When AI Borrows Money from Wall Street: The Tech Giants' 'CapEx Cycle' Accelerates Financialization

AI infrastructure is evolving from the tech industry's 'CapEx cycle' into a financing cycle supported by the entire financial system.

Written by: Jim, MSX Maitong

Edited by: Frank, MSX Maitong

In the past two years, Wall Street's discussions about AI have revolved around one key figure: GPU, data centers, and CapEx.

However, by 2026, another figure is rapidly coming into focus: debt.

Once, the most profitable companies in Silicon Valley were accustomed to using their own cash to purchase GPUs and build data centers. Now, as the scale of AI infrastructure continues to approach hundreds of billions and even trillions of dollars, even the world's most cash-generating companies like Alphabet, Amazon, and Meta are increasingly stepping into the bond market.

Morgan Stanley predicts that global AI-related debt issuance could approach $57 billion by 2026, more than doubling from last year; as of the end of May, the scale has already reached about $236 billion, four times that of the same period last year.

Meanwhile, Alphabet, Amazon, Microsoft, and Meta are expected to spend about $700 billion this year, with Hyperscalers' capital expenditures potentially exceeding $1 trillion by 2027.

The financing methods are also continuing to extend outward.

* On August 10, NVIDIA announced a partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, aiming to leverage over $500 billion in third-party capital for AI infrastructure. * Almost simultaneously, Microsoft, Meta, Oracle, Amazon, and Alphabet disclosed future lease payment commitments that have yet to be executed, already reaching the trillion-dollar level;

When the world's wealthiest companies begin to change their financing methods, it signifies that the AI narrative is entering a new phase. The question arises: Why are these companies, which previously had hundreds of billions in cash, suddenly so keen on borrowing money?

1. The Scale of AI's Cash Burn is Becoming More Exaggerated

Alphabet is the best example of understanding this change.

From a business perspective, its latest quarter can be described as strong—Q2 revenue reached $119.8 billion, a 24% year-on-year increase; Google Cloud revenue reached $24.8 billion, with an astonishing 82% growth.

However, on the other hand, Alphabet's quarterly capital expenditures have reached about $44.9 billion, resulting in the company experiencing negative quarterly free cash flow for the first time, with Q2 free cash flow dropping to -$5.9 billion, while Alphabet has raised its 2026 capital expenditure guidance to $195 billion—$205 billion.

This is precisely the biggest difference between AI capital expenditures and the traditional software era.

Large tech companies in the software era were essentially cash machines: after the initial R&D was completed, the marginal cost of adding each user was limited, and a large portion of revenue could ultimately be retained as free cash flow.

In other words, AI is transforming tech companies back into 'heavyweights'.

GPUs, servers, high-speed networks, data centers, substations, cooling systems, and land all require substantial cash investment before real revenue is generated.

Amazon CEO Andy Jassy once explained that data centers often incur construction expenses about two years before they officially go live, while revenue must wait until the facilities are actually operational.

Thus, a natural mismatch in funding timelines arises: cash must be spent today, but revenue needs to be gradually recouped over many years. In this case, even if a company has a large amount of cash on hand, relying entirely on internal cash flow for financing may not be the most reasonable choice.

In February, Alphabet completed approximately $31.5 billion in global bond financing, including a rare 100-year bond; in August, it completed a $25 billion investment-grade dollar bond issuance. Amazon's actions are even more aggressive, first raising about $37 billion in the U.S. bond market in March, followed by a €14.5 billion bond issuance the next day, totaling nearly $54 billion, and then issuing $25 billion in dollar bonds in July.

At the same time, Amazon has raised its 2026 capital expenditure plan to $220 billion. AWS's latest quarterly revenue grew by 37%, marking the fastest growth in over four years, but its free cash flow has dropped from $18.2 billion a year ago to -$7.6 billion over the past 12 months.

Meta is also showing similar changes. In April, Meta completed a $25 billion bond issuance; Q2 revenue still grew by 28%, reaching $60.8 billion, but free cash flow plummeted from $8.55 billion in the same period last year to only $784 million, with the 2026 CapEx guidance raised to $130 billion—$145 billion.

These companies have not suddenly lost their ability to generate profits, and they still have plenty of profits; they just have less and less cash available for distribution.

This is why, in the AI era, free cash flow is becoming a metric that is harder to ignore than simple EPS.

2. Borrowing Money: Google and Oracle Are Not the Same Story

However, borrowing itself does not imply danger.

For companies like Alphabet and Amazon, debt is more of a capital structure tool.

They have large core businesses, stable cash flows, and high credit ratings. In the case of extreme front-loading of capital expenditures, spreading construction costs into the future through long-term bonds is a normal form of term matching.

What truly needs attention is when capital expenditures begin to consistently exceed their cash-generating capacity, whether financing is optimizing the balance sheet or starting to put pressure on it.

Oracle is currently one of the most extreme examples.

By the end of the 2026 fiscal year, Oracle's annual capital expenditures reached about $55.66 billion, while operating cash flow was only about $32 billion, resulting in an annual free cash flow drop to -$23.69 billion. Meanwhile, the company has completed approximately $43 billion in debt financing and $5 billion in equity financing for FY2026, with total borrowings amounting to about $130.1 billion as of the end of May.

On July 9, S&P Global Ratings downgraded Oracle's long-term credit rating from BBB to BBB— the lowest level within the investment-grade rating, with the next tier entering speculative grade.

