Goldman Sachs warns that $5.3 trillion in AI capital expenditure has approached credit saturation limits, with enterprises increasingly shifting focus to compute cost optimization and efficiency improvements.

The surge in artificial intelligence infrastructure investment is reshaping the global capital markets landscape, with hidden debt risks that cannot be overlooked.

According to Goldman Sachs' latest forecast, hyperscale cloud computing enterprises will collectively spend $5.3 trillion on capital expenditures in artificial intelligence and data centers from 2025 to 2030, marking an unprecedented capital expenditure supercycle.

Goldman Sachs anticipates that these hyperscale enterprises may face constraints due to saturation in liquidity credit markets and will therefore be forced to seek financing from various sources.

Gary Marcus, Professor Emeritus at New York University, characterized Goldman Sachs' statement as a "terrifying sentence" when sharing related analysis and commented:

"For me, the question is no longer whether hyperscale models will collapse, but how severe the collateral damage will be.

These hyperscale cloud service providers cannot possibly recover their $5.3 trillion investments unless they extract this money from taxpayers through large-scale government subsidies. This is precisely what they intend to do."

Meanwhile, Morgan Stanley estimates that by 2028, global data center construction capital expenditures alone will reach approximately $2.9 trillion, with a substantial proportion relying on debt financing. This means that once market adjustment occurs, losses will no longer be confined to shareholders but could spread through credit markets to society as a whole.

The flip side of this investment boom is increasingly tight corporate budgets. Early large-scale AI adopters such as Uber, Amazon, and Walmart have already imposed caps on employee AI usage or are implementing cost-cutting measures.

After Anthropic switched to token-based pricing, Workato Chief Information Officer Carter Bass saw the company's daily spending spike sevenfold.

According to Goldman Sachs analysts, AI capital expenditure growth is outpacing actual data center construction, which means future bottlenecks may shift from model demand to financing capacity, power supply, and project execution.

Morgan Stanley's analysis is more detailed. They anticipate that by 2028, among the $2.9 trillion in global data center construction capital expenditures, financing sources will be distributed accordingly.

This structure means that AI infrastructure investment is substantially credit-driven.

AI social media commentator Rohan Paul noted on X that because only a few hyperscale cloud companies can issue public debt without restriction, investors are growing concerned about issuer concentration risk.

The complexity of data center financing further exacerbates this issue.

It is not a single asset but a combination of land, power access, network connectivity, buildings, cooling systems, and AI servers, so financing needs spill across infrastructure funds, real estate funds, private credit, and corporate bonds in multiple markets.

Should systemic market adjustment occur, the transmission chain of losses would be far more complex than during the internet bubble era.

On the demand side, AI's high operating costs are forcing companies to reassess the value of each query and automated workflow.

Uber is the most representative case. Early reports indicate that this ridesharing giant exhausted its entire annual AI budget for 2026 in just a single fiscal quarter.

After using up its budget as early as April, Uber announced a monthly cap of $1,500 per token for any single AI tool used by employees. Uber President and Chief Operating Officer Andrew McDonald acknowledged:

"It is becoming increasingly difficult to justify spending on AI tokens, and it is hard to establish clear causation between such spending and actual product feature improvements."

Walmart has also imposed caps on token usage for internal AI assistants. Walmart Global Chief Technology Officer Suresh Kumar noted that usage of the company's Code Puppy programming platform has "surged" and it's time to "step back and reassess."

This trend is supported by a structural shift in billing models. Major AI labs such as Anthropic and OpenAI have transitioned some services from fixed subscriptions to token-based billing, making enterprises more cost-sensitive to each prompt and automated workflow.

Deloitte Global Generative AI Lead Kosti Perikotos stated:

"The cost of computing power has begun to capture the attention of chief financial officers and corporate boards. Consumers and enterprises have been told that AI is cheap or free, but this is not the case."

Sam Altman, Chief Executive Officer of OpenAI, acknowledged this month that cost has become a "major challenge" for customers this year—a topic rarely discussed last year.

Enterprise-level cost-cutting actions are also exerting significant pressure on upstream players in the AI industry chain.

Both Anthropic and OpenAI plan to conduct initial public offerings later this year at valuations close to $1 trillion. However, the trend of enterprises cutting AI spending is potentially pressuring these companies' revenue growth expectations.

Major AI platforms have begun responding by guiding users toward cheaper, non-frontier models to maintain adoption rates.

GitHub Chief Operating Officer Kyle Daigle revealed that Microsoft has proactively engaged customers on pricing changes, discussing "adaptability and appropriate use cases," emphasizing that "not every task requires frontier models."

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

Some enterprises have turned to open-source models, running them on local servers or personal devices to reduce reliance on AI labs and cloud providers.

Patel from Cisco articulated the dilemma facing many enterprises:

"Our engineers want more tokens, and we have to find ways to pay for them."

This statement reflects the dilemma facing the entire industry: while the strategic value of AI is widely recognized, the business case for sustaining payments for it remains subject to market scrutiny.