The global bond market has been repeatedly sold off, with yields approaching the highest levels of this century, until a U.S. Treasury auction on October 8, where the winning bid rate was very high, but overseas demand was strong, stimulating capital inflows to buy the dips, causing the 10-year Treasury yield to drop sharply by 10 basis points. Despite bond market turmoil, U.S. stocks continued to rise, led by AI-related stocks, but the market remains concerned about the impact of high interest rates on the banking sector, corporate financing, and the real estate market.
FOMC meeting minutes revealed that U.S. policymakers still intend to raise interest rates, and futures markets expect a second rate hike in December. The Bank of Japan's governor stated that core inflation has reached target levels, suggesting no rate hike in October. France witnessed large-scale street demonstrations, and the Spanish government announced early elections, causing the euro to continue weakening last week amid bond market volatility. Houthi militants attacked an airport in Saudi Arabia, pushing up Brent crude oil prices. U.S. West Texas crude prices remained relatively stable, but supply and demand imbalances persisted in oil products such as diesel, and natural gas prices surged. Amid market and geopolitical turmoil, gold prices rebounded last week.
Last week, the U.S. Treasury auctioned $39 billion in 10-year Treasury bonds, with a winning bid rate of 5.3%, the highest since November 2000, and oversubscribed 2.77 times, significantly exceeding the average of 2.54 times in the past five auctions. Overseas demand was particularly strong, with market analysis suggesting that overseas sovereign funds, insurance funds, and pension funds are re-entering the market, receiving 80.3% of the total allocation from underwriters. This was a remarkably strong auction and subscription, with high-net-worth funds beginning to discuss whether U.S. long-term bond yields are approaching a peak, and the pace of entry clearly accelerating.
First, this auction dispelled the weak sentiment seen over the past few months, as domestic and overseas real-money investors perceive yields are starting to become relatively attractive. Second, conflicts in Iran, Saudi Arabia, and the Russia-Ukraine war continue, creating significant uncertainty, yet capital appears willing to enter, demanding sufficient yield levels to compensate for risk premiums. Third, the U.S. fiscal situation is not optimistic, but other major developed nations are similarly struggling, with even weaker risk resilience, prompting capital to flow into dollar-denominated assets for safe-haven purposes.
This auction shows that there is demand for U.S. Treasuries, but prices must reflect premiums arising from fiscal and inflation risks. Yields around 5.3% are beginning to attract some funds to increase their U.S. bond allocations. Note, however, that this represents only a portion of available capital, not all. The chief investment officer of the giant bond fund PIMCO is discussing the risks of 10-year yields rising to 6%. Nevertheless, the market is no longer uniformly bearish, and the tug-of-war between bulls and bears will determine the future direction of the Treasury market and the timing of yield peaks.
Although the auction revealed demand from certain funds, especially large overseas funds, it has not resolved the awkwardness on the supply side. The U.S. fiscal situation shows no sign of fundamental improvement, and the government appears to lack even the will to improve it. Bond supply imbalances could continue to push up issuance costs. Meanwhile, inflation improvement is progressing slowly, and the U.S. Federal Reserve estimates further rate hikes will still be needed, which remains a significant uncertainty for the bond market.
Bond market volatility has so far had limited impact on U.S. stocks. First, major tech companies' profitability is strong. Second, the tech sector is less sensitive to interest rate cycles. Third, global cash is afraid of missing out on AI-related concepts. In reality, U.S. stocks outside of AI, including bank stocks, have already been deeply hurt by high interest rates, though this is not reflected in stock indices. As for AI companies issuing bonds or raising funds in private markets, the damage from high interest rates is already evident, as movements in risk-free assets inevitably trigger chain reactions across other assets.
The author believes that real demand is gradually emerging, and U.S. long-term bond yields have entered a phase of searching for a peak. However, the exact level of the peak still depends on the trajectory of inflation and the Federal Reserve's stance. Attention should be paid to the process of peaking and the spillover risks that high U.S. bond yields pose to stock markets and overseas bond markets.
Turning to differences between China and the U.S. in the AI sector. The U.S. AI industry is more driven by market-based capital, thus capital returns and commercialization pathways are subject to stronger market constraints, making heavy investment, closed-source weights, and closed-source models the prevailing industry norms. In China, AI industry funding sources are more diverse, with local government industrial funds and state-owned capital platforms playing significant roles alongside internet platforms and private capital. Most Chinese AI models adopt open weights and open-source models, focusing more on the social and inclusive value of the AI revolution, with funding levels far below those of their U.S. counterparts.
There are significant directional differences between the two countries in computing power utilization. The U.S. focuses more on R&D, while China emphasizes monetization through applications. According to public API token consumption data, over half of token usage in the U.S. is for programming development and enterprise automation, whereas in China, about half of token usage is for short videos, short dramas, and visual content production, with a significantly lower proportion dedicated to technological R&D compared to the U.S. The U.S. is brewing faster and larger technological breakthroughs, while China attempts to feed back AI industry development through application monetization, while closely following technological leaders.
