CITIC SEC: The "politicization" of data centers may be the biggest obstacle to the short-term AI narrative.

CITIC SEC released a research report stating that the halt of data center construction in Texas, USA, and politicization may be the biggest obstacle to the short-term AI narrative. For Hong Kong stocks, since 2026, due to limited incremental funds, the rotation effect between different industries has significantly strengthened, especially under the current situation of continuously tightening overseas liquidity. With a relatively limited total amount of funds, technology and biotechnology, as two growth tracks in the Hong Kong stock market with higher elasticity and valuations more sensitive to liquidity, are more likely to form a seesaw allocation relationship. At the same time, since southbound capital inflows in 2026 have been relatively weak, especially since June when foreign capital has continued to flow back into Hong Kong stocks, they have dominated marginal pricing power. Therefore, against the backdrop of foreign capital continuously flowing out of Hang Seng Tech and into biotechnology, a pattern has formed in which Hang Seng Tech is relatively weak while biotechnology is relatively strong. Overall, under a high interest rate environment, both the Hang Seng Tech Index and the Hang Seng Biotech Index are under significant pressure. Under the current expectation of tightening global liquidity, investors are advised to prioritize quality sectors with strong defensive attributes and stable dividends, such as power, telecommunications, and utilities. The main views of CITIC SEC are as follows: AI companies' financing channels are diversified, and the potential financing volume they can absorb in 2027 may exceed USD 740 billion. This week, US Treasury yields surged again, while credit spreads on US corporate bonds rated B and below also began to rebound. The simultaneous rise in risk-free rates and credit spreads once again led some investors to worry about the sustainability of US AI financing. However, the bank believes that US AI companies' financing is relatively diversified. In addition to debt financing, they can support future CAPEX through equity financing, industrial capital, and reduced share buybacks. Over the past two years, the main financing channel for US AI companies' CAPEX has been the debt market, while the only equity market financing case was Google's USD 85 billion this year, possibly because US listed companies have long sought to avoid diluting existing shareholders' equity. From the perspective of the debt market, US AI companies' financing process over the past two years has also been quite smooth. According to Bloomberg, in the first half of this year, the five major US hyperscalers had raised USD 167.8 billion in the global public bond market; Meta and Anthropic also raised USD 27 billion and USD 35 billion in the private bond market in 2025 and this year, respectively. In addition, if subsequent financing demand in the equity and debt markets is exhausted, US listed companies may also consider reducing share buybacks to support CAPEX. According to data from Wind, Bloomberg, and Norges Bank, taking into account the US stock market, global bond market (public & private bonds), syndicated loan channels, potential conversion of US listed companies' buyback funds into CAPEX, and financing from US industrial capital and sovereign wealth funds, under a neutral scenario the bank estimates that the volume available to absorb US AI companies' CAPEX in 2027 will be USD 742 billion. The halt of data center construction in Texas, USA, and politicization may be the biggest obstacle to the short-term AI narrative. This week, the United States halted data center construction in Texas, and Oracle also issued a force majeure notice to the developer of Project Jupiter, hoping to delay payments if the project is not brought online by 2028 as stipulated in the contract. Combining these two developments, the bank judges that data center construction will be one of the most important directions of bipartisan contention around the midterm elections. In addition to the "timing" constraint of power grid interconnection and the weakening of "geographic" conditions caused by stricter administrative approvals, the "human" conditions for US data center construction are also showing a trend of marginal deterioration. The reasons why US residents oppose data centers being built in their communities mainly fall into two categories: one is factors affecting living experience such as noise and water use, and the other is factors directly affecting household economic interests such as rising electricity prices and land and property depreciation. As NIMBY sentiment in the United States heats up, the risks faced by data center projects are no longer limited to stricter early-stage government approvals and higher resident communication costs, but also include legal disputes that may continue to occur during project construction, further triggering restart of approvals, site relocation, or even project cancellation, thereby increasing the delay risk in the process of converting commercial commitments such as signed but not yet leased contracts by hyperscalers into actual capacity. Therefore, the 2026 US midterm elections may become an important political variable that further amplifies industry regulatory risk. If the Democratic Party wins the House of Representatives, it may push data center-related issues to receive greater congressional attention and extend the fulfillment cycle of projects from planning to production, thereby increasing the computing power delivery delay risk faced by hyperscalers that have already formed commercial commitments. US stocks and bonds diverged significantly in the third quarter, and there may be rebalancing demand at quarter-end. From the third quarter through September 24, the S&P 500 rose 2.73%, while TLT, which tracks US Treasuries with maturities of more than 20 years, fell 8.10%. The single-quarter return gap between the two reached 10.83 percentage points, at the 80th percentile since 2002. Looking back at history: among the 34 historical samples in which the single-quarter stock-bond return gap exceeded 8 percentage points, TLT rose by an average of 0.8% in the next quarter, with 18 positive instances; raising the threshold to 10 percentage points, among the 23 samples, TLT rose by an average of 0.4% in the next quarter, with 11 positive instances. Based on this, the bank judges that after a large stock-bond divergence, US Treasuries have a certain tendency toward mean reversion, but the magnitude is limited and the win rate is only slightly above half. What rebalancing brings is more likely a moderate stabilization of TLT rather than a trend reversal. By contrast, momentum continuation in US stocks has been more stable historically. In the above two types of event samples, the S&P 500 rose by an average of 2.4% to 2.6% in the next quarter, with declining samples accounting for less than 30%. Based on this, the bank judges that the marginal impact of quarter-end rebalancing on US stocks is weaker than market intuition, and there is insufficient basis for significantly reducing US equity holdings and substantially increasing US Treasury allocations; a more reasonable judgment is that the bond market stabilizes in the short term, the relative return of US stocks continues, and the medium-term pattern of stocks outperforming bonds has not changed. Since June 2026, the seesaw effect between Hang Seng Tech and Hang Seng Biotech has significantly strengthened. Since 2026, due to limited incremental funds, the rotation effect of funds among different industries has significantly strengthened. With a relatively limited total amount of funds, technology and biotechnology, as two growth tracks in the Hong Kong stock market with higher elasticity and valuations more sensitive to liquidity, are more likely to form a seesaw allocation relationship. At the same time, because southbound capital inflows in 2026 have been relatively weak, especially since June when foreign capital has continued to flow back into Hong Kong stocks, they have dominated marginal pricing power. Therefore, against the backdrop of foreign capital continuously flowing out of Hang Seng Tech and into biotechnology, this has pushed forward the formation of a pattern in which Hang Seng Tech is relatively weak while biotechnology is relatively strong. Overall, under a high interest rate environment, both the Hang Seng Tech Index and the Hang Seng Biotech Index are under significant pressure. Under the current expectation of tightening global liquidity, investors are advised to prioritize quality sectors with strong defensive attributes and stable dividends, such as power, telecommunications, and utilities. Risk factors: 1) Global central banks continue to tighten monetary policy beyond expectations; 2) Global geopolitical conflicts escalate again; 3) Frictions in the China-US technology, trade, and financial fields intensify; 4) The United States tightens approvals for data center construction beyond expectations; 5) China's policy intensity, implementation effects, and economic recovery fall short of expectations.