AI wants money, and Western governments want money too! The global "capital battle" has begun, and the bond storm has "just started"

In the past two years, the market's most crowded narrative has been AI: computing power, chips, data centers, cloud services—every asset is being revalued around artificial intelligence. But now, a more fundamental issue has begun to surface: AI doesn’t just need computing power, it needs money; and what it requires is massive, long-term, and low-tolerance-for-error capital.

Meanwhile, the US government, European governments, and the Japanese government are all seeking money. Soaring fiscal deficits, debt rollovers, and skyrocketing interest expenditures are pushing the global bond market into a new phase: capital is no longer cheap, nor abundant.

This means that the capital market logic established over the past decade, based on "low interest rates, ample liquidity, and unlimited financing capability," is being rewritten.

The real risk is not just that a single AI company's debt issuance costs are rising, nor only that the US deficit has exceeded $2 trillion. Behind these surface-level stories, two enormous financing demands are converging on the same market: on one side is the AI super capital expenditure cycle, on the other is the government fiscal deficit cycle.

Behind this storm is a key variable: Japan. The world had become accustomed to cheap Japanese money, but now must pay a higher price to compete for this capital. As domestic interest rates rise in Japan, Japanese investors are becoming more selective—French government bonds were the first to be sacrificed, followed closely by US Treasuries.

When these forces collide, the most likely outcome

is a systemic rise in the global cost of capital, a forced “rationing of demand” in the bond market, a subsequent impact on capital markets, and ultimate transmission to the real economy.

AI Financing Enters the “Massive and Expensive” Stage

For the past two years, market discussions about AI have mostly focused on chips, models, applications, and valuations. But as AI competition moves into deeper waters, the variable that determines victory is capital spending capacity. Training and deploying large models require high-end chips, servers, networking equipment, cooling systems, land, electricity access, and vast data centers. Each category means heavy asset investment, which occurs even before revenue is realized.

This has quickly transformed AI from a technological revolution into a wave of global “financing revolution” for infrastructure.

Data shows that the five largest hyperscale AI data center operators have issued about $220 billion of debt this year, more than double last year’s figure. Wall Street expects AI-related capital spending to rise from about $700 billion this year to nearly $1 trillion next year. Over the past two years, companies have borrowed about $625 billion for data center and AI investments, and Wall Street projects the total construction cycle will cost close to $5 trillion by the end of 2030.

The pressure on financing is already showing up in the cash flows of tech giants: Alphabet reported its first ever negative free cash flow in the second quarter, burning $5.9 billion and raising its 2026 capital expenditure forecast to as high as $205 billion; Amazon has lifted its 2026 capex plan to about $220 billion, with its free cash flow turning negative by $7.6 billion in the past twelve months.

Of even greater concern is that the AI financing curve is trending higher, with lower-rated AI-related borrowers already bearing noticeably higher capital costs. For smaller, lower-credit-rated “new cloud” companies like CoreWeave, Lambda, and Crusoe, the high-cost leveraged loan market is their only option.

A typical example is Volta Infrastructure Holdings. For its Norwegian data center project, the company offered a $5 billion leveraged loan, with market quotes showing a blended financing yield close to 11% at one point.

For a company in expansion, still needing to continually invest in data centers and computing equipment, this is an extremely high profit threshold. It means a project must not only be completed and fully operational, but also generate sufficiently high and stable cash flows over a long period of time to cover interest, depreciation, and ongoing refinancing pressure. The AI debt market has already shifted from the stage of “capital willing to pay” into “capital demanding high compensation.”

According to Goldman Sachs, about $88 billion of low-rated AI-related borrowing has already entered the market this year. AI financing is gradually moving down the credit spectrum: the highest-quality companies issue investment-grade bonds; those with slightly weaker credit depend on leveraged loans; and for even weaker borrowers, double-digit financing costs have become the norm.

This is precisely the signal of the beginning of a stratified credit cycle.

AI may not be a bubble, but conditions for AI financing are no longer cheap.

The US Government Is Also Desperately “Seeking Money”

If only the AI sector’s financing was expanding, the market might still be able to digest it. The real trouble is that the world’s largest borrower—the US government—is also absorbing capital at a staggering pace.

