GTHT has released a research report initiating coverage on ILUVATAR COREX (SEHK: 09903) with an 'Add' rating.

The report states the company is a leading domestic provider of general-purpose GPU products and AI computing power solutions, with its business focused on AI training GPUs, AI inference GPUs, AI computing servers/clusters, and supporting software ecosystems.

The firm projects the company's revenue for 2026 through 2028 to be RMB 2.578 billion, RMB 5.002 billion, and RMB 8.323 billion, respectively.

Using a price-to-sales ratio valuation, the report assigns the company a 50x PS multiple for 2026, corresponding to a target market capitalization of RMB 128.9 billion, or approximately HKD 149.393 billion, leading to a target price of HKD 588.16.

The report notes that ILUVATAR COREX achieved high revenue growth last year, with its general-purpose GPU products becoming the primary growth driver.

In 2025, the company reported revenue of RMB 1.034 billion, a year-on-year increase of 91.6%; gross profit of RMB 558 million, up 110.5% year-on-year; and an adjusted net loss of RMB 438 million, narrowing by 32.1% compared to the previous year.

Within this, revenue from general-purpose GPU products reached RMB 923 million, surging 149.6% year-on-year, and its contribution to total revenue increased to 89.3%.

GTHT indicated that the expansion in AI computing power demand is benefiting domestic general-purpose GPUs, driven by strong industry momentum and an acceleration in domestic substitution.

Citing Frost & Sullivan, the report states that China's general-purpose GPU market revenue is projected to grow from RMB 154.6 billion in 2024 to RMB 715.3 billion in 2029.

The compound annual growth rate from 2025 to 2029 is forecasted at 29.5%, with the share of domestic general-purpose GPUs expected to reach 31% by 2029.

The company continues to adhere to its "software-hardware co-design" philosophy, advancing the iterative upgrades of its TianGai and ZhiKai product series.

It has released a new-generation software development platform that natively supports mainstream industry GPU programming models, providing software support for unleashing the computing power of its self-developed general-purpose GPU chips and enabling large-scale commercial deployment, having completed adaptation for over 610 algorithm models through its DeepSpark open-source community.