At Haitong International's autumn strategy conference, chief strategist Zhang Yidong stated that the AI bull market which began in 2023, along with the global technology J-curve cycle that started in 2021, has now entered its "autumn phase."

Zhang believes the AI bull market has not yet run its full course, but a major peak in 2027 is highly probable. The current AI rally has shifted into "autumn," and winter "is not far off." His logic is grounded in the J-curve cycle theory, suggesting that the global technology cycle, which bottomed and turned upward in Q1 2021, will likely complete its full trajectory by early 2028. Since stock markets typically peak about half a year ahead of fundamental conditions, Zhang projects the AI bull market's high point will most likely arrive around mid-2027.

"But autumn is also a season of vibrant colors," Zhang noted, emphasizing that the "autumn phase" still offers abundant investment windows. Investors can seize opportunities from now through the first half of next year, while remaining vigilant about "top-forming" market dynamics. He advises that investors "should focus on the tech mainline while also recognizing non-tech diffusion opportunities, avoiding overly crowded tracks."

Zhang flagged risks in the TMT sector, pointing out that within crowded segments like optical modules and memory storage, roughly 70% of stocks have already seen their highs during the June-July rally. Only about 30% of names, backed by earnings delivery, still have potential for new highs. In terms of stock selection, priority should go to China's "filling the gaps" sectors and areas where China holds global competitive advantages. Specific opportunities include AI applications, equipment, and consumer electronics; biopharma CXO, innovative drugs, and new technology platforms; new-quality manufacturing (commercial aerospace, robotics); and energy technology (AI data centers, transformers, power generation equipment).

On non-tech diffusion opportunities, Zhang highlights traditional industries "blooming anew." First, in resources, he holds a strategically bullish view on gold, expecting it to challenge the $5,000 mark within the year, with substantial upside potential in 2027-28. Second, in defense, he sees "gaming returns," particularly as AI-empowered military and defense AI adoption represent a long-term mega-trend. Third, in fintech, he focuses on brokers with high AI content and digital asset-related businesses. Fourth, in consumer sectors, he targets specific names that have undergone sufficient corrections.

The AI "autumn rally" signals that "AI applications" are entering a phase of explosive diffusion. Zhang believes the AI industry is now crossing the penetration rate inflection point, entering an accelerated demand expansion period. "Compared to the 30% penetration threshold of the Internet 1.0 era in 1998, current U.S. consumer-side AI usage has surpassed 30%, while corporate AI procurement stands at about 20%. AI is spreading from code generation to industry-specific applications."

While application-side adoption continues to deepen, AI capital expenditure growth is also set to expand. Zhang argues that even long-term returns may not cover the massive Capex investments accumulated so far. He predicts that over the next 1-1.5 years, large model development will descend into intense competition until corporate funding runs dry. However, before capital is exhausted, the continued decline in large model costs and ongoing version iterations will keep empowering various industries and sustaining the market rally.

"Currently, cloud vendors and semiconductor tech firms are seeing strong short-term earnings growth," Zhang said. "This can support the AI industry, but strong short-term results do not necessarily mean the long-term logic holds. The prerequisite for healthy industry development is a two-way connection between upstream computing infrastructure and downstream industry applications. If computing hardware costs keep rising, it will accelerate the end of the large model 'arms race.'" Zhang cautioned that with AI sector risk premiums at historical lows, investors must maintain humility when the cycle turns.