Thus, while both Alphabet and Oracle are burning cash in the AI space, Alphabet is more about utilizing its balance sheet, whereas Oracle is beginning to challenge its balance sheet. This is a new framework that must be established in the next phase of the AI market.

Previously, the market focused on how much AI revenue was growing, how many orders were increasing, and how fast cloud business was growing. In the future, several more questions need to be added, such as how much money the company spent to achieve this growth? How much capital expenditure is needed for every additional dollar of revenue? How much free cash flow is left? How much needs to be borrowed? When interest, depreciation, and lease costs all start to enter the profit statement, how much profit can ultimately be left?

Ultimately, what truly determines valuation will not just be growth itself, but the capital efficiency of that growth.

3. More Important Changes than Issuing Bonds: AI is Becoming a Financial Asset

If the bond issuances by Alphabet, Amazon, and Meta are merely a change in financing methods, then NVIDIA's latest move takes this a step further.

On August 10, NVIDIA announced a strategic partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, aiming to establish an independent computing financing platform to mobilize over $500 billion in third-party capital for AI infrastructure in the long term.

NVIDIA itself has the option to provide up to 25% of potential transactions, which is about $125 billion as a backstop, although the specific contributions and funding deployment timelines from various institutions have yet to be disclosed.

What is truly worth noting is how NVIDIA defines this model. In the official announcement, NVIDIA directly describes AI Compute and AI Factory as a new 'investable asset class'.

The logic is not complicated. In the past, if a Neocloud needed to purchase billions of dollars in GPUs, it first had to raise a large amount of capital itself. In the future, this structure may increasingly resemble traditional infrastructure financing, where data centers and GPUs form assets, clients long-term lease computing power to generate cash flow, and institutions like Apollo, BlackRock, and KKR provide long-term capital, with projects repaying financing costs using future computing power income.

Thus, GPUs are no longer just chips that are sold off; instead, the entire AI Factory, composed of GPUs, data centers, electricity, and long-term computing contracts, begins to be packaged as an infrastructure asset capable of generating long-term cash flow that can be priced and financed by capital markets.

The implications behind this are significant.

Because once AI infrastructure can enter the investment scope of pension funds, insurance capital, private credit, infrastructure funds, and asset management institutions, the capital pool that AI can tap into will no longer be limited to the cash of tech companies.

This could allow this round of AI infrastructure construction to last longer than the market expects.

However, at the same time, it will also change the nature of risk. After all, the biggest risk in the first phase of the AI cycle was that AI was not used by anyone, but in the latest second phase, a more concerning risk may be that AI is indeed being used, but the price of computing power drops too quickly, and revenue growth does not keep pace with debt, depreciation, and financing costs.

Especially when the speed of GPU updates remains rapid, today's asset return models designed for five years or even longer must be based on an important assumption: that these devices can maintain sufficiently high utilization rates and economic value in the coming years.

This is also a new variable that the AI industry will face once financial capital truly begins to enter.

In addition, there is another capital commitment that is easily overlooked.

According to statistics from Reuters on company documents, Microsoft, Meta, Oracle, Amazon, and Alphabet have disclosed about $1.09 trillion in future lease payment commitments that have yet to be executed, including approximately $329.1 billion from Microsoft, $279 billion from Meta, $260 billion from Oracle, $137.2 billion from Amazon, and $85.2 billion from Alphabet; Meta subsequently signed a new data center lease agreement worth approximately $68 billion in July, further raising the known scale to about $1.16 trillion.

Of course, these figures cannot be simply understood as 'tech companies owe $1.16 trillion'. Many contracts span over a decade and have not yet officially begun execution, so they have not all entered the balance sheet as lease liabilities; Amazon's related disclosures also include assets such as warehouses, offices, planes, and vehicles, not all of which belong to AI data centers.

However, it still reveals an important fact: a significant portion of AI infrastructure investment in the coming years has already been locked in advance. As long as computing power demand continues to grow, these commitments will form the basis for future revenue growth. But if model efficiency improves rapidly, unit computing power prices continue to decline, or the pace of enterprise AI commercialization falls short of expectations, then today's locked-in long-term capacity may become fixed costs that are difficult to quickly reduce in the future.

And this is the biggest difference between the AI financing cycle and a simple technology cycle.

Final Thoughts

It is still difficult to simply interpret these changes as negative signals.

On the contrary.

The entry of the bond market, private credit, pension funds, insurance capital, and global asset management institutions may further expand the scale of funds that AI can tap into, allowing this round of infrastructure construction to last longer.

As of August 12, U.S. stocks are still not far from their historical highs. After the latest earnings report, Amazon's stock rose nearly 9% in after-hours trading due to a 37% growth in AWS; Microsoft received a significant market reward after proving the growth of its cloud business and cash-generating ability; meanwhile, Alphabet faced pressure after announcing continued increases in CapEx, and Meta encountered a sell-off after a 91% drop in free cash flow.

The market has not begun to reject AI investments; it is simply transitioning from a pure 'demand trade' to a more stringent 'capital return trade':

• In the first phase, it was about who dared to spend money;
• In the second phase, it is about who can obtain more computing power at a lower capital cost and ensure that every dollar invested ultimately generates sufficient cash flow;

And when AI transitions from capital expenditures of tech companies to assets that Wall Street can purchase, finance, and price, the real question that this competition needs to answer will eventually become 'who builds something that can truly make money?'

And this may be the true dividing line in the next year or two that determines the valuation gaps between Google, Amazon, Meta, Oracle, CoreWeave, and even NVIDIA.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.