What Is the Impact of DeepSeek in the AI Space?

By Daniel Morgan, Synovus Trust Senior Portfolio Manager
Synovus Trust Company, N.A.

The news about DeepSeek’s launch of its latest AI models created seismic waves throughout the Technology Sector. Reports that the AI Assistant — powered by DeepSeek-V3 — has rivaled ChatGPT to become the top-rated free application available on Apple's App Store in the U.S. sent doubts through investor’s minds on how a less expensive AI model may impact this red-hot sector. It is reported that DeepSeek-V3 and DeepSeek-R1 are on par with OpenAI and Meta's most advanced models.

DeepSeek’s introduction has raised questions about the “AI Trade” among investors as it raised worries that an AI model could be trained for a fraction of the cost of other models from — OpenAI and Meta. This created concerns that DeepSeek’s models could cut demand for data centers and high-end chips. DeepSeek’s popularity has raised concerns that it could impact the demand for the hefty computation needs required for the Infrastructure-as-a-Service (IaaS) data providers, like Amazon’s AWS, Microsoft’s Azure and Alphabet’s GCP. DeepSeek is expected to use less memory for AI computation. That could impact the long-term demand for High Bandwidth Memory (HBM) used for large GPU-clusters (manufactured by the likes of Micron Technology). The real money in AI is providing the chips that power the massive data centers from the likes of Nvidia, AMD and Broadcom. Further, there have been concerns that Nvidia would be impacted as well, as DeepSeek requires less computation capacity.

Bear in mind that DeepSeek’s AI Assistant is focused mostly on mobile, personal computer and laptop devices. So we could see some threat from DeekSeek’s AI models to Microsoft’s ChatGPT and Alphabet’s Gemini. Alphabet’s Gemini, which runs on both iOS and Android, would most likely be impacted by the less expensive DeepSeek’s AI Assistant.

Putting all these concerns aside, DeepSeek’s biggest impact has been to push the AI industry even harder toward resource-intensive reasoning models, shifting away from conventional large language models. Reasoning models, which are based on large language models, use far greater resources in terms of microchips and electricity. Further, it allows AI technology to be used by a greater mass of potential users, thus increasing overall demand.

DeepSeek AI’s reported development, which uses a fraction of the resources markets assumed needed, was originally interpreted as a potential “eureka moment” for China’s burgeoning technology capabilities. However, at this point we stand unconvinced that meager resources DeepSeek is credited with possessing in development of CapEx, chip count, talent pool, for instance, is enough to completely dislodge the AI CapEx Investment Wave.

In summary, the top four hyperscalers — Amazon, Microsoft, Meta and Alphabet — are collectively estimated to spend $240 billion (+60% Year-over-Year) in CapEx in 2025 to build out their AI presence. Meta announced during its recent 4Q24 earnings call that “We anticipate our full year 2025 capital expenditures will be in the range of $60-65 billion (an increase from $40 Billion). We expect capital expenditures growth in 2025 will be driven by increased investment to support both our generative AI efforts and core business.” This runs completely contradictory to concern’s that CapEx would be adjusted lower to take advantage of the less expensive DeepSeek AI Models. Weighing all the facts at this point, we see no negative change in current AI spend rates from the impact of DeepSeek’s Open AI Models.