The AI Revolution’s New Challenger

DeepSeek, a startup founded in 2023 by Liang Wenfeng, is challenging industry giants with a bold claim: their large language models (LLMs) deliver performance comparable to the best in the market—at a fraction of the cost.

  • The company's rise comes at a time of rapid expansion for AI, where efficiency and scale are increasingly in tension.
  • Wall Street analysts are now weighing the potential consequences of DeepSeek’s advancements, debating whether its breakthroughs represent a fundamental shift or merely a temporary ripple in the sector.

Innovation or Overstatement

DeepSeek’s models, which include reasoning capabilities designed to articulate their decision-making processes, are open source and have rapidly gained traction. Their mobile app has topped download charts in major markets, including the U.S., UK, and China. However, some analysts question whether DeepSeek’s purported cost reductions are as dramatic as advertised.

  • Citi analysts, for instance, expressed skepticism, noting that high-performance LLMs typically rely on advanced GPUs to achieve their results. “While DeepSeek’s achievement could be groundbreaking, we question the notion that its feats were done without the use of advanced GPUs to fine tune it and/or build the underlying LLMs the final model is based on through the Distillation technique.” Citi remarked in a note to clients.
  • Similarly, Bernstein analysts were quick to dampen market hype, cautioning that while efficiency gains are essential to sustaining AI progress, the idea that DeepSeek has revolutionized the cost structure of AI development may be overstated. “If we acknowledge that DeepSeek may have reduced costs of achieving equivalent model performance by, say, 10x, we also note that current model cost trajectories are increasing by about that much every year anyway (the infamous “scaling laws…”) which can’t continue forever. In that context, we NEED innovations like this (MoE, distillation, mixed precision etc) if AI is to continue progressing.”

The Jevons Paradox

Efficiency gains in AI, ironically, may fuel greater demand rather than curb it—a phenomenon reminiscent of the Jevons paradox in economics. OpenAI’s introduction of ChatGPT Pro in 2024, which delivered an 86% accuracy on math problems, drove a surge in token consumption despite its improved efficiency. Within a year, weekly token usage soared from 8 billion to over 300 billion, enabling new applications like continuous-operation AI agents and automated workflows.

  • DeepSeek’s potential to make AI even more accessible could further amplify this trend. Bernstein analysts underscored that "[...] for those looking for AI adoption, as semi analysts we are firm believers in the Jevons paradox (i.e. that efficiency gains generate a net increase in demand), and believe any new compute capacity unlocked is far more likely to get absorbed due to usage and demand increase vs impacting long term spending outlook at this point, as we do not believe compute needs are anywhere close to reaching their limit in AI.
  • The semiconductor industry thus remains at the heart of AI’s expansion. Analysts generally agree that even with efficiency gains, demand for chips will remain robust as AI adoption spreads across industries and devices.

Morgan Stanley also pointed to the possibility of AI models running on smaller hardware, from office computers to smartphones, as a potential growth driver for semiconductor manufacturing equipment. UBS analysts noted that lower AI training costs could spur adoption among retail users, diversifying demand for data centers and enhancing pricing power for infrastructure providers in high-demand locations.

The Chips Diplomacy

DeepSeek’s rise also underscores China’s growing prominence in the AI landscape. Analysts from Goldman Sachs highlighted the potential for Chinese companies to expand their influence, particularly in consumer-facing applications. Companies like ByteDance have already launched dozens of AI-powered tools, with applications like Doubao gaining traction among millions of users.

  • However, geopolitical tensions loom large. Jefferies analysts pointed out that U.S. export restrictions on advanced chips may inadvertently accelerate Chinese innovation. “China is the only market that pursues LLM efficiency owing to chip constraint. Trump/Musk likely recognize the risk of further restrictions is to force China to innovate faster. Therefore, we think it likely Trump will relax the AI Diffusion policy.”
  • Such developments could pressure the U.S. to reassess its policy of tariffs and export restrictions on advanced chips. By inadvertently driving rapid Chinese innovation, these measures may undermine their intended effect and weaken the strategic leverage they were designed to provide.

For now, DeepSeek’s emergence highlights a broader shift in the global AI landscape, challenging assumptions about technological dominance and economic leverage.

Disclaimer

Please note that Benchmark does not produce investment advice in any form. Our articles are not research reports and are not intended to serve as the basis for any investment decision. All investments involve risk and the past performance of a security or financial product does not guarantee future returns. Investors have to conduct their own research before conducting any transaction. There is always the risk of losing parts or all of your money when you invest in securities or other financial products.

Credits

Photo by Christian Wiediger / Unsplash.