DeepSeek: Nvidia and TSMC Poised to Benefit from Rising AI Inference Demand

Piper Sandler analysts have identified Nvidia and Taiwan Semiconductor Manufacturing Company (TSMC) as the biggest beneficiaries of the surging demand for inference compute power, driven by the rapid expansion of AI applications.

Following discussions with a data center engineering expert, the analysts reaffirmed their belief that AI infrastructure spending will continue to rise at a significant pace.

"As inference applications for large language models (LLMs) keep expanding, the demand for compute power and capital expenditure requirements will only increase," the expert explained.

While DeepSeek’s advancements have captured industry attention, Piper Sandler noted that top AI labs like OpenAI and Anthropic still maintain a six-month lead in development.

Inference Costs Outpacing Training Investments

One of the most crucial takeaways from the discussion is that inference—rather than training—is becoming the primary cost driver for AI labs.

"Inference, which serves as the revenue-generating mechanism for most AI labs, is responsible for driving capital expenditures higher, and there are no shortcuts in deploying inference compute infrastructure," the expert stated.

Currently, OpenAI and Anthropic allocate roughly 75% of their compute resources to inference, a figure expected to rise to 90% as training stabilizes.

The analyst report underscores Nvidia's dominant position in this shift. "According to our expert, Nvidia, followed by TSMC, stand out as the clear winners for capital expenditure dedicated to inference," the note highlights.

Additionally, the report emphasizes that networking, particularly optical solutions, will play a crucial role in supporting AI's increasing inference requirements.

DeepSeek’s Cost Efficiency Under Scrutiny

Despite growing excitement around DeepSeek, analysts were less impressed with its cost efficiency when compared to OpenAI.

"A similar training run on OpenAI’s infrastructure would have cost approximately $10 million, making DeepSeek’s cost efficiencies less remarkable relative to its U.S. counterparts," the expert observed.

Nvidia and TSMC Remain Top AI Investment Picks

With inference applications expanding rapidly, analysts remain bullish on Nvidia and TSMC, viewing them as key beneficiaries of AI’s ongoing evolution. As demand for AI-driven computing power accelerates, these two companies are expected to play a leading role in shaping the future of AI infrastructure.

(Sources: investing.com, reuters.com, ChatGPT)