According to a research report released by Sinolink Securities, long-end U.S. Treasury yields have surged recently, with both 30-year and 10-year yields climbing to multi-year highs. The core issue driving this rise is the crowding-out effect of AI capital expenditures on long-term capital, as tech giants transition from cash generators to long-term capital demanders, competing with sovereign debt for funds and significantly reducing market absorption capacity.
A secondary factor stems from concerns over Federal Reserve credibility and long-term inflation risks triggered by elevated oil prices. Looking ahead, the continued expansion of AI financing demand, deteriorating global long-term capital supply-demand dynamics, and negative feedback from Japanese asset allocation could force the Fed to adopt accommodative policy and drive gold prices even higher.
Where the Pressure Is Coming From
Long-end Treasuries have once again become the focal point of global markets this week. On August 18, the 30-year Treasury yield surged above 5.32%, reaching its highest level since 2007, while the 10-year yield also broke through 4.7%. The current upward move in long-end rates can be attributed to one primary contradiction, two secondary contradictions, and one gray rhino risk.
The primary contradiction is that massive external financing needs from technology companies have pushed up real interest rates, crowding out sovereign debt. Over the past decade, major U.S. tech firms have been characterized by exceptionally strong cash flows, serving as some of the largest cash generators in the capital markets. However, AI capital expenditure is now fundamentally altering this dynamic.
Market consensus currently projects capital expenditures for the five major hyperscale data center operators to reach $750–800 billion in 2026, climbing further to $1.0–1.1 trillion by 2027. PIMCO previously estimated that 2026-2027 capex would equal approximately 94% of the five hyperscalers' operating cash flow, and with recent upward revisions to investment plans, this ratio may now approach 95%–100%.
The five companies' bond issuance in 2026 is estimated at around $250 billion, equivalent to roughly one-third of their capex, with 2027 issuance potentially rising to $400 billion, representing about 35% of capex. This means tech giants are transforming from cash creators, share repurchasers, and financial asset buyers into long-term capital demanders.
Meanwhile, the U.S. government's fiscal financing needs have not declined, even with a shift toward short-end financing strategies. When the world's highest-credit-quality private enterprises begin competing with the world's largest sovereign debt issuer for long-term capital, crowding-out effects are inevitable.
The Dallas Fed has estimated that if AI-related investment-grade corporate bond issuance reaches $300 billion this year, it could create approximately $360 billion in 10-year-equivalent duration supply, roughly one-eighth the size of U.S. Treasury duration supply. On August 7, when Alphabet announced a $25 billion bond issuance plan, Treasury yields rose 3-4 basis points across the curve that day, with the 10-year climbing to around 4.65% and the 30-year reaching 5.21%. On the same day, the G-spread on Alphabet's previously issued 5.65% bonds maturing in 2056 widened from approximately 95 basis points to 101 basis points, confirming the crowding-out effect on long-duration assets.
Market absorption capacity for AI-related debt is deteriorating at the margin. Among the 91 hyperscale computing enterprise bonds issued so far in 2026, 78 had higher yields at the end of July compared to their issuance levels. In terms of subscription demand, the bid-to-cover ratio for new hyperscaler bonds has fallen from nearly 5 times in February to less than 2 times in July, while new issue concessions have expanded from 2-3 basis points to around 12 basis points. Amazon's dollar bonds issued in March saw approximately 3.4 times subscription, while July issuance attracted only about 1.6 times. Credit default swap spreads and secondary market credit spreads for tech companies have also begun widening again since June.
Fed Credibility and Oil Price Risks
One secondary contradiction is that the predictability of the Fed's reaction function has declined, with long-end rates beginning to price in a "credibility discount." Although the July FOMC meeting kept rates unchanged, Warsh's continued de-emphasis of forward guidance has raised concerns about the predictability of future monetary policy. Consequently, post-meeting Treasury rates moved in a "short-end down, long-end up" pattern, with reduced bets on near-term rate hikes but increased risk compensation for long-term inflation uncertainty and Fed credibility—shifting from pricing "rate hikes" to pricing "Fed credibility."
However, this rate extreme presents an opportunity to test Warsh. When extreme scenarios emerge, markets are eager to see whether a "Fed Put" will materialize and in what form, particularly in the post-Powell era. Before any "Warsh put," markets first witnessed the familiar "Bessent put."
On August 19, at a sensitive juncture when long-end Treasury yields were surging, the U.S. Treasury Department announced it would at least double the liquidity support repurchase scale for 10-20 year and 20-30 year nominal Treasury bonds from September 9 to November 4, raising the single-repurchase cap from $2 billion to at least $4 billion. Based on the previously published schedule, seven long-end repurchases are planned during this period, corresponding to at least $14 billion in additional purchase capacity.