Based on the author's subjective assessment of frontier model capabilities and product iteration pace, the gap between the two was about six months at the beginning of the year, narrowed to three to four months mid-year, and has recently widened again to about six months. The main reason for the widening gap lies in computing power; U.S. export restrictions on AI hardware have hindered China's expansion of computing capacity, while China is already on par with the U.S. in talent, algorithms, and data. In fact, differences in model parameters are no longer important—sufficient capability is enough. Competition is extending to the application layer, where profitability certainty matters more.
In AI applications, both China and the U.S. face similar bottlene游戏副本s. The low-hanging fruit, such as programming benefits, have already been harvested, but progress in enterprise adoption has been slower than initially expected. Enterprises feel uncertain about scenario setting and value confirmation in AI transformation, and friction is high when integrating transformation with existing architectures and processes. Each industry and each enterprise presents non-standardized application scenarios. The author expects a gradual and iterative process of 5 to 10 years. Users seem to find existing products sufficient, and most are unwilling to pay higher fees for new products.
In building data centers, the U.S. annual capital expenditure is about six times larger than China's, though the gap in investment efficiency is much smaller. Chinese tech giants also face a situation of nearly negative free cash flow this year, though pressure from funding gaps appears temporarily manageable. China's capital markets lack the depth and liquidity of U.S. markets, but local state capital, telecom operators, and the banking system offer financing channels different from U.S. capital markets, somewhat alleviating funding pressures on AI companies.
The endgame of the AI revolution is power supply. This year, the U.S. has entered an era of power shortages, while China's power supply remains abundant. China's annual new installed capacity is equivalent to about half of the total installed capacity across the entire U.S., while the U.S. faces dual constraints of insufficient power generation and transmission limitations. China's 46 ultra-high-voltage transmission lines efficiently and cheaply convert and transmit wind and solar power generated in the west to coastal regions, an advantage the U.S. cannot match.
There are few major data releases this week. U.S. September CPI is expected to show a technical rebound, with forecasts of a 0.6% month-on-month increase and a 3.6% year-on-year increase, core CPI rising 0.2% month-on-month and 2.5% year-on-year, slightly higher than the previous period. Watch the Fed's Beige Book, U.S. retail data (expected to weaken significantly), and China's monthly data. Tao Dun
*This week's article expresses the author's understanding and views on the economy, policy, and markets, representing personal opinions only, and is not investment advice or solicitation.
*Articles published in ET, signed or unsigned, represent the authors' personal opinions and do not reflect ET's position. ET serves as a platform for free expression of views.
FOMC meeting minutes revealed that U.S. policymakers still intend to raise interest rates, and futures markets expect a second rate hike in December. The Bank of Japan's governor stated that core inflation has reached target levels, suggesting no rate hike in October. France witnessed large-scale street demonstrations, and the Spanish government announced early elections, causing the euro to continue weakening last week amid bond market volatility. Houthi militants attacked an airport in Saudi Arabia, pushing up Brent crude oil prices. U.S. West Texas crude prices remained relatively stable, but supply and demand imbalances persisted in oil products such as diesel, and natural gas prices surged. Amid market and geopolitical turmoil, gold prices rebounded last week.
Last week, the U.S. Treasury auctioned $39 billion in 10-year Treasury bonds, with a winning bid rate of 5.3%, the highest since November 2000, and oversubscribed 2.77 times, significantly exceeding the average of 2.54 times in the past five auctions. Overseas demand was particularly strong, with market analysis suggesting that overseas sovereign funds, insurance funds, and pension funds are re-entering the market, receiving 80.3% of the total allocation from underwriters. This was a remarkably strong auction and subscription, with high-net-worth funds beginning to discuss whether U.S. long-term bond yields are approaching a peak, and the pace of entry clearly accelerating.
First, this auction dispelled the weak sentiment seen over the past few months, as domestic and overseas real-money investors perceive yields are starting to become relatively attractive. Second, conflicts in Iran, Saudi Arabia, and the Russia-Ukraine war continue, creating significant uncertainty, yet capital appears willing to enter, demanding sufficient yield levels to compensate for risk premiums. Third, the U.S. fiscal situation is not optimistic, but other major developed nations are similarly struggling, with even weaker risk resilience, prompting capital to flow into dollar-denominated assets for safe-haven purposes.
This auction shows that there is demand for U.S. Treasuries, but prices must reflect premiums arising from fiscal and inflation risks. Yields around 5.3% are beginning to attract some funds to increase their U.S. bond allocations. Note, however, that this represents only a portion of available capital, not all. The chief investment officer of the giant bond fund PIMCO is discussing the risks of 10-year yields rising to 6%. Nevertheless, the market is no longer uniformly bearish, and the tug-of-war between bulls and bears will determine the future direction of the Treasury market and the timing of yield peaks.
