Sinolink: AI capital expenditure squeezes sovereign debt, and the rise in long-term U.S. Treasury rates may force the Federal Reserve to ease policy.

Sinolink has released a research report stating that long-term U.S. Treasury yields have recently surged significantly, with 30-year and 10-year yields both rising to multi-year highs. The core issue driving this upward trend is the crowding-out effect of AI capital expenditure on long-term capitaltechnology giants are shifting from cash generators to long-term capital demanders, competing with sovereign debt for capital, which has significantly reduced market absorption capacity. A secondary issue arises from the discount on the Federal Reserve's credibility and long-term inflation concerns triggered by high oil prices. Looking ahead, the continuous expansion of AI financing demand, the deterioration of global long-term capital supply and demand, and negative feedback from Japanese asset allocation may compel the Federal Reserve towards easing and further increases in gold prices. Sinolink's main points are as follows: This week, long-term U.S. Treasuries have once again become the focus of global markets. On August 18, the yield on 30-year U.S. Treasuries briefly rose above 5.32%, hitting a new high since 2007, while the 10-year yield also surpassed 4.7%. We believe this phase of rising long-term rates can be summarized with one primary contradiction, two secondary contradictions, and one gray rhino. The primary contradiction is that the immense external financing demands of technology companies are driving up real interest rates, thus exerting a crowding-out effect on sovereign debt. For the past decade and more, a significant characteristic of large U.S. technology companies has been their abundant cash flow; they have been one of the largest cash generators in the capital markets. However, AI capital expenditures are now changing this dynamic. The market currently widely expects that capital expenditures by the five major hyperscale data center operators will reach $750 billion to $800 billion by 2026, further climbing to $1 trillion to $1.1 trillion in 2027. PIMCO previously projected that capital expenditures for 2026-2027 would account for about 94% of the operating cash flow of the five hyperscalers, and with recent adjustments to investment plans, this ratio may now be approaching 95%-100%. The bond issuance scale for these five companies in 2026 is about $250 billion, which is roughly one-third of their capital expenditures; bond issuance in 2027 may further rise to $400 billion, accounting for about 35% of capital expenditures. This shift implies that technology giants are transforming from cash flow creators, stock repurchasers, and financial asset buyers into long-term capital demanders. Meanwhile, the financing demands of the U.S. government have not decreased, even as it has shifted to a short-term financing strategy. When the best-quality private firms in the world start competing for long-term capital with the largest sovereign debt issuer globally, it inevitably leads to a crowding-out effect. The Dallas Federal Reserve has estimated that if investment-grade corporate bond issuance related to AI reaches $300 billion this year, it could generate approximately $360 billion of 10-year equivalent duration supply, which is about one-eighth of the supply of U.S. Treasury durations. On August 7, when Alphabet announced a $25 billion bond issuance plan, U.S. Treasury yields rose across the board by 3-4 basis points, with the 10-year yield approaching 4.65% and the 30-year rising to 5.21%; on that day, the G-spread of Alphabet's previously issued 5.65% bonds maturing in 2056 widened from around 95 basis points the previous day to 101 basis points, confirming the presence of a crowding-out effect for long-duration assets. The market's absorption capacity for AI bonds is marginally decreasing. Among the 91 super-large-scale computing enterprise bonds issued so far in 2026, 78 had a yield to maturity at the end of July that was higher than its issuance level. Subscription data indicates that the bid-to-cover ratio for new hyperscaler bonds has dropped from nearly 5 times in February to less than 2 times in July, and the new issuance concession has expanded from 2-3 basis points to around 12 basis points. Amazon's dollar bonds issued in March had a subscription ratio of approximately 3.4 times, while July's issuance only achieved about 1.6 times. The credit default swap (CDS) spreads and secondary market credit spreads for technology companies have also widened since June. One of the secondary contradictions is the decreasing predictability of the Federal Reserve's response function, leading long-term rates to factor in a "credibility discount." Although the Federal Open Market Committee (FOMC) meeting in July kept interest rates unchanged, the continued weakening of forward guidance by Warsh raised market concerns regarding the predictability of future Federal Reserve monetary policy. Thus, in the aftermath of the meeting, Treasury rates moved in a "short-end down, long-end up" manner: reducing bets on short-term rate hikes while simultaneously increasing risk premiums for long-term inflation uncertainty and Federal Reserve credibilityshifting from pricing in "rate hikes" to pricing in "Fed credibility." However, this interest rate extreme presents an opportunity to assess Warsh. When faced with extreme scenarios, will the Fed put materialize, and in what form? The market is eagerly awaiting Warsh's expressions and stabilizing attitude, which are particularly important in the post-Powell era. Before the "Warsh Put," the market first observed the familiar "Bessent Put." On August 19, at a sensitive point when long-term Treasury yields had significantly risen, the U.S. Treasury announced that from September 9 to November 4, it would at least double the liquidity support repo scale for nominal Treasury bonds of 10-20 years and 20-30 years, increasing the per-repo cap from $2 billion to at least $4 billion. According to a previously announced schedule, there will be