The core purpose of this Treasury repurchase program is to reduce the long-duration supply the market needs to absorb in the short term and improve the supply-demand structure of the long bond market through maturity swaps. If financing is increasingly completed through short-dated or shorter-maturity new issues, this effectively replaces part of the existing long-duration debt with more liquid short-duration debt, which can alleviate long bond liquidity pressure and term premiums in the near term. The market quickly traded this policy backstop—the 10-year yield fell approximately 7 basis points after the announcement, and the 30-year yield dropped nearly 10 basis points.
However, this operation does not change the U.S. fiscal deficit or overall government financing needs; it only alters the maturity structure of debt issuance. Therefore, this resembles more of a phased "peak-shaving" of long-end supply pressure, which can temporarily ease market imbalances but cannot fundamentally reverse the upward pressure on long-term rates stemming from fiscal expansion and increased debt supply.
Another secondary contradiction involves high oil prices, which have reduced short-term rate disturbances but intensified long-term inflation concerns. On August 18, Brent crude rose for a third consecutive day, climbing to a monthly high of $92 per barrel. Unlike previous episodes, markets have not simultaneously raised expectations for near-term Fed rate hikes—in fact, pricing for September hikes actually declined from the prior week. Instead, the greater concern has become that high oil prices could make the inflation path more sticky over the coming years, prompting investors to demand higher inflation risk compensation and term premiums on long bonds.
The Gray Rhino Risk From Japan
Finally, the gray rhino of Japanese government bonds is amplifying risk contagion and sentiment resonance across global long-duration bond markets. Recent U.S. fiscal pressures have re-emerged and entered the market pricing framework for long bonds. Last week, the Congressional Budget Office raised its FY2026 deficit forecast from $1.9 trillion in February to $2.1 trillion, primarily due to Supreme Court rulings causing tariff revenues to fall significantly below earlier projections. Additionally, through the first ten months of FY2026, the U.S. fiscal deficit has already reached approximately $1.798 trillion, exceeding the $1.629 trillion recorded in the same period of FY2025, with further increases likely in the coming two months.
Beyond Treasuries, G10 sovereign bond yields have also been rising broadly. On one hand, monetary policy direction across major economies remains tilted toward tightening. Swap and futures markets pricing policy rates 6-12 months ahead shows higher monetary tightening expectations in Korea, Japan, Canada, Europe, and the UK compared to the U.S. Beyond rate hike expectations, fiscal sustainability concerns are also widespread. For instance, Japanese markets are preemptively pricing in BOJ rate hike possibilities in a weak yen environment, as well as risk pricing for high-government expansionary fiscal policy, while European long-duration bonds face dual pressures from fiscal uncertainty and inflation.
The 30-year Treasury is not an isolated market. Ultra-long bonds and high-grade corporate debt from the U.S., Germany, the UK, and Japan fundamentally belong to the same category of long-duration assets favored by global insurance companies, pension funds, and sovereign wealth funds. Therefore, when risk contagion occurs, the term compensation demanded by investors tends to rise collectively.
Yen depreciation is also viewed as a potential "gray rhino" risk for Treasuries. After the joint U.S.-Japan intervention, the yen appreciated temporarily but quickly weakened again, indicating Japan still faces a thorny policy trilemma—wanting to prevent continued yen depreciation, unable to tolerate excessively rapid domestic rate increases, while maintaining financial system and fiscal stability. Going forward, the likelihood of continued Japanese intervention in currency markets remains high.
Repatriation of Japanese domestic capital could also impact Treasuries. As Japan's risk-free rate continues to rise, the yield advantage of U.S. Treasuries over Japanese government bonds after currency hedging may narrow further. This could trigger a slow structural shift, with the world's largest overseas holder of U.S. Treasuries seeing reduced marginal allocation demand for long-end U.S. debt. Japan's holdings of U.S. Treasuries fell approximately 2.3% month-over-month to $1.116 trillion in June.
When the Fed's control over long-end rates diminishes and overseas demand for U.S. Treasuries weakens at the margin, quantitative easing may be the last resort—which partially explains the recent rally in gold prices.
What Lies Ahead
Looking forward, the Bessent Put and the upcoming Jackson Hole meeting provide a near-term window for long-end rates to repair, but given fiscal supply pressures and the Fed's credibility discount, the sustainability of such repair should not be overestimated. The key focus is that AI capital expenditure scale and corresponding financing growth will likely increase further. On this basis, the global supply-demand relationship for long-term capital may continue to deteriorate, and Japanese long-end yields, the yen, and Japanese institutional overseas asset allocation could form new negative feedback loops. All these factors are pushing toward a more accommodative Fed and higher gold prices.
Comments