As businesses begin to "reduce computing power costs," Goldman Sachs warns that $5.3 trillion in AI capital expenditures is approaching credit saturation!

```

The wave of AI infrastructure investments is reshaping the global capital markets landscape, with hidden debt risks that cannot be ignored.

According to Goldman Sachs' latest forecasts, from 2025 to 2030, capital expenditures by hyperscale cloud computing companies in the AI and data center sectors will total $5.3 trillion, creating an unprecedented super cycle of capital spending.

Goldman Sachs expects hyperscale companies will need to seek financing from multiple markets, as they may face limitations due to saturation in the liquidity credit market.

When sharing a related analysis, NYU Professor Emeritus Gary Marcus described Goldman’s statement above as "a terrifying sentence." He said:

To me, the question is no longer whether the hyperscale model will collapse, but how severe the collateral damage will be.

Gary Marcus further warned:

Hyperscale cloud providers cannot possibly recoup their $5.3 trillion investment unless they extract it from taxpayers through massive government subsidies. That is exactly what they intend to do.

Meanwhile, Morgan Stanley estimates that the capital expenditure for global data center construction alone will approach $2.9 trillion by 2028, with a significant proportion relying on debt financing. This means that if there is a market correction, losses will not be limited to shareholders but may spread through credit markets to society at large.

The other side of this investment boom is enterprises tightening their budgets. Early AI adopters like Uber, Amazon, and Walmart have started to cap employee AI usage or implement cost-saving measures.

After Anthropic switched its billing model to charge by token, Carter Busse, CIO of software company Workato, saw daily expenses soar sevenfold and exclaimed:

We have created a monster.

A $5.3 Trillion Super Cycle, Financing Pressure Spreads to Credit Markets

According to Goldman Sachs analysts, AI capital spending is rising at a rate faster than actual data center construction, which means future bottlenecks may shift from the demand for models to financing capacity, electricity supply, and project execution.

Morgan Stanley’s estimates are more specific. It projects that of the $2.9 trillion in capital expenditures for global data center construction by 2028, funding sources will be as follows:

About $1.4 trillion from hyperscalers’ own cash flow;Corporate debt about $200 billion;Asset securitization credits about $150 billion;Private credit, asset-backed financing, and joint venture debt about $800 billion;Other capital about $350 billion.

This structure indicates that AI infrastructure investment is largely credit-driven.

AI commentator Rohan Paul noted on X that since only a few hyperscalers can issue bonds in the public market without limit, investors are now concerned about issuer concentration risks.

The complexity of data center financing further aggravates the issue.

It is not a single asset, but incorporates land, electricity access, network links, construction, cooling systems, and AI servers. Thus, financing needs naturally spill over into multiple markets, including infrastructure funds, real estate funds, private credit, and corporate bonds.

Should a systemic correction occur, the transmission chain for losses will be much more complex than during the Internet bubble era.

Enterprises Hit the Brakes: From "Unlimited Use" to "AI Financial Responsibility"

On the demand side, the high operational costs of AI are forcing enterprises to reassess the value of each query and automation workflow.

Uber is the most representative case. Wallstreetcn reports that the ride-hailing giant spent its entire 2026 AI budget in just one quarter.

After exhausting its budget as early as April, Uber announced a $1,500 per month limit for employees’ usage of a single AI tool. Uber President and COO Andrew Macdonald admitted:

It’s increasingly hard to justify AI token spending or to draw a clear causal line between expenditure and actual product improvement.

Walmart has similarly capped token usage for its internal AI assistant. Walmart Global CTO Suresh Kumar noted that usage of its Code Puppy coding platform “has surged dramatically,” and now it’s time to “take a step back and reassess.”

Behind this trend is a structural shift in billing models. Leading AI labs like Anthropic and OpenAI have switched some services from fixed subscriptions to token-based pricing, making companies more sensitive to the cost of every prompt and automation.

Deloitte’s Global Generative AI leader Costi Perricos said:

Compute costs are now on the radar of CFOs and boards. Consumers and businesses have always been told AI is cheap or free, but that’s far from the truth.

OpenAI CEO Sam Altman also admitted this month that costs have become a “major issue” for customers this year—a topic hardly mentioned last year.

The Contradiction Between Enterprise Cost Reduction and Lab Valuations

Cost-cutting efforts at the enterprise level are also having a significant impact on the upstream of the AI industry chain.

Both Anthropic and OpenAI plan to go public later this year with valuations approaching $1 trillion. However, the trend of companies cutting AI spending is putting potential pressure on these two firms’ revenue growth expectations.

Major AI platforms have started countermeasures, such as guiding users to adopt cheaper, non-cutting-edge models to maintain adoption rates.

GitHub COO Kyle Daigle said that Microsoft is already discussing pricing changes with clients, exploring "fit and applicable scenarios," and stressing that “not every task requires a leading-edge model.”

Microsoft, Amazon, and Google have also launched tools to automatically route user requests to the most cost-effective appropriate models.

Some enterprises are turning to open-source models, running them on local servers or personal devices to cut payments to AI labs and cloud providers.

Cisco’s Patel summed up the plight many companies face:

Our engineers want more tokens, and we have to figure out how to pay for them.

This reflects the dilemma of the entire industry: while AI’s strategic value is widely accepted, the business logic for continually paying for it still awaits market validation.

Risk Warning and DisclaimerMarket risk exists, and investments require caution. This article does not constitute personal investment advice, nor does it take into account the particular investment objectives, financial situation or needs of any individual user. Users should consider whether any opinions, views or conclusions in this article are suitable for their specific circumstances. Invest at your own risk. ```