CICC: Fed lacks basis for consecutive large rate hikes; AI investment returns to be validated in the first half of next year.

CICC released a research report stating that the Fed lacks the necessity and basis for consecutive and large rate hikes, for the following reasons: First, traditional demand does not support large rate hikes. Second, inflation will first fall rapidly driven by a high base. CICC roughly estimates that US CPI will return to 2% in May 2027, significantly below the current 3.7%. Third, the investment returns and financing pressure of AI itself have also wavered, and the first half of next year is a key validation point. The current prolonged lack of new narratives and catalysts for AI to open up the "ceiling" of application scenarios is the key reason why weak links such as CoreWeave's credit risk continue to rise. This is exactly the same as the situation at the end of last year when AI likewise lacked progress and the market worried about private equity credit risk. At the October FOMC meeting concluded in the early hours of this morning Beijing time, the Fed held rates steady as expected, maintaining the benchmark rate at 3.75-4%. The incremental information released at this meeting was very limited. Fed Chair Warsh remained "sparing with words" and once again said he does not like forward guidance. In addition, no rate hike at this meeting was entirely within market expectations, because market changes were also very limited. After the meeting, US Treasury yields were basically unchanged, with the 10-year yield rising slightly above 5%, US stocks also maintained a range-bound performance, and the probability of a December rate hike implied by CME rates remained around 85%. Although after the last rate hike in September, rate hike expectations fluctuated repeatedly, and September CPI and PPI beating expectations during the period also triggered market concerns, the overall policy direction was also in line with CICC's previous judgment, namely that the September rate hike was necessary because it needed to safeguard its credibility (this was basically achieved, with the term premium returning to 60bp before concerns escalated in July), but consecutive and large rate hikes were unnecessary ("Is One Fed Rate Hike Enough?"). After the September rate hike, the market environment has quietly changed: rising financing costs have already squeezed traditional demand such as real estate, and weak links in AI such as NeoCloud representative CoreWeave's CDS and high-yield bond spreads have also risen rapidly, all reflecting a tightening of financial conditions. The outlook for AI demand and the sustainability of capital expenditure has therefore also been somewhat clouded. US Treasury yields seem to be topping out, oil prices remain high, so is a December Fed rate hike a foregone conclusion? How many more times will the Fed hike? Chart: October rate hike expectations rose above 70% at the end of September, but had fallen to less than 10% before the meeting Source: Bloomberg, CICC Research Department Core information from the October FOMC: pause rate hikes, follow data going forward; strong economy, but tightening financial conditions; reduce the number of meetings going forward The incremental information provided by this meeting was very limited, mainly conveying the following signals: 1) Maintain the benchmark rate at 3.75%-4%, in line with market expectations; 2) The economy remains strong; 3) Financial conditions have begun to tighten, reflected in some credit spreads, especially high-yield bonds; 4) Future decisions depend on data, with 2 nonfarm payroll reports and 1 inflation report before the December FOMC. The September meeting minutes showed that participants generally emphasized that inflation remains elevated, while labor market risks have diminished, so most officials indicated that one more rate hike may still be needed before the end of the year; 5) Warsh once again emphasized that he does not like forward guidance, and that the number of meetings will be gradually reduced in 2027 and 2028. Future policy path: consecutive and large rate hikes are also unnecessary, with the first half of next year a key validation point; market expectations are basically all unreliable The Fed's pause in rate hikes this time was already within market expectations and also consistent with CICC's previous judgment. Before the September rate hike was implemented, CICC noted that the Fed needed to hike rates to rebuild market trust in it, but also emphasized that US fundamentals did not support consecutive and large rate hikes ("What Happens If Rates Are Hiked?"). Since the purpose of rebuilding trust has now basically been achieved for the time being (the 10-year US Treasury term premium has fallen back to the market expectation level before Warsh's remarks at the July FOMC meeting), nonfarm payrolls are weakening, AI concerns are increasing, financial conditions are tightening in some areas, and supply constraints are not significantly tight for now, then why shouldn't the Fed observe for a while, especially with the midterm elections just around the corner. Looking ahead, CICC still believes that the Fed lacks the necessity and basis for consecutive and large rate hikes, for the following reasons: First, traditional demand does not support large rate hikes. The US economy