The surge in technology stocks this year has been remarkable, yet market discussions about a "tech bubble" have never cooled down. History repeatedly shows that inflection points in tech waves often begin when imagination outpaces industrial reality—when expectations significantly lead actual developments, volatility and pullbacks inevitably follow. So, is this latest semiconductor boom ignited by AI just another bout of emotional exuberance, or the starting point of a sustained trend?

On July 3rd, at the Harvest Fund "Friends of Harvest" media communication session, Wang Guizhong, the Director of Big Tech Research at Harvest Fund, framed the rhythm of AI hardware investment with an analogy: "Computing power is the shovel, storage power is the vault, and transmission power is the highway." In his view, investing in technology is investing in the future, and it's not yet time to conclude a bearish stance—the "length, intensity, and duration" of this semiconductor upcycle may far surpass any previous demand recovery driven by consumer electronics or automobiles. However, he also emphasized that high volatility is inherent to technology's DNA, and investors must first learn to accept volatility before pursuing returns.

Behind the Divergence: Stocks as Voting Machines Short-Term, Weighing Machines Long-Term

The market divergence has reached an extreme year-to-date. Wang Guizhong finds this quite understandable.

On one hand, there is broader consensus on the value of AI. The global AI industry is moving into a commercial closed loop, with major domestic players shifting from a model competition to an ecosystem competition, burning cash to secure user entry points, and accelerating user adoption. On the other hand, there is a contrast in the fundamental industry sentiment—AI-related hardware is experiencing extremely high demand, with bandwidth and storage becoming the genuine physical bottlenecks; meanwhile, May's retail sales data showed temporary negative growth, indicating weakness in the traditional consumer sector.

"Stocks are voting machines in the short term and weighing machines in the long term," Wang Guizhong said. With market participants like quant funds and hot money involved, short-term stock prices naturally tend to follow the "votes" of market sentiment. He remains optimistic about Chinese tech assets and holds a positive view on China's accelerated catch-up in AI.

Will this divergence converge? Wang Guizhong did not answer directly but instead extended his perspective to a longer industrial cycle.

The Three Cornerstones of a "Super Cycle"

The market's most pressing question is: can semiconductors, which have led the market for most of the year, continue to run?

Wang Guizhong characterizes the current phase as a "super upcycle" and provides three layers of logic supporting this judgment.

The first layer is the "visibility" of aggregate growth. He cited forecast data from the World Semiconductor Trade Statistics organization—the global semiconductor market size is projected to reach $1.51 trillion by 2026, representing approximately 90% growth from current levels; of this, memory chips alone account for about $800 billion, a 250% year-on-year increase. By 2027, this figure may further climb to $1.9 trillion, with AI investment maintaining a relatively high growth rate. "Based on current observations, it's not yet time to conclude a bearish view," Wang Guizhong stated.

The second layer is the acceleration of domestic substitution. He indicated that domestic computing power cards can already meet inference-side demands, and capacity expansion in advanced process nodes is driving equipment investment. The capital expenditure intensity of leading domestic memory and advanced logic companies is expected to be revised upward from approximately $40 billion to around $70 billion, presenting historic expansion opportunities for upstream equipment, materials, and component companies.

The third layer involves the spillover effects from overseas capacity withdrawal. Driven by AI technology breakthroughs prompting capacity structure adjustments, major overseas semiconductor manufacturers are withdrawing from mature process nodes beyond expectations. Domestic companies are absorbing this spillover demand, with capacity utilization rates for mature processes nearing full capacity. Under this supply-demand imbalance, "there is hope for a wave of intensive price increases in the second half of the year."

With these three layers of logic combined, Wang Guizhong believes the "length, intensity, and duration" of this cycle may far exceed any previous demand recovery driven by consumer electronics or automobiles.

Following the Path of "Shovels, Vaults, and Highways"

Within such a large cycle, where should capital be allocated? Wang Guizhong's thinking unfolds along the line of "computing power is the shovel, storage power is the vault, and transmission power is the highway."

He judges that AI's demand pull on hardware follows a clear sequence—computing power first, storage power follows, and transmission power comes later. This sequence serves as the roadmap for sector rotation among sub-segments.

The first stop is optical communication and high-speed rate upgrades. Whether it's the interconnection needs within AI clusters or the iteration of optical modules from 800G to 1.6T, demand in this segment continues to expand. In Wang Guizhong's view, this is "one of the segments with the strongest growth certainty within computing power infrastructure."

The second stop is the price increase chain, including memory, optical chips, PCB upstream materials, etc. Supply-demand mismatch is the underlying logic. The explosion in high-bandwidth storage demand driven by AI has reversed the industry's supply-demand structure, pushing prices into an upward channel. He judges that this period of high prosperity "is expected to continue until 2028" and will transmit upstream to equipment, materials, and components.

The third stop is the global resonance in semiconductor equipment. Shortages lead to price increases, price increases lead to capacity expansion, and expansion first requires buying equipment—in this chain, the semiconductor equipment segment benefits earliest.

After outlining these three paths, a more practical question arises: how can ordinary investors grasp the opportunities of this tech wave?

Volatility is the Admission Ticket, Not the Surprise

"The first lesson in investing in technology is to accept its inherently high-volatility nature."

In his view, the root of tech bubbles often lies in "imagination running faster than the industry." Faced with uncertainties at the emotional and trading levels, the method he offers is to use long-term industry trends to counter short-term disturbances—while maintaining dynamic calibration, focusing on indicators like the revenue growth trend of AI applications, the capital expenditures and cash flow sustainability of cloud providers.

For ordinary investors, his advice is straightforward: "Systematic investment, long-term investment, contrarian investment." Do not chase highs, maintain restraint during sharp rallies; dare to buy during significant declines, especially the dips created by macro black swan events—as long as the fundamentals haven't changed, it's an opportunity; use systematic investment to smooth out costs.

Regarding tools, he categorizes them into two types: for investors with deep research into specific sub-sectors, thematic industry indices are a sharp spear; for those without the time for deep research, entrust capital to a truly trusted fund manager.