CITIC SEC: Bullish on computing metal, fluoride industry, as well as storage chain, optical modules, and other sectors.

CITIC SEC has released a research report stating that they maintain the view that the Federal Reserve will not raise interest rates in the mid-term, and there is room for valuation repair in types of carbon + silicon raw materials that are currently under liquidity discount. As for extreme silicon-based cyclical products with direct exposure, it is still a bit early to discuss the bursting of the bubble, but after valuations rise, it will require a balance of performance. It is no longer a comfortable trading range, and preference is given to types with strong supply constraints and low valuations, leaning towards the middle and lower end of the industry chain. Based on this, they are most bullish on the longer narrative, with AI exposure but under macro discount, in the carbon-based cyclical products (computing power metals, fluoride chemicals, phosphorus chemicals), as well as undervalued, middle to lower end products (storage chain, combustion engine chain, optical modules, PCB, cloud providers). Why does this AI cycle resemble the lithium battery cycle rather than the internet bubble? Both cycles belong to the "cyclical products market under accelerated penetration rate," where the difference lies in the fact that in the electric vehicle sector, demand is directly transmitted, while in AI, token consumption accelerates Capex indirect transmission of hardware in the middle to upper stream, making it more sensitive to fluctuations in penetration rate and having multiple acceleration points. The essence of their fluctuations is "supply-demand gap + price increase," falling under the framework of cyclical stocks rather than valuation framework - unlike the internet bubble that relied on inflated valuations (NASDAQ fwd PE 60x), both lithium batteries and AI are built on solid capital expenditure/consumption and real high profitability, with current valuations not exhibiting obvious signs of a bubble (NASDAQ dynamic PE is less than 30x). Both can be broken down into upstream, midstream, and downstream subcategories based on their bulk properties (resources materials/components end applications); in terms of position, AI electric vehicle sector in 2021midstream (NVIDIA) will take the lead, with upstream (storage/materials) expected to ignite by the end of 2025. How to interpret mid-term performance and valuation? CITIC SEC uses a holistic approach to calculate the market value changes of upstream, midstream, and downstream companies over a complete cycle. Typically, during the mid-term of a market cycle, there will be a period where the upstream outperforms the mid and downstream, but eventually returns. The point of initiation is when the industry capacity becomes large enough to affect the supply-demand balance sheet (corresponding to lithium batteries from September 2020 to September 2021, and AI starting from June 2025 when the materials market ignites, with amplitude and coverage already exceeding electric vehicles); more importantly, the market value peak of electric vehicles leads the performance peak by about 6 quarters - the PE of core upstream companies as a whole, which was pushed down from 137x to 18x during the surge in performance. Most real peaks occurred in Q3 of 2021, indicating that the selling point is not in performance/gross profit realization, but rather when the market value, valuation, and stake loosen. The global PE of AI upstream companies has now fallen from 100x to around 40x, which is comparable to the electric vehicle sector in mid-2021, where while performance will continue to materialize, valuation is unlikely to rise further. In addition, sell-side expectations show a significant deviation, with analyst models consistently being more conservative towards the upstream in the early stages and more aggressive in the late stages compared to the mid and downstream. Once there is a significant increase in profit/ASP consistency expectations in the upstream, the stock price is likely not far from its peak. For AI upstream companies, it is expected that the increase in foreign profits will not be widespread until the first half of 2026, and the dynamic PE of the middle-to-upstream has not yet begun to collapse (reflecting a stronger performance breakout in this cycle with tighter mid-term supply expectations). How to track in the short term: Four topping signals 1) Watch closely for the topping of "least scarce" products in the short term - In the electric vehicle cycle, the most scarce lithium carbonate and cobalt commodity prices lagged behind stock prices for nearly a year, with little reference value. The real leaders were actually products with less tight supply-demand characteristics (such as electrolyte in 2021: the most relaxed in supply-demand, with prices leveling off in 4Q21 as demand slowed, and stock prices simultaneously weakening); applying this to AI, DRAM lithium carbonate (direct exposure, constrained capacity, quickest price increases), silicon wafers electrolyte (slowest and smallest price increases, structural oversupply, large consumption exposure), and NAND spot prices have already seen a marginal slowdown in price increases, so focus on silicon wafers - once they enter a price stabilization period, it could be a short-term topping signal. 2) Next, observe the price game between upstream and downstream: In the electric vehicle cycle, upstream stock prices peaked in Q4 of 2021, while Tesla first raised prices at the end of March 2021, and both domestic and international automakers collectively raised prices or halted production mainly from 4Q21 to 1Q22; in the AI sector, Apple's price increases for new devices and the ongoing battle with Micron over storage procurement prices took place around June 2026. 3) Keep an eye on Capex announcement density, which has a stronger explanatory power for upstream stock prices during the electric vehicle cycle than commodity prices (disclosures affect mid-term supply-demand expectations). However, this round of AI is dominated by overseas entities, with more challenging capacity expansion and tighter monetary conditions, so the forward-looking effectiveness of domestic Capex announcements may be weaker than in the electric vehicle sector. Large-scale overseas capacity expansion would be a warning signal. 4) Lastly, the collapse of crowdedness and diffusion is the final signal of the cycle's end, which has not yet been triggered: Crowdedness is a "necessary but not sufficient" indicator, and relaxation is the endpoint, while diffusion looks at the proportion of new highs (leading stock price peaks). In the electric vehicle sector, the proportion of new highs peaked at 62% in 2021, leading stock prices by 34 months, while in the AI sector, while crowdedness has already reached near historic highs, the proportion of new highs has not declined, and the clustering of leading stocks remains strong. Which types currently offer better value for money? Two rules determine the value for money of the latter stages: the greater the rise in the previous bubble, the deeper the fall after it bursts, and in the electric vehicle cycle, midstream processing materials with low barriers and easy capacity expansion hit an average market value low about 21% of the peak within 2 years of the top, while upstream resources with resource/quota constraints and slow capacity expansions hit about 32% of the peak. In the past couple of weeks, there have been expectations of tightened macro liquidity due to a more ambiguous FOMC and expectations of interest rate hikes brought about by high oil prices, May CPI, and non-farm data being influenced by the World Cup. CITIC SEC maintains the view that the Federal Reserve will not raise interest rates in the mid-term and believes that there is room for valuation repair in types of carbon + silicon raw materials that are currently under liquidity discount. As for extreme silicon-based cyclical products with direct exposure, it is still a bit early to discuss the bursting bubble, but once valuations rise, a balance of performance will be required. It is no longer a comfortable trading range, and the preference is towards strong supply constraints and low valuations, focusing on the middle and lower end of the industry chain. Based on this, CITIC SEC is most bullish on the longer narrative, with AI exposure but under macro discount, in carbon-based cyclical products (computing power metals, fluoride chemicals, phosphorus chemicals), as well as undervalued, middle to lower end products (storage chain, combustion engine chain, optical modules, PCB, cloud providers). Key charts: [charts not included in the translation]