According to the Congressional Budget Office, for the fiscal year ending September 30, 2026, the US federal budget deficit rose to $1.993 trillion, up 12% year-over-year and the highest level since 2021. Federal spending reached $7.4 trillion while receipts were $5.4 trillion. The deficit as a percentage of GDP is expected to exceed 6%.

Historically, US deficits above 6% have typically occurred during periods of war, financial crisis, or economic recession. But at present, the US economy is not in a recession and has been expanding for years. This means that US fiscal policy has entered a crisis-like deficit state during a non-crisis period.

Even more difficult is that the deficit is creating a self-reinforcing effect. US net interest expenditures have surpassed $1.1 trillion, accounting for over one-fifth of all tax revenue and now exceeding defense and Medicare spending. High interest rates raise refinancing costs, interest expenses devour tax revenues, and the deficit expands further, which requires additional debt issuance—forming a dangerous cycle.

It is worth noting that the recent rise in US bond yields has not yet been fully reflected in the budget. Higher interest rates will only gradually enter the fiscal accounts with new debt and refinancing of existing bonds, meaning that fiscal pressure will continue to be released with a lag.

Beyond the two major financing waves, the global bond market also faces a long-underestimated variable on the demand side: Japanese investors are becoming more selective.

Data shows that as of July, Japanese investors held about 23 trillion yen ($145 billion) of French government bonds, the largest overweight position in the eurozone. But as Japan’s 10-year government bond yield exceeded 3% last month, hitting a 30-year high, hedged for currency, the yield advantage of French 10-year government bonds over Japanese government bonds has narrowed to just about 40 basis points, drastically reducing their appeal to Japanese holders.

With the yield advantage vanishing and French fiscal conditions continuing to deteriorate, Japanese capital is starting to move away. Japanese holdings of French bonds have dropped 2.5% since last year. The global fixed income team at Sumitomo Mitsui DS Asset Management, headed by Shinji Kunibe, has liquidated its French bond holdings out of fiscal concerns. Mizuho’s Masayuki Nakajima said: “Even if valuations appear cheaper, Japanese investors’ motivation to rebuild positions is decreasing.”

US Treasuries are similarly affected. In the first half of this year, Japanese domestic investors were net sellers of $4.45 trillion yen (about $28 billion) in US long-term Treasuries, the first half-year net selloff since 2022, while massively increasing holdings of local government bonds, with retail purchases hitting a new high since 2014.

Reidy points out, Japan doesn’t need to “dump” for US borrowing costs to rise—“buying less is enough.” The world has become used to Japanese cheap capital, but must now pay a higher price to compete for it.

The Price of AI and Governments Competing for the Same Capital Pool—Rising Interest Rates

JPMorgan CEO Jamie Dimon gave the most direct assessment: "Rates are rising, capital demand is robust, and government borrowing is substantial."

AI needs money, and so do governments. Corporations want to build data centers, governments want to cover deficits. The private sector wants financing to buy GPUs, the public sector needs financing to pay for welfare, interest, and national defense. Everyone is looking for capital in bond, loan, and private credit markets.

When capital demand surges simultaneously, unless savings supplies increase sharply as well, the only possible result is higher prices.

And the price of capital is the interest rate.

Dimon does not deny the long-term value of AI. He even believes AI will, like the internet, eventually create massive productivity gains. But he emphasizes that the issue isn’t simply whether AI has value; it’s the cost of funding AI infrastructure, the construction cycle, regulatory resistance, legal challenges, power bottlenecks, and changes in technology paths that matter.

In other words, AI could be correct in the long run, but short-term financing costs could truly be high.

This is exactly what the market currently tends to overlook: a long-term correct industry trend can also create serious short- to medium-term capital market stress.

The internet ultimately changed the world, but the capital bubble around 2000 still burst. Railroads ultimately changed the United States, but in the 19th century, railroad debts triggered multiple financial crises. Infrastructure revolutions and capital market turbulence are not mutually exclusive—indeed, they often occur simultaneously.

In addition, Dimon also warned that US inflation carries structural risks of rising. Facing rearmament, immigration policy, and supply chain restructuring, he bluntly noted that there is a risk that inflation could remain “sticky” and rates may keep rising.

