Greg Brockman put it, is “becoming the new oil.” 

AI companies say they are seeing a surge in demand for their services. AI-centric cloud provider Nebius shared there was more demand than they’ve been able to fulfill for some time now, while the CEO of processor maker Cerebras Systems, Andrew Feldman, noted that “the demand for compute far outstrips available capacity” contributing to shortages of data-center capacity and other inputs.

“I somewhat think of AI demand as almost unlimited,” former Intel CEO Pat Gelsinger told CNBC, adding that energy availability is “the only real limiter.” 

Some observers acknowledged the circular nature of the deals, but pointed out that  the companies involved have sound financial numbers. “The interconnected nature of the AI ecosystem is real,” said Gary Tan, portfolio manager at Allspring Global Investments. “But near term risks are mitigated by the strong balance sheets, cash flows and credit quality of the major providers funding the buildout.”

So far, the numbers have supported this assessment. While the S&P 500 needs $2 to $3 in assets to generate $1 in sales, a gigawatt of computing power costing $40 billion to build can produce $40 billion in sales, an impressive 1:1 asset-to-sales ratio, according to Barrons. Meanwhile, JPMorgan analysts predicted that AI cloud providers, model providers, and neoclouds could make a combined revenue run-rate of about $1.6 trillion by the end of 2026, then grow that by 10% to 20% annually to reach $2.5 trillion to $3 trillion by 2030.

Nonetheless, the rising capex of the AI frontrunners is triggering questions from the investors that finance them.

With AI-related capex currently in the trillions, AI companies’ investors are beginning to demand more results from the money being spent on advancing the technology, according to CNBC. These companies are on track to have bigger capex than free cash flow by 2027. 

This unease has been unfolding in recent AI-centric financing deals. In July, Meta secured a $12.3 billion bond to build a data center in Texas. This bond has a 7.5% rate, considered higher than the going rates for previous projects. Nvidia’s recent $105 billion deal for OpenAI’s data center buildout was originally worth up to $250 billion, marking a decrease in commitment as deal negotiations went on.

When asset manager Fidelity surveyed financial advisors in July, 42% of them said AI companies’ valuations have been too high compared to financial performance. Fidelity added that advisors are increasingly seeking evidence of AI-driven growth, a noticeable shift from previous enthusiasm about AI investments. 

This more cautious stance may end up clashing with even higher capex projections in the years ahead. Dom Rizzo, the global technology equity fund manager at T. Rowe Price, forecasted that companies are halfway through the spending cycle, and are on the cusp of an acceleration in capex. 

Nevertheless, the evolving financial story for AI companies may assuage capex concerns. JPMorgan recently said that the acceleration in revenue at AI companies has strengthened the economic case behind the sector’s enormous infrastructure spending.

Meanwhile, as investors debate about AI capex, the wider AI buildout narrative may unearth other opportunities for investors. After all, spending is one thing, but getting the resources needed for a buildout is another.

In the AI race, capital isn’t the only limit. Power infrastructure, chip manufacturing, and enterprise integration bottlenecks dictate real-world outcomes. The evolving AI landscape has resulted in a few unlikely winners, like Japanese toilet maker Toto and food seasoning brand Ajinomoto, which found new lucrative roles in the AI capex boom by making chipmaking equipment parts and insulating materials, respectively.

As the cases of Toto and Ajinomoto illustrate, the AI investment landscape also extends to companies producing materials and components used in power grids, semiconductor facilities, data centers and related infrastructure.

For instance, let’s focus on the power angle. Energy systems require significant mineral inputs. Wind turbines use rare earth elements in magnets. Solar supply chains require materials such as silicon, silver, copper, and aluminium. Power grids require large amounts of copper and aluminium. As electricity demand grows from data centers, EVs, industrial electrification, and renewables, investors are paying closer attention to these resources. SmallCap Asia discussed this in detail in a recent article.

While it may be too early to identify which specific resource shortages could slow AI infrastructure deployment, it is clear that AI will require more physical materials, more power, and more resilient supply chains, according to fund manager VanEck.

As the AI story continues to evolve, companies need to clearly communicate how their products and services fit within the broader economy and support emerging technologies. Global One Media works with public and pre-IPO companies involved in exploring for and supplying the physical materials that support technology, energy and infrastructure markets.

Contact us to learn how Global One Media can help communicate your company’s investment story to a global audience.