seven long-end repos during this period, providing at least $14 billion in additional purchasing capacity. The core of this Treasury repo operation is to reduce the amount of long-duration supply that the market needs to absorb in the short term while improving the supply-demand structure in the long bond market through maturity swaps. By increasing the repo for 10-30 year Treasury bonds and completing financing more through short bonds or shorter-term new bonds, it effectively replaces some existing long-duration bonds with more liquid short-duration debt, temporarily alleviating liquidity pressures and term premiums in the long bond market, having a stabilizing effect. Following the announcement, the yield on 10-year U.S. Treasuries temporarily declined by about 7 basis points, and the yield on 30-year Treasuries dropped by nearly 10 basis points, indicating that the market quickly responded to this policy support. However, this operation did not alter the U.S. fiscal deficit or the government's overall financing needs; it merely changed the term structure of debt issuance. Therefore, it is more like a phase-based "peak shaving" of long-term supply, which can balance the market in the short term but is unlikely to fundamentally reverse the upward pressure on long-term rates caused by fiscal expansion and increased debt supply. Another secondary contradiction is that while high oil price risks have reduced short-term interest rate disturbances, long-term inflation concerns have intensified. On August 18, Brent crude oil prices rose for the third consecutive day, reaching a monthly high of $92 per barrel. Unlike before, this round of market conditions did not lead to a significant increase in expectations for short-term rate hikes by the Federal Reserve; rather, market pricing for the September rate hike has decreased compared to the previous week. However, greater concern has shifted to the notion that high oil prices could make the inflation path more persistent over the ensuing years, leading investors to demand higher inflation risk compensation and term premiums from long bonds. Finally, the "gray rhino" of Japanese bonds has amplified the risk contagion and emotional resonance in the global long bond market. Recently, U.S. fiscal pressures have resurfaced and entered the pricing framework for long bonds. Last week, the Congressional Budget Office (CBO) raised its deficit forecast for the fiscal year 2026 from $1.9 trillion in February to $2.1 trillion, primarily due to a Supreme Court ruling leading to significantly lower tariff revenues than previously estimated. Moreover, as of the first ten months of fiscal year 2026, the U.S. fiscal deficit has already reached approximately $1.798 trillion, exceeding $1.629 trillion during the same period of the fiscal year 2025, and is likely to continue rising in the next two months. In addition to U.S. Treasury yields, recent yields on G10 sovereign bonds have also been climbing. On one hand, monetary policy directions from major economies are generally tightening. Swap and futures markets indicate that the policy rate expectations for the next 6-12 months are higher in South Korea, Japan, Canada, Europe, and the UK than those for the U.S. Aside from interest rate hike expectations, there are also widespread concerns regarding fiscal sustainability. For instance, the Japanese market is speculating on the possibility of BOJ rate hikes in a weak yen environment and the risks associated with expansive government fiscal policies, while Europes ultra-long bonds are facing dual pressures from fiscal uncertainties and inflation. The 30-year U.S. Treasury market is not isolated. The ultra-long bonds and high-rated corporate bonds of countries such as the U.S., Germany, the UK, and Japan are fundamentally attractive long-duration assets favored by global insurance companies, pension funds, and sovereign wealth funds. Therefore, when risks spread, the term premiums demanded by investors are likely to rise collectively. The depreciation of the yen is also regarded as a potential "gray rhino" risk for U.S. Treasuries. The yen briefly strengthened after the U.S.-Japan joint intervention but quickly weakened again, indicating that Japan is still facing a particularly tricky policy trianglewanting to prevent the yen from continuous depreciation while also not bearing excessively rapid domestic interest rate rises, all while needing to maintain the stability of the financial system and fiscal integrity. Therefore, Japan is likely to continue its demand for currency market interventions. Further repatriation of Japanese domestic funds could also impact U.S. Treasuries. As Japans risk-free interest rates continue to rise, the yield advantage of U.S. Treasuries over Japanese bonds may further diminish after currency hedging. This could bring about a gradual structural change, with the worlds largest foreign holder of U.S. Treasuries reducing its marginal allocation demand for U.S. long bonds. The amount of U.S. Treasuries held by Japan had decreased by approximately 2.3% month-on-month to $1.116 trillion in June. When the Federal Reserve's control over long-end rates diminishes, and overseas demand for U.S. Treasuries weakens marginally, quantitative tools (QE) may become the last resortthis also explains part of the recent rise in gold prices. Looking ahead, the Bessent Put and the upcoming Jackson Hole meeting provide a short-term window for the repair of long bond rates, but in the context of fiscal supply pressures and the credibility discount of the Federal Reserve, sustainability should not be overestimated. The key point is that the scale of AI capital expenditures and the corresponding financing growth are likely to further increase; on this basis, the global long-term capital supply-demand relationship may continue to deteriorate; Japanese long bond yields, the yen, and Japanese institutions overseas asset allocations may also generate new negative feedback. All these factors are pushing the Federal Reserve to adopt a more accommodative stance and could result in higher gold prices.