After HBM's high-level consolidation, where will the funds go next?

In the past cycle of AI market momentum, the core has been very concentrated, with trading centered almost entirely around two main themes: compute power and storage — specifically NVIDIA and HBM. After Micron's earnings reinforced demand expectations and SK Hynix continued to lead in HBM, the storage sector became one of the most crowded areas for capital. However, by the end of June, a noticeable shift began: although HBM remained in a strong trend, its upward slope slowed, volatility increased markedly, and capital began seeking new sources of returns. This change does not signify the end of the AI rally but rather a classic signal that the market is transitioning from a "single-theme-driven" phase to a "sector rotation" phase.

The Nature of HBM's High-Level Consolidation: From Expectation Trading to Realization Phase

The reason HBM entered a consolidation phase is not a sudden deterioration in fundamentals, but a change in market trading structure. Previously, the bullish logic for HBM was mainly based on "supply-demand gap + AI expansion expectations," a classic expectation-driven phase. However, as Micron and SK Hynix continuously released signals of capacity expansion and long-term orders, some of those expectations have already been realized ahead of time, and the market has naturally entered a rebalancing phase.

At the same time, the storage industry has inherent cyclical characteristics. Even under AI demand drivers, price elasticity goes through a process of "rapid rise—high-level stabilization—structural divergence." The market is clearly in the early stage of the third phase, where prices no longer move unilaterally upward but oscillate around high levels, awaiting new demand increments or new narrative expansions.

First Capital Migration: Data Centers and Cloud Infrastructure

When HBM enters consolidation, the first sector to absorb capital tends to be its upstream demand side—data centers and cloud infrastructure. The reason is that the logic chain is very clear: HBM demand comes from GPUs, and GPU demand comes from data center expansion.

Therefore, when the marginal gains in storage slow, capital naturally flows back to the "source of demand." In this phase, the market's focus shifts from a single chip to the overall capital expenditure on AI infrastructure, including cloud providers' server expansion, growth in AI inference demand, and accelerated enterprise AI deployment. The characteristics of this asset class are longer cycles and higher certainty, making it a more stable absorber during rotation.

Second Rotation: Repricing Compute Leaders

During the storage sector's consolidation, another common path is for capital to return to core compute assets. Although leaders like NVIDIA have already experienced significant gains, their dominant position in the AI ecosystem remains unchanged.

The role of HBM is essentially to "enhance GPU efficiency," not replace GPUs. Therefore, when storage enters a consolidation phase, the market often refocuses on the long-term demand curve for compute power itself. Especially as AI expands from training to inference, compute demand continues to grow, but the pace shifts from explosive to steady.

This phase is characterized by a rebalancing of valuation and growth, rather than the start of a new unilateral rally.

Third Diffusion: AI Applications and Software Layer Catching Up

When upstream infrastructure enters consolidation, capital typically spreads downstream to the AI application and software layer. This layer is characterized by relatively lagging valuations, larger narrative space, but longer cycles for earnings realization.

AI applications include enterprise AI tools, vertical industry models, AI agents, and various software service platforms. These assets often benefited less in the early stages but are prone to periodic catch-up during sector rotation. The market logic shifts from "infrastructure certainty" to "application-side elasticity."

Therefore, capital behavior in this phase is more driven by thematic and expectation repair rather than rapid fundamental validation.

Fourth Extension: Optical Communications and Energy Infrastructure

Beyond compute and applications, there is another direction gaining gradual market attention: the peripheral systems of AI infrastructure, including optical communications, high-speed interconnects, and energy systems.

As AI cluster scales increase, chip-to-chip connection efficiency and data transmission capacity become increasingly important, making optical modules and high-speed interconnect technology new bottleneck links. At the same time, the rise in power demand from large-scale data centers makes energy infrastructure a long-neglected but increasingly important variable.

This part is characterized by "hidden demand," which is often fully priced by the market only in the later stages of industrial expansion.

Structural Change in AI Industry Chain: From "Single-Point Outburst" to "Chain-Linked Rotation"

The most important change in the current AI rally is not the price movement itself, but the structural transformation. In the past, the core market model was a single-point outburst (e.g., GPU or HBM dominating). Now, it is shifting to a chain-linked rotation structure, where capital continuously spreads along the upstream and downstream of the industry chain.

This shift means that, on the one hand, excess returns from a single asset diminish; on the other hand, the overall extensibility of the industry chain increases, resulting in a longer-lasting rally. In other words, the AI rally is transitioning from a "strong trend concentrated rally" to a "multi-phase rotational rally."

Gate Stock Trading: Improved Trading Efficiency Under Cross-Market Rotation

As AI enters the rotation phase, a practical issue is that the market is more dispersed. US stocks, Korean stocks, and Hong Kong stocks correspond to different segments of the industry chain (e.g., storage, compute, applications), making cross-market tracking more important.

In this context, Gate Stock Trading supports 7×24-hour trading of US, Hong Kong, and Korean stocks, allowing investors to more flexibly switch asset allocation across different markets, thereby participating in the AI industry chain rotation process more timely. For highly event-driven technology rallies, this trading mechanism enhances cross-market response efficiency.

Conclusion: The AI Rally Enters a More Complex but Longer-Lasting Phase

HBM's high-level consolidation does not mean the end of the AI rally; instead, it indicates that the market has moved from a single theme to a structural rotation phase. Capital is spreading from storage to compute, data centers, applications, and infrastructure peripherals. This change makes the AI rally more complex, but possibly also more sustainable.

The future investment logic for AI is no longer about finding a single theme but understanding how the industry chain rotates and how capital reallocates among different segments.

FAQs

1. Why did HBM shift from a strong trend to consolidation?

Mainly because earlier expectations were quickly realized, while the industry entered a supply-demand rebalancing phase.

2. Why is capital flowing out of HBM?

Because after concentrated returns were released, capital started seeking new sources of growth elasticity.

3. Which directions are most likely to benefit in the next phase?

Data centers, compute leaders, AI applications, and optical communication infrastructure.

4. Has the AI rally ended?

No, it has not ended; it has transitioned from a single theme to a phase of industry chain rotation.

5. What is the significance of Gate Stock's 7×24-hour trading?

It improves cross-market participation efficiency, enabling investors to more flexibly capture opportunities at different stages of the AI industry chain.