remains in a K-shaped divergence of "strong tech but weak traditional demand," and the current rate level has already begun to significantly squeeze traditional demand. Take real estate as an example: the 30-year mortgage rate has risen to a multi-year high of 7.5%, significantly above the 6% rental yield, and existing home sales have weakened again since May; the short-term loan rate for small businesses has also risen to 8.2%. Chart: The US economy remains in a K-shaped divergence of "strong tech but weak traditional demand" Source: Haver, CICC Research Department Chart: Existing home sales continue to weaken under the pressure of rising US mortgage rates Source: Haver, CICC Research Department Chart: Both US 30-year mortgage rates and small business short-term loan rates are rising rapidly Source: Bloomberg, CICC Research Department However, the strength of AI has offset and masked the pressure on traditional demand at the aggregate level. But when AI begins to need to draw resources from the whole of society, its crowding out of traditional demand in physical and financial factor resources will also become increasingly obvious. Previously, the positive externalities brought by AI demand supported US economic growth and also provided the Fed with the "confidence" to hike rates. But as AI begins to need more external funding to support investment, crowding-out effects are also unavoidable, and it may gradually turn from a "friend" of traditional demand into an "enemy" ("AI from a 'Friend' to an 'Enemy' of Traditional Demand?"). AI's massive bond issuance and the Fed's need to keep hiking rates are typical examples. Once AI's investment returns and positive externalities weaken, the combined effect will no longer be supportive, and it will be difficult for the Fed to continue hiking rates. Chart: In the second quarter, US real GDP grew at an annualized quarterly rate of 2.2%, with AI-related demand contributing 28% of growth Source: Haver, CICC Research Department Second, inflation will first fall rapidly driven by a high base. The high base in the first half of this year through May due to the Iran situation will cause overall US inflation to fall rapidly. Assuming oil prices remain at the status quo (around US$95/barrel), CICC roughly estimates that US CPI will return to 2% in May 2027, significantly below the current 3.7%. Chart: CICC estimates that the high base will push CPI year-on-year down to around 2% in the first half of 2027 Source: Haver, CICC Research Department Geopolitics and oil prices are indeed short-term disturbances that affect the path, but over a longer horizon they do not affect the trend. If short-term geopolitical tensions escalate again and push oil prices higher, this would indeed increase the pressure of CPI running above expectations and of rate hikes, and this is the biggest variable in the short-term path. But over a longer horizon, geopolitical conflicts do not necessarily lead to a sustained contraction in crude oil supply, because the current high premium means that even if the Strait of Hormuz is not open, various alternative transport routes (such as pipelines, ship-to-ship transfers, etc.) have basically restored Gulf exports excluding Iran to pre-conflict levels. The same was true during the 2022 Russia-Ukraine conflict: although the geopolitical situation had not fully eased, Russia's crude oil supply gradually recovered to pre-conflict levels, and oil prices gradually fell back. Therefore, there may be short-term path disturbances, but the medium-term trend may not be affected. Chart: According to Kpler data, crude oil exports from Gulf countries other than Iran have also recovered to pre-conflict levels Source: Kpler, CICC Research Department Chart: The gradual release of medium- and long-term crude oil supply will push oil prices lower Source: Haver, CICC Research Department Third, the investment returns and financing pressure of AI itself have also wavered, and the first half of next year is a key validation point. The current prolonged lack of new narratives and catalysts for AI to open up the "ceiling" of application scenarios is the key reason why weak links such as CoreWeave's credit risk continue to rise. This is exactly the same as the situation at the end of last year when AI likewise lacked progress and the market worried about private equity credit risk. The first half of next year is a key node, similar to how Anthropic's results greatly exceeded expectations in the first quarter of this year and proved to the market that the closed loop of the coding business had been achieved. Otherwise, if it does not appear for a long time, market patience will be exhausted, and the outlook for capital expenditure and returns in 2028 will also become further dim. As the main driver currently boosting CKH HOLDINGS' profit, if one does not have strong confidence in AI, then how can one expect the Fed to continue hiking rates? In summary, traditional demand beginning to come under pressure plus the rapid fall in inflation in the first half of the year both seem not to support consecutive and large Fed rate hikes. Moreover, if AI demand slows more than expected, the Fed may even need to cut rates, and the first half of next year is a key validation point for AI and the policy path. The upside risks to CICC's view are persistently higher-than-expected oil prices and AI progress greatly exceeding expectations. As for market expectations, historical experience shows that they are almost always unreliable. Chart: From historical experience, the rate path implied by rate futures deviates from the eventual actual path Source: Bloomberg, CICC Research Department How high a rate can the US withstand? 