The era of the "inference economy" is now underway. As NVIDIA integrates its super chips into laptops and LENOVO GROUP reports massive order backlogs, AI is making a powerful transition to local, edge devices.

This surge in edge-side AI has left many investors confused by the array of Hong Kong-listed tech indices. Which one truly offers the highest concentration of "hard tech"?

Let's break down the underlying differences between four major Hong Kong tech indices and examine a core ETF with significant exposure to semiconductors and AI PCs, highlighting a key opportunity in the next phase of the computing revolution.

The Rise of the Inference Economy

The inference economy represents a new industrial phase where AI shifts from model training to practical deployment and utilization.

For the past two years, capital has been heavily focused on the training stage—investing in NVIDIA GPUs, building massive computing clusters, and processing data to create smarter models.

Now, with models maturing, the game has changed. The real scale lies in inference—every time a model answers a question, generates an image, or processes code, it consumes computing power. Think of a trained model as a PhD graduate now performing work; each task is an inference. The demand for inference far exceeds training because it occurs every single time a model is used.

Why is this suddenly gaining momentum? Three key drivers: First, the cost of running large model inference has dropped significantly, making widespread use feasible. Second, applications like AI agents, AI programming, and autonomous driving are now actively consuming computing power, moving beyond conceptual stages. Third, NVIDIA's collaboration with MediaTek on the RTX Spark super chip brings robust inference capabilities directly to PCs and laptops, making local, edge-side inference a tangible reality.

The AIPC Surge: The Battle for the Inference Endpoint

The recent strong performance of LENOVO GROUP—up nearly 110% over one month and 177.55% over three months—alongside gains for Dell and HP, reflects the inference economy thesis playing out at the device level.

The core idea is that inference happens not just in the cloud, but also on the device in front of you. At a recent industry event, NVIDIA's CEO announced that brands including LENOVO GROUP, Dell, and HP will launch PCs featuring the RTX Spark chip. This co-developed chip integrates CPU and GPU capabilities to run the Arm version of Windows locally.

This enables numerous daily inference tasks—document summarization, image generation, code completion—to be processed locally without constant cloud calls, enhancing privacy and reducing latency. Earlier, Intel launched its third-generation Core Ultra processors, boasting a 2.7x improvement in NPU AI performance and 64% lower power consumption, with support for hybrid edge-cloud AI.

Intel China has stated that the core goal for AI PCs is achieving local task completion, long-term memory evolution, hybrid edge-cloud inference, and local security control. In this AI race, LENOVO GROUP has positioned itself advantageously. Its AI-related revenue reached 38% of total sales in its last fiscal quarter, with annual group revenue surpassing $80 billion for the first time. AI server revenue hit record highs, growing 50% year-over-year, supported by an order backlog exceeding 140 billion yuan.

This trend signifies the inference economy's evolution from pure "cloud computing" to "edge-cloud synergy." Analysts note that NVIDIA's entry into the PC platform will accelerate AI's penetration to the edge, positioning AI PCs as a major new demand driver and growth engine for cloud AI.

Four Hong Kong Tech Indices: Similar on Surface, Different in Substance

These four indices essentially form a spectrum from highest to lowest "hard tech density." The performance of the Hang Seng Stock Connect Information Technology Index, tracked by the Hang Seng Stock Connect Information Technology ETF, has notably diverged since April, warranting a closer look.

The indices differ significantly in their selection criteria, number of constituents, and weighting rules. The Hang Seng Stock Connect Information Technology Index focuses solely on the information technology sector, selecting 30 constituents weighted by free-float market cap with a 15% single-stock cap.

In terms of sector allocation, this index shows high concentration in "hard tech" areas. Its top three sectors are semiconductors (33%), computer equipment (16%), and consumer electronics (15%). In contrast, the Hang Seng Stock Connect Internet Index is heavily concentrated in internet software and services. The Hang Seng Tech Index is more diversified, covering IT, consumer discretionary, and healthcare. The Hang Seng Stock Connect Technology Index offers a balance across technology, healthcare, and new energy sectors.

Regarding top holdings, all indices include major Hong Kong-listed tech leaders, but with different weightings. The Hang Seng Stock Connect Information Technology Index shows more concentrated exposure to semiconductors and communications. The Hang Seng Tech and Hang Seng Stock Connect Technology indices have lower individual stock concentration and higher diversification.

Looking at recent market favorites like SMIC, Hua Hong Semiconductor, and LENOVO GROUP, the Hang Seng Stock Connect Information Technology Index holds the highest weights among the indices for these companies—13.71% for SMIC, 10.00% for Hua Hong, and 16.10% for LENOVO GROUP—highlighting its "overweight" stance on specific core tech leaders.

From a long-term return perspective, the Hang Seng Stock Connect Information Technology Index has consistently outperformed the other three indices. Historical annual returns from 2021 to 2025 show it delivered positive returns in 2023, 2024, and 2025, with a 39.30% gain in 2025.

Key Takeaways for Investors

For exposure to the AI hardware/semiconductor/AI PC theme with higher potential volatility, consider the Hang Seng Stock Connect Information Technology Index (highest purity in hard tech, up 34.89% over three months).

For a broad bet on the recovery of China's major tech giants, including new energy vehicle players, the Hang Seng Tech Index is an option.

For a focus on the recovery of internet platform businesses (e-commerce, local services, advertising), consider the Hang Seng Stock Connect Internet Index.

For a more balanced approach across tech sectors, the Hang Seng Stock Connect Technology Index may be suitable.

Among ETFs tracking the Hang Seng Stock Connect Information Technology Index, the HuaBao Hang Seng Stock Connect Information Technology ETF stands out in terms of size and average daily trading volume, and is currently the only such ETF with an associated feeder fund.