Currently, more and more Wall Street institutions are warning that the expansion of US AI debt has begun to put competitive pressure on the US treasury bond market. According to J.P. Morgan's Michael Cembalest estimates, up to 2026, the five largest cloud computing companies plus Nvidia's debt issuance scale is about US$320 billion, of which the long-term portion is equivalent to a ten-year equivalent of about US$303 billion, which is equivalent to 68% of the US Treasury's newly growing long-term loans during the same period. Goldman Sachs fixed income credit trader and investment-grade credit expert Jeffrey Papai warned in the latest report that after a total of about 300 billion US dollars of AI-related bonds have been issued during the year, supply will slow significantly in the fourth quarter of 2026, providing a brief respite for AI credit spreads. However, this easing is extremely short-lived. It is estimated that in 2027, the scale of bond issuance by hyperscale cloud computing companies and chip makers will increase by about 40% compared to 2026, reaching about 340 billion US dollars. At that time, the market will face a greater impact. The core assumptions include: the capital expenditure of hyperscale cloud computing enterprises will be about US$930 billion in 2027; debt financing will rise to 37.5% of capital expenditure from about 30% in 2026; the share of dollar financing will drop from about 80% to 70%; and advanced chip debt will increase from about US$35 billion to US$75 billion in 2026.

Currently, more and more Wall Street institutions are warning that the expansion of US AI debt has begun to put competitive pressure on the US treasury bond market. According to J.P. Morgan's Michael Cembalest estimates, up to 2026, the five largest cloud computing companies plus Nvidia's debt issuance scale is about US$320 billion, of which the long-term portion is equivalent to a ten-year equivalent of about US$303 billion, which is equivalent to 68% of the US Treasury's newly growing long-term loans during the same period. Goldman Sachs fixed income credit trader and investment-grade credit expert Jeffrey Papai warned in the latest report that after a total of about 300 billion US dollars of AI-related bonds have been issued during the year, supply will slow significantly in the fourth quarter of 2026, providing a brief respite for AI credit spreads. However, this easing is extremely short-lived. It is estimated that in 2027, the scale of bond issuance by hyperscale cloud computing companies and chip makers will increase by about 40% compared to 2026, reaching about 340 billion US dollars. At that time, the market will face a greater impact. The core assumptions include: the capital expenditure of hyperscale cloud computing enterprises will be about US$930 billion in 2027; debt financing will rise to 37.5% of capital expenditure from about 30% in 2026; the share of dollar financing will drop from about 80% to 70%; and advanced chip debt will increase from about US$35 billion to US$75 billion in 2026.