As physical AI moves toward generalization, where can opportunities near the bottom be found in metals, functional sugars and building materials?

Why is progress in physical AI shifting from hardware to learning capability?

Sinolink Securities is focusing on changes in how robots acquire new skills: GEN-1.5 uses a single demonstration lasting 3 to 12 seconds as a prompt and can perform new tasks without updating its parameters. Its average success rate across 10 short-duration tasks is 59%; after an update using data from approximately 50 demonstrations, it rises to 83%.
The evaluation criteria are also shifting toward sustained output. Effective intelligent working hours, proposed by AgiBot, combines the number of robots, autonomous working time and output efficiency; data on failures, recoveries and human takeovers generated during continuous operation can also be used for model iteration.
Tesla’s Optimus has begun early production, targeting capacity of approximately 20 thousand units per week in about a year and deliveries to external enterprise customers by the end of 2027. The company expects it will still need about a year to resolve electromechanical design, reliability and AI software issues.

Changjiang Securities believes that overseas interest rate increases and tightening liquidity can periodically suppress metal prices and valuations, but overseas electrolytic aluminum expansion is constrained by resources, energy and outbound investment. As companies reduce leverage and lower their capital expenditure levels, more cash flow is being directed toward dividends.
Based on annualized interim results for 2026 and an average payout ratio of 52%, leading electrolytic aluminum stocks have an estimated average price-to-earnings ratio of 6.1 times and a corresponding dividend yield of 8.5%. At a 5% dividend yield, estimated average upside is approximately 70%.
Profit pressure is more immediate in steel: rising coking coal prices and weak steel prices have narrowed the spread between selling and purchase prices, and only about 11% of steelmakers have positive cash flow. After production cuts, iron ore and coke prices have begun to soften, and the brokerage expects profitability may improve in the fourth quarter.
Hydrometallurgical nickel production consumes about 9 to 10 tonnes of sulfur per tonne of nickel. Sulfur prices in Indonesia rose from about 550 dollars at the start of the year to 1100 dollars per tonne, increasing the cost of hydrometallurgical nickel by nearly 6000 dollars per tonne. A profit recovery depends on sulfur prices falling or companies increasing their sulfur self-sufficiency.

CSC Financial believes opportunities depend on the supply structure and cycle position of each product. Jinhe Industrial and CuHua jointly control approximately 95% of the global acesulfame potassium market. Utilization of CuHua’s designed capacity of 15 thousand tonnes is 57.9%, and idle capacity provides profit upside when prices rise.
Erythritol, meanwhile, faces excess capacity: nominal capacity is 300 thousand to 350 thousand tonnes against demand of about 150 thousand tonnes, and mainstream quotes have fallen to the cost line of approximately 10 thousand yuan per tonne. Some production lines have switched products, halted construction or exited, and the brokerage expects the shakeout to take another two to three years.
Profitability also varies among dietary fiber companies: in the first half of 2026, Bailong Chuangyuan’s gross margin for the business was 49.29%, versus 17.39% at Baolingbao. Over the same period, Bailong Chuangyuan’s sales volume in the business rose 5.52% year on year, while Baolingbao’s fell 2.34%.

Changjiang Securities sees low holdings, valuations and expectations as the starting point for allocating to building-materials dividend stocks. As of the second quarter of 2026, excluding fiberglass, institutional holdings in consumer building materials and cement accounted for only 0.2% and 0.1%, respectively; cement and glass price-to-book ratios were near historical lows.
Building-materials dividends are less stable than those of hydropower and banks, and industry profitability is generally low. The cement industry recorded an aggregate loss of approximately 4 billion yuan in the first half of 2026. If stronger support for domestic demand drives a recovery in infrastructure investment and supply constraints tighten, the brokerage expects industry profitability may improve in 2027.
Dividend capacity is also affected by funding needs: Huaxin Building Materials is expanding and needs funds for mergers and acquisitions. The brokerage expects dividends may rise once it reaches a stable phase and capital expenditure slows. Beijing New Building Materials has a debt-to-asset ratio below 30% and ample cash flow, providing a basis for higher dividends.