Over $5 trillion! JPMorgan: Global AI infrastructure is "unprecedented in scale" and will impact all capital markets

```

Author: Bao Yilong

Source: Hard AI

JPMorgan has issued a warning: the $5 trillion AI feast will "drain" every credit market.

On November 10, a team of JPMorgan strategists led by Tarek Hamid released a major report on the financing needs for AI data centers. The report emphasized that at least $5 trillion will be needed for AI data center construction over the next five years—and the amount could reach as high as $7 trillion.

This massive capital influx will drive further accelerated growth in the bond and syndicated loan markets. The scale of financing required means that no single financing market can "swallow it all" alone.

The report predicts that over the next five years, investment-grade bond markets will need to provide about $1.5 trillion, leveraged finance markets about $150 billion, and data center asset securitization can handle at most $30–$40 billion per year. Nevertheless, there remains a huge shortfall of $1.4 trillion, which will need to be filled by private credit and even government funding.

(Sources of funds for AI/data centers)

A Capital Feast

The research report points out that global construction of AI and data centers will be a "remarkable and sustained capital market event".

The baseline forecast from the report shows that from 2026 to 2030 alone, the world will need an additional 122 GW of data center infrastructure capacity. A more optimistic forecast based on semiconductor orders suggests the scale of growth over the next three years may reach 144 GW.

(The relationship between data center installed capacity and annual capital expenditure)

However, this feast faces "hard constraints" in the physical world, the biggest bottleneck being electricity supply.

The delivery cycle for new orders of natural gas turbines has been extended to 3-4 years, while nuclear power plant construction cycles are over ten years.

Balancing the need to meet new electricity demand and keeping residential electricity prices reasonable will become a sensitive political and economic issue in the US. While these physical constraints slow construction, they also mean the capital demand would be even larger without such limitations.

(Average price of residential electricity in the US)

Sources of Trillions in Funding

Such massive capital expenditures are far beyond what any single market can bear. JPMorgan believes the future will be shaped by a financing pyramid constructed by multiple capital markets, each layer playing an irreplaceable role.

(Capex growth trends for hyperscale cloud service providers)

First, the cornerstone is the cash flow of tech giants.

Large tech companies generate over $700 billion in annual operating cash flow, of which nearly $500 billion is reinvested as capital expenditure. JPMorgan assumes some $300 billion of annual cash flow will be invested directly into AI and data centers.

Next, the main force is the investment-grade bond market, with public bonds shouldering most of the financing "burden".

JPMorgan expects the investment-grade bond market to absorb about $300 billion in AI-related bonds in the coming year, and reach a cumulative $1.5 trillion over the next five years.

(Most hyperscale cloud firms are rapidly increasing debt levels)

Currently, AI and data center-related industries account for 14.5% of the JULI index—surpassing the US banking sector. By 2030, this share may exceed 20%.

The JULI index is JPMorgan's benchmark for the performance of investment-grade, dollar-denominated corporate bonds. Simply put, it's an important indicator tracking the overall market for high-quality US corporate bonds.

Next come supporting forces: leveraged finance and securitization markets.

Leveraged finance markets (high-yield bonds and leveraged loans) are capable of providing about $150 billion over the next five years. But the report issues a historic warning: in the 1990s, when telecom became high-yield's largest segment, the sector collapsed; from 2010–2015, energy's expansion also ended poorly.

(Tech now accounts for 16% of the leveraged loan index and 7% of the high-yield bond index.)

The securitization market is the "natural home" for data center financing, expected to absorb $30-40 billion in risk capital each year. However, currently its main function is to provide construction-phase financing, not permanent capital, which limits its role to some extent.

(Data center-related securitized products grew 83% year-on-year and now represent 5% of total ABS and CMBS issuance)

Finally, attention must be paid to the crucial "gap-fillers": private credit and alternative capital.

After tapping all public markets, JPMorgan estimates there is still a $1.4 trillion funding gap. This gap will mostly be filled by private credit and alternative capital.

The private credit market has a pool of approximately $466 billion, and its structural flexibility allows it to tailor solutions for complex projects.

Meta recently completed a $27.3 billion private financing via an instrument called "Beignet Investor LLC", a typical innovative example that cleverly shifts construction funding and long-term lease obligations off the balance sheet.

(Organizational structure of Beignet Investor LLC)

Echoes of Past Manias: Lessons from the Telecom Bubble

The report compares the current AI craze to the telecom and fiber-optic network bubble of the early 2000s:

Similarities:Back then, markets believed internet data would "double every 100 days," leading to massive investment in fiber networks. Many financially weak dot-coms and highly leveraged telecom builders took part.Today, expectations for exponential AI compute demand are fueling a similar investment frenzy.

Key Differences:The telecom bubble collapsed because "the revenue curve couldn’t keep up with the investment curve." End-users and enterprises were slow to adopt high-speed networks, resulting in bankruptcies like Global Crossing and a sector crash.Today's tech giants—Amazon, Google, Meta, Microsoft—have powerful free cash flow and balance sheets, making them unlike the weak dot-coms of that era.

Nevertheless, the report sees this history as a strong warning: no matter how attractive the technology's prospects, if it can't be turned into real revenue, massive capital outlays may ultimately lead to nothing but losses.

Winner-Takes-All and Inevitable Losers: The Ultimate Investment Risk

JPMorgan concludes the report with two core risks that will determine the final outcome of this capital feast.

The first is monetization risk. JPMorgan calculates that to earn a 10% annual return on AI investments over the next five years, $650 billion in new annual revenue will be needed.

This is equivalent to 0.58% of global GDP, or every iPhone user worldwide paying $34.72 per month. While enterprises and governments will pay most of it, the sheer scale reveals how tough monetization will be.

The second is disruptive technology risk. Analysts note that if an investment boom relies on a particular technology path, it becomes highly sensitive to breakthroughs in efficiency.

The report cites the "DeepSeek moment"—when a startup claimed to match top model performance at ultra-low cost, briefly triggering market panic over the risk of pricey GPU investments becoming obsolete "dark fiber" (already-laid fibers that went unused).

JPMorgan’s report makes clear: the wave of AI infrastructure building is now irreversible, and will inject unprecedented vitality into capital markets.

However, in this $5 trillion-plus gamble, not all participants will win. The "winner-takes-all" nature of the AI ecosystem means that—even in the most optimistic monetization scenario—some companies will inevitably be eliminated.

For investors, understanding capital flows, identifying companies with true "moats" and monetization capability, and maintaining a healthy respect for historical bubbles is key to winning in this great era of opportunity and risk.

This article comes from WeChat Official Account "Hard AI". For more cutting-edge AI insights, click here.

Risk DisclaimerThe market has risks, investment requires caution. This article does not constitute personal investment advice and has not taken into account individual users' specific investment goals, financial situations, or needs. Users should assess whether any opinions, views, or conclusions in this article are appropriate for their particular circumstances. Investment based on this article is at your own risk. ```