DeepSeek: Stop the Panic

  • DeepSeek is a precursor of many more disruptors coming in the AI world. Its importance may be overblown compared with what comes next.

  • What’s at stake: DeepSeek nearly sank Nvidia and other AI model players. Except DeepSeek itself is barely known by anyone, its true story still a mystery, and the likely impact on the artificial intelligence market undetermined. It won’t be the last enterprise to put a dent on the AI market, however, according to Market Traction’s Steve Carr who says many more such surprises are waiting in the wings, both in the West and in China.

The race for leadership of the artificial intelligence market just got scrambled. The debut of China’s DeepSeek AI model sparked a selling frenzy on Wall Street as spooked investors dumped Nvidia and other AI-related stocks. The implications of DeepSeek’s platform are not clear, however. Not enough is known about how it was developed, meaning a full assessment of its market implications are months away.

What is clear is that DeepSeek’s AI model isn’t a result of any mega innovation in semiconductor or software processing, considering the Chinese company is leveraging resources from other models, including ChatGPT, and depended on chips from Nvidia Corp., the market’s leading vendor of AI semiconductors. This doesn’t take away from the disruptive effects DeepSeek has had on the market already, though. Other questions were raised in a webcast discussion with Steve Carr, founder of Markettraction.io and TalkingIoT.

For example, what are the geopolitical implications of DeepSeek’s launch, and how can enterprises outside China best position themselves to navigate the evolving AI development dispute between the US and China? Are investments in Nvidia, Advanced Micro Devices, and other AI accelerator vendors still worth the elevated valuations?

Calm down, said Carr during a chat with the Ojo-Yoshida Report. First, there’s a lot more to the story of how DeepSeek was able to achieve its performance against the OpenAI model, Carr said. In fact, evidence is growing that DeepSeek leveraged data it gleaned from other models to train its own. Which means the lower costs reported by DeepSeek may not be quite what they seem. He also questioned the timing of the announcement of the DeepSeek model, coming as it did on the same day as the US was swearing in a new president. In essence, conspiracy theories aside, there’s a lot more to the DeepSeek saga than meets the eye.

But let’s dig deeper. Does the release of DeepSeek mean the end of the hefty margins Nvidia has been reporting? Carr believes the only certainty we have now is that AI has become the new Wild, Wild West, marked by uncertainties and rapid change.

That’s the puzzle facing investors. Should they keep pouring billions of dollars into AI accelerators or should they wait for the inevitable market shakeout? Also, what are the implications of DeepSeek’s apparently lower costs – and those of others unknown – to semiconductor vendors? In addition, must the industry trim down currently high sales forecasts for processors and high bandwidth memory (HBM) or are the numbers safe?

DeepSeek or deep-six?

DeepSeek, trained on OpenAI’s ChatGPT, cost only $5.6 million to develop using 2,000 Nvidia chips, contrasting with OpenAI’s $120 million for GPT-4.

This revelation led to a $1 trillion loss in US AI-related stocks. Just ahead of this wobbly performance, Donald Trump’s US administration had announced a $500 billion investment fund in AI led by Softbank and OpenAI. DeepSeek put a hole under those numbers by implying that America may not need such lofty figures to maintain its leadership of the AI market. It also implies that perhaps America should be humbler and stop thinking it is the world’s AI leader, able to shut out others from the sector.

Carr isn’t interested in the geopolitics of AI. He believes AI is in its infancy with many iterations to come. Furthermore, he suggests we stop focusing on DeepSeek and look under the sofa. Many other companies that can match or beat DeepSeek’s performance are already at work in somebody’s garage – in the West. Open your eyes to new possibilities, Carr says.

In fact, DeepSeek may have been launched and announced to tamp down America’s AI enthusiasm, he notes. Otherwise, why announce this disruptive technological advancement on the date a preening president was being sworn into office?

DeepSeek’s success, for Carr, highlights the potentials for “democratizing AI”, with other Chinese companies and other enterprises likely to follow suit. The conversation with Carr also touched on the broader implications for AI development, the need for significant investment, and the competitive landscape between Western and Chinese AI enterprises.

The potential geopolitical implications of DeepSeek’s launch are significant, he said. The following key points are important, according to Carr:

  1. DeepSeek’s rapid development and low-cost model could be seen as a challenge to American AI dominance. This could heighten tensions and spur increased competition and investment in AI between the US and China;
  2. The timing of DeepