Martin Lau, President of Tencent, has confirmed that the company has built up a ´pretty strong stockpile´ of NVIDIA Artificial Intelligence GPUs, a move designed to buffer its operations against the effects of recent US export controls targeting high-end compute hardware. Speaking during a mid-week earnings call, Lau indicated that these chips will be essential as Tencent prepares to activate a new Artificial Intelligence strategy. The admission stems from mounting concerns over the US Bureau of Industry and Security (BIS) tightening licensing requirements, which have curtailed supplies of potent GPUs such as the NVIDIA H20 to Chinese firms.
Lau detailed the company´s response to this rapidly shifting regulatory climate, describing how Tencent prioritized gathering GPUs ahead of anticipated bans and regulatory adjustments. He outlined a two-pronged approach to leveraging their GPU reserves: immediate deployment in revenue-generating applications like digital advertising and content recommendation engines, followed by the training of Tencent’s proprietary large language models—tasks that depend on high-end compute resources. Importantly, Lau signaled confidence that Tencent’s current inventory is sufficient to sustain Artificial Intelligence model training for several product generations.
Beyond NVIDIA, Tencent is keeping a close watch on the evolution of Chinese-made alternatives such as Huawei’s Ascend 910C accelerator, considered a leading domestic competitor as Artificial Intelligence chip development accelerates in China. Such homegrown solutions are gaining traction amid tighter US export controls and the growing influence of local Artificial Intelligence ecosystems, exemplified by the rise of DeepSeek. Lau also noted a strategic pivot away from the ´scaling law´ philosophy dominant in US tech circles, suggesting that effective model training does not always require ever-larger clusters of chips. This shift, combined with Tencent’s robust pre-ban stockpile, places the company in a relatively secure position to pursue its Artificial Intelligence ambitions amid ongoing geopolitical and regulatory headwinds.