A Massive Capital Squeeze May Already Be Underway

Global capital tightening is shifting from macro pressure to market damage.

As the competition for capital intensifies between sovereign financing and AI financing needs, European bank stocks have plunged; on October 9, the SX7E European banking sector index abruptly fell by about 4% intraday, breaking below its 100-day moving average. French assets are also being repriced; the spread of Italian and Spanish 10-year government bonds over French government bonds has shrunk sharply, even turning negative.

Goldman Sachs analyst Privorotsky believes that beneath the surface of market moves a much deeper problem is brewing.

In the post-financial crisis era, the global economy depended on ample savings and structurally low interest rates. National fiscal systems were also built on the assumption of cheap capital existing for the long term. Now, this assumption is breaking down.

Sovereign financing needs remain enormous, and artificial intelligence is creating an unprecedented demand for capital. For example, SpaceX is reportedly seeking $40 billion in financing, while Broadcom is exploring a financing program of more than $50 billion. These massive demands will compete for the same pool of funds as government bonds and other corporate financings.

If AI investments can produce internal rates of return of 20–30%, then borrowing at near-double-digit interest rates remains reasonable. But then, what about other sectors that also need capital? This may explain why real yields are hard to drive lower. If the market expects a large wave of high-quality private sector bonds to come to market, the appeal of buying sovereign debt falls.

Privorotsky offers two possible outcomes: First, AI gradually achieves self-sustaining financing, and the funding pressure eases; second, capital becomes rationed in other areas and the more fragile segments of the economy begin to break down.

CCC-rated credit spreads may already be sending early warning signals. If the weakest credit tier continues to be under pressure, it means that capital tightening is no longer just a rates market issue,

but is being transmitted to companies and assets with weaker financing capacities.

Who Gets “Crowded Out” First?

In a recent appearance on the The Julia La Roche Show, Dalio described this structure as a debt-induced "heart attack": when the cost of capital rises enough to weed out marginal borrowers, the market enters a passive and brutal "rationing of demand."

On the supply side, the pressure is quantifiable: the US federal government spends about $7 trillion annually against about $5 trillion in tax revenue, creating a $2 trillion deficit that must be patched by ongoing bond issuance—and a significant portion of the fiscal room is already eroded by sky-high interest payments. Meanwhile, corporations are borrowing heavily to fund massive AI data center capital expenditures. The combination means the supply side of bonds faces ongoing shock.

The demand side, however, is unstable. Foreign investors hold about one-third of US Treasuries, but Japan and China, the two main holders, are both reducing their allocations. If traditional buyers continue to step back, the bond market must offer higher yields to attract new buyers or force some borrowing demand to exit the market.

Dalio believes that government and AI behemoth financing expansion pushes up the risk-free rate and credit spreads; lower- and mid-tier credit issuers are then crowded out; real estate, consumption, manufacturing, and small-to-medium-sized business investment weakens; credit risks surface and, ultimately, pressure is transmitted to the real economy. Mortgage borrowers and ordinary homeowners may be the first to be forced out, with broader capital markets feeling the strain only later, and finally delivering a heavy blow to the real economy.

Capital Has a Price, and Growth Comes at a Cost

The long-term value of AI is not in doubt. As Dimon said, AI could create massive value, even fundamentally changing productivity and social structures like the internet did.

But “AI is valuable” does not mean “any price of AI investment is reasonable”; and “the government can issue debt” does not mean “debt can expand forever without consequences.”

When building AI data centers requires money, the US government needs trillions to fill its deficit, European fiscal repair chews up more capital, and Japan’s normalization of interest rates is sucking back domestic funds, the competition in global capital markets will inevitably intensify.

Rising interest rates may not mean the economy collapses immediately; but it does mean capital will no longer be universally available, financing will no longer be cheap, and asset prices can no longer be built atop expectations of “eternal easing.”

The most profound change in the bond market may not be that yields hit a specific level, but that markets are rediscovering a simple fact: money is not infinite.

When capital is forced to be rationed, the first to be hit are marginal assets and weak credit entities; next is the capital market in general; and, in the end, the pressure transmits to jobs, consumption, real estate, and real investment.

The global “capital war” is already underway, and the bond market storm may have only just begun.