6% may be the watershed On one side is the strength of AI, and on the other is the gradual pressure on and crowding out of traditional demand, so when might AI's positive externalities no longer be able to offset the weakness of traditional demand? CICC estimates in "AI from a 'Friend' to an 'Enemy' of Traditional Demand?" that this watershed may be around a 6% 10-year US Treasury yield (AI can withstand 7% or even very high levels, while traditional demand can withstand around 4-4.5%). This also explains why the recent rise in US Treasury yields has not yet fully hurt overall growth, but erosion of local weak links has already begun to appear. If this inflection point is exceeded, AI's positive externalities can no longer continue to offset the negative externalities, causing overall growth to come under pressure, and then the Fed may even need to begin cutting rates. How will the market trade? A brief rate hike still fits the "sell the news" pattern The recent topping and pullback in US Treasury yields basically fits the pattern that once a rate hike is implemented, expectations are gradually realized. In September's "Is One Fed Rate Hike Enough?", CICC noted that the implementation of a rate hike is not bad for the market, but may instead mean "bad news exhausted." A typical case is the 1997 "preventive rate hike," after which US Treasury yields quickly peaked and fell back; conversely, the "preventive rate cuts" of 2019, 2024, and 2025 also played out a similar "script," with US Treasury yields gradually rising at the moment the cuts were implemented or shortly thereafter ("What Happens If Rates Are Hiked?"). Compared with this time, although October inflation data caused rate hike expectations to "rebound" and US Treasury yields to rise again, if a brief rate hike is the baseline assumption, it still fits this pattern. Based on the dot plot of one more rate hike, a benchmark rate of 4-4.25% plus a term premium of 70-90bp corresponds exactly to a 10-year US Treasury yield of around 5.3%. Therefore, if CICC's judgment above on the Fed's future rate hike path is accurate, then the moment of tightest financial conditions is also gradually approaching. Chart: A rate expectation of 4-4.25% plus a term premium of 70-90bp corresponds to a peak US Treasury yield of around 5.3% Source: Bloomberg, CICC Research Department US stocks instead reflect more the AI narrative and sentiment disturbances from the midterm elections. In its June mid-year outlook, CICC further raised its S&P 500 target to 7,800-8,000, and after the market reached this level it has been caught in a struggle and range-bound trading. In addition to the liquidity disturbance brought by rising rates, this is more because the recent rise in AI financing risks and the slowdown in revenue realization have both weighed on the overall AI narrative, so the market has also "narrowed its focus" to cloud vendors with not-high valuations and expectations that do not need to rely entirely on future high-growth assumptions. Since US stock earnings remain solid while valuations have instead contracted, CICC remains not pessimistic on the US stock outlook. Chart: The AI stress index constructed by CICC has also continued to rise since the end of September Source: Bloomberg, FactSet, CICC Research Department The dollar strengthened amid the pullback in US Treasury yields and easing rate hike expectations, partly reflecting the "foil" of a weaker Europe and the return of capital. Although US Treasury yields have begun to fall, the dollar has not weakened in tandem. The main reason is that Europe faces greater uncertainties in politics and fiscal policy, so some capital has flowed back, with the only question being whether it chooses US Treasuries or US stocks. Chart: The dollar is range-bound in the baseline scenario, and in the short term a Fed rate hike to rebuild trust still supports the dollar Source: Bloomberg, CICC Research Department Gold is currently more like a "bullish option," with three meanings: First, a certain "option premium" must be paid, namely time cost; second, downside pressure is controllable, with the biggest risk being consecutive and large rate hikes; third, option realization requires "exercise conditions," namely a narrative. For example, at the end of last year, market concerns about AI investment returns, US policy stability, and Fed credibility jointly catalyzed a sharp rise in gold. Chart: Greater upside for gold comes from grand narratives Source: Polymarket, CICC Research Department Judging from the rate hike expectations over the next year priced into various assets: rate futures (3 hikes) > gold (2.4 hikes) > copper (1.9 hikes) > Dow Jones (1.7 hikes) = US Treasuries (1.7 hikes) > S&P 500 (1.1 hikes) > dot plot (1 hike) > Nasdaq (0.4 hikes). Chart: Judging from the rate hike expectations over the next year priced into various assets, rate futures > gold > copper > Dow Jones = US Treasuries > S&P 500 > dot plot > Nasdaq Source: Bloomberg, CICC Research Department