Overall, our prior views continued to play out as expected, and our core thesis remains unchanged. We believe the current AI cycle is gradually transitioning from being primarily expectation-driven to increasingly fundamentals-driven.

Market Environment

Markets performed strongly this month, supported by easing geopolitical tensions in the Middle East and continued improvement in AI fundamentals. Compared with March, valuation attractiveness has moderated somewhat. However, from a long-term perspective — especially considering the continued upward revisions in AI-related fundamentals — overall market valuations remain within a relatively reasonable range.

 

AI Fundamentals Continue to Strengthen

Despite the significant appreciation in AI-related assets, fundamentals have continued to improve alongside prices, with commercialization and investment increasingly forming a positive feedback loop.

On one hand, Anthropic reported ARR and token consumption significantly above expectations. According to Bloomberg and various third-party sources, Anthropic’s ARR increased rapidly from approximately US$9-10 billion at the end of 2025 to roughly US$30 billion by April 2026, with recent market discussions suggesting the figure may be approaching US$40 billion. This reflects accelerating real-world demand for AI usage.

Meanwhile, OpenAI continues to iterate rapidly on Codex capabilities and has quickly emerged as a first-tier platform, with commercialization progress also expected to accelerate further.

On the other hand, as commercialization pathways become clearer, AI model capability itself is no longer the primary bottleneck. Instead, the key challenge increasingly lies in “harnessing” AI within real-world workflows and enterprise deployment. AI applications remain in the early stages of adoption, leaving substantial room for efficiency gains and productivity improvement.

From a business perspective, investment and returns are beginning to enter a self-reinforcing cycle, significantly improving the durability and visibility of future AI spending.

Against this backdrop, expectations for overall capital expenditures (capex) continue to move higher.

 

Structural Opportunities

Structurally, we continue to believe that storage and networking-related segments remain among the most certain beneficiaries within the AI value chain and continue to offer compelling long-term structural opportunities.

As companies across the ecosystem continue reporting results, industry fundamentals remain firmly in an upward trajectory. Both revenue growth and profit margins are improving across many areas of the supply chain, leading to continued upward revisions in long-term growth expectations.

Taking networking infrastructure as an example, the long-term demand outlook for data center optical fiber continues to be revised upward:

• As AI cluster sizes continue to expand, network architectures are evolving from traditional two-layer designs toward three-layer structures, causing network growth to outpace compute growth itself;

• Bandwidth upgrades, which tend to double every one to two generations, are expected to further increase both fiber usage and value content;

• As future Scale-Up architectures continue to evolve, intra-rack interconnect demand may eventually exceed inter-rack demand by a meaningful margin.

In addition, the rapid expansion of AI inference workloads is increasing the physical distance between data center resources and end users. Compared with training workloads, inference infrastructure is also inherently more geographically distributed. As a result, we believe demand for Data Center Interconnect (DCI) infrastructure could enter a phase of rapid long-term growth, which still appears to be in its early stages today.