Bullish on Chinese AI chips! Goldman Sachs "again" raised the target prices for SMIC and HUA HONG SEMI

Goldman Sachs' latest research report has once again cast a heavy "bullish" vote on China's semiconductor industry, raising the target prices for SMIC and Hua Hong Semiconductor for the fourth time in the past month, believing they will benefit in the long term from the growth in domestic AI-driven chip demand.

According to news from the Chasing Wind Trading Desk, Goldman Sachs, in a report released on October 5 by analysts Allen Chang and others, raised the 12-month target price for SMIC's H-shares to HKD 117.0 and the A-shares target price to RMB 211.0; at the same time, it significantly raised the target price for Hua Hong Semiconductor by 34% to HKD 117.0, maintaining a "buy" rating for both companies.

The report pointed out that DeepSeek recently released its experimental new model DeepSeek V3.2-Exp, which significantly reduced training and inference costs, leading to a decrease of over 50% in its API (Application Programming Interface) fees. Goldman Sachs believes this will effectively lower the application threshold for AI technology, promoting its broader commercial implementation.

In addition, the explosive growth of AI applications in China will generate massive demand for power management chips (PMIC) for data centers, Bluetooth/WiFi related to AI devices, image sensors (CIS), radio frequency (RF), and microcontrollers (MCU).

AI Model Cost Revolution Ignites Downstream Application Demand

The bullish logic of the report begins with technological breakthroughs in China's AI models. Recently, DeepSeek released its latest experimental model DeepSeek V3.2-Exp. This model introduced the innovative DSA (DeepSeek Sparse Attention) technology, significantly improving training and inference efficiency when processing long texts.

The most direct result is a substantial reduction in costs. According to the company's announcement, its API (Application Programming Interface) costs have decreased by 50% to 75%. Specifically, its input cost has now dropped to RMB 0.2-2 per million tokens, while the output cost is RMB 3 per million tokens.

Goldman Sachs believes that the launch of this efficient AI model will greatly lower the usage threshold for AI technology and promote its widespread adoption. The decrease in costs means that more enterprises and developers can integrate and use large models in practical business, which will directly translate into incremental demand for computing infrastructure and related hardware.

What makes Goldman Sachs even more optimistic is that China's local AI ecosystem is showing strong synergies. With the release of the new DeepSeek model, local chip suppliers, including Cambricon, Huawei Ascend, and Haiguang Information, quickly announced that their products have been adapted to DeepSeek V3.2-Exp.

This close cooperation between chip suppliers and model developers has formed a rapidly iterating development closed loop, ensuring that chip computing power is optimally utilized. In addition, AI server suppliers, such as Huawei and Alibaba, are working to develop more efficient networking capabilities through technologies like SuperPod, to connect and utilize more chips in AI training, thereby compensating for the computing power limitations of a single chip.

Steady Capacity Expansion, Long-term Benefits for Two Major Foundries

Although SMIC and Hua Hong Semiconductor's core capacities are focused on mature processes rather than the most advanced AI GPUs or ASICs, Goldman Sachs believes that both companies will still gain long-term growth momentum from AI-driven demand increases. The report analyzes that the proliferation of AI will drive an increase in demand for a range of peripheral chips. This includes power management chips (PMIC) for data centers, as well as Bluetooth/WiFi, image sensors (CIS), radio frequency (RF), and microcontrollers (MCU) for various AI devices. Most of these chips are produced using mature processes, which are the core of the two major foundries' businesses.

To meet the growing demand, both companies are steadily expanding their capacities and upgrading their technologies. Goldman Sachs points out that SMIC is expanding its 7nm/14nm capacity, while Hua Hong Semiconductor has announced plans to migrate to 28nm in its next foundry and may further advance its technology in the future.

Based on optimistic expectations for the development of China's AI ecosystem, Goldman Sachs believes that the market is re-evaluating the valuations of Chinese semiconductor companies and has updated its valuation model accordingly. For Hua Hong Semiconductor, its target price has been raised from HKD 87.50 to HKD 117.0, an increase of 34%. The new target price is derived from a discounted expected price-to-earnings ratio of 68.8 times for 2028 (previously 51.5 times). Goldman Sachs stated that the higher valuation multiple reflects the "ongoing re-evaluation of Chinese semiconductor companies."

For SMIC, its H-share target price has also been raised to HKD 117.0 (previously HKD 95.0), with the valuation model similarly based on the updated expected price-to-earnings ratio of 62.9 times for 2028, aiming to reflect the "ongoing re-evaluation of the Chinese semiconductor sector." At the same time, the firm has set the target price for SMIC's A-shares at RMB 211.0 based on an average A-H share premium of 196%