Chinese Tech Giant’s Strategy Head Claims DeepSeek Reduces Need for NVIDIA AI GPUs

The following content is not intended as investment advice. The author holds no positions in any mentioned stocks.

As U. S.megacap corporations pour billions into capital expenditures for AI GPUs, James Mitchell, the Chief Strategy Officer at Tencent, suggests that the recent advancements from DeepSeek may indicate that such immense spending could be unnecessary. DeepSeek claims to have crafted AI models that rival those of leading American companies, but at a dramatically reduced cost, a claim that has profoundly influenced the stock market landscape.

Perhaps the most affected entity is NVIDIA, which has struggled to recover from nearly $600 billion in losses since the January selloff. Despite efforts during the recent GTC conference led by CEO Jensen Huang to highlight potential trillion-dollar markets for NVIDIA’s offerings, investor sentiment remains cautious, and the stock price has stagnated.

Impact of DeepSeek’s Innovations on Chinese Technology Firms’ GPU Spending

In a recent discussion, Tencent’s CSO, James Mitchell, articulated that one of the primary reasons for investing in NVIDIA’s AI GPUs was the need for training large language models (LLMs).Shortly after DeepSeek’s technologies gained traction in January, Tencent introduced its Hunyuan Turbo S AI model, claiming response times of under a second, tailored for the Chinese market.

While seeking to outpace DeepSeek in developing superior AI models, Tencent acknowledges that DeepSeek’s innovative training techniques have significantly lowered AI development expenses. By leveraging sophisticated software engineering, DeepSeek reportedly enhances efficiency while minimizing costs associated with AI model training. Traditionally, engineers have relied on NVIDIA’s CUDA software to utilize GPUs effectively, but this has often meant compromising on fine-tuned control over their products.

Regarding capital allocation, Mitchell pointed out that investments in GPUs for training large language models had been vital prior to DeepSeek’s revelations. He recalled a time last year when there was a prevailing belief that each new generation of LLM necessitated significantly more GPUs. However, he indicated that DeepSeek has shifted this narrative, particularly among Chinese tech firms. As Mitchell noted, “that period of time ended with the breakthroughs that DeepSeek demonstrated.”

He revealed that, following these advancements, “the industry is now achieving much higher productivity for LLM training using existing GPUs, eliminating the need to acquire additional GPUs at the previously anticipated rate.”Notably, due to restrictions on purchasing NVIDIA’s latest AI GPUs, including the Blackwell and Hopper products, Chinese companies are compelled to rely on older GPU models or large clusters to mitigate limited computing resources.

Tencent has proclaimed that its Turbo S model excels in mathematics, reasoning, and other AI functionalities compared to DeepSeek’s offerings. Industry insiders suggest that Chinese companies might consider partnerships with Huawei and its Ascend AI chips while navigating the ongoing chip embargo.

While like NVIDIA, Huawei provides its chip users with software to manage the chips, reports indicate that DeepSeek found Huawei’s software performance lacking compared to NVIDIA’s solutions. In the meantime, NVIDIA shares continue to flounder, down 14% year-to-date, as investors await more conclusive data to stimulate demand. Tencent, while traded on OTC markets, boasts an impressive market capitalization of $601 billion.

James Mitchell of Tencent quite clearly laying out the bear case for $NVDA – DeepSeek’s innovations mean fewer GPUs needed vs.what Tencent previously thought, despite the explosion in AI usage.@doodlestein

“There was a period of time last year when there was a belief that… pic.twitter.com/f80UXyQ8Ny

— Timothy Liu (@timothyhliu5) March 20, 2025