Although the auction revealed demand from certain funds, especially large overseas funds, it has not resolved the awkwardness on the supply side. The U.S. fiscal situation shows no sign of fundamental improvement, and the government appears to lack even the will to improve it. Bond supply imbalances could continue to push up issuance costs. Meanwhile, inflation improvement is progressing slowly, and the U.S. Federal Reserve estimates further rate hikes will still be needed, which remains a significant uncertainty for the bond market.
Bond market volatility has so far had limited impact on U.S. stocks. First, major tech companies' profitability is strong. Second, the tech sector is less sensitive to interest rate cycles. Third, global cash is afraid of missing out on AI-related concepts. In reality, U.S. stocks outside of AI, including bank stocks, have already been deeply hurt by high interest rates, though this is not reflected in stock indices. As for AI companies issuing bonds or raising funds in private markets, the damage from high interest rates is already evident, as movements in risk-free assets inevitably trigger chain reactions across other assets.
The author believes that real demand is gradually emerging, and U.S. long-term bond yields have entered a phase of searching for a peak. However, the exact level of the peak still depends on the trajectory of inflation and the Federal Reserve's stance. Attention should be paid to the process of peaking and the spillover risks that high U.S. bond yields pose to stock markets and overseas bond markets.
Turning to differences between China and the U.S. in the AI sector. The U.S. AI industry is more driven by market-based capital, thus capital returns and commercialization pathways are subject to stronger market constraints, making heavy investment, closed-source weights, and closed-source models the prevailing industry norms. In China, AI industry funding sources are more diverse, with local government industrial funds and state-owned capital platforms playing significant roles alongside internet platforms and private capital. Most Chinese AI models adopt open weights and open-source models, focusing more on the social and inclusive value of the AI revolution, with funding levels far below those of their U.S. counterparts.
There are significant directional differences between the two countries in computing power utilization. The U.S. focuses more on R&D, while China emphasizes monetization through applications. According to public API token consumption data, over half of token usage in the U.S. is for programming development and enterprise automation, whereas in China, about half of token usage is for short videos, short dramas, and visual content production, with a significantly lower proportion dedicated to technological R&D compared to the U.S. The U.S. is brewing faster and larger technological breakthroughs, while China attempts to feed back AI industry development through application monetization, while closely following technological leaders.
Based on the author's subjective assessment of frontier model capabilities and product iteration pace, the gap between the two was about six months at the beginning of the year, narrowed to three to four months mid-year, and has recently widened again to about six months. The main reason for the widening gap lies in computing power; U.S. export restrictions on AI hardware have hindered China's expansion of computing capacity, while China is already on par with the U.S. in talent, algorithms, and data. In fact, differences in model parameters are no longer important—sufficient capability is enough. Competition is extending to the application layer, where profitability certainty matters more.
In AI applications, both China and the U.S. face similar bottlene游戏副本s. The low-hanging fruit, such as programming benefits, have already been harvested, but progress in enterprise adoption has been slower than initially expected. Enterprises feel uncertain about scenario setting and value confirmation in AI transformation, and friction is high when integrating transformation with existing architectures and processes. Each industry and each enterprise presents non-standardized application scenarios. The author expects a gradual and iterative process of 5 to 10 years. Users seem to find existing products sufficient, and most are unwilling to pay higher fees for new products.
In building data centers, the U.S. annual capital expenditure is about six times larger than China's, though the gap in investment efficiency is much smaller. Chinese tech giants also face a situation of nearly negative free cash flow this year, though pressure from funding gaps appears temporarily manageable. China's capital markets lack the depth and liquidity of U.S. markets, but local state capital, telecom operators, and the banking system offer financing channels different from U.S. capital markets, somewhat alleviating funding pressures on AI companies.
The endgame of the AI revolution is power supply. This year, the U.S. has entered an era of power shortages, while China's power supply remains abundant. China's annual new installed capacity is equivalent to about half of the total installed capacity across the entire U.S., while the U.S. faces dual constraints of insufficient power generation and transmission limitations. China's 46 ultra-high-voltage transmission lines efficiently and cheaply convert and transmit wind and solar power generated in the west to coastal regions, an advantage the U.S. cannot match.
There are few major data releases this week. U.S. September CPI is expected to show a technical rebound, with forecasts of a 0.6% month-on-month increase and a 3.6% year-on-year increase, core CPI rising 0.2% month-on-month and 2.5% year-on-year, slightly higher than the previous period. Watch the Fed's Beige Book, U.S. retail data (expected to weaken significantly), and China's monthly data. Tao Dun
*This week's article expresses the author's understanding and views on the economy, policy, and markets, representing personal opinions only, and is not investment advice or solicitation.
*Articles published in ET, signed or unsigned, represent the authors' personal opinions and do not reflect ET's position. ET serves as a platform for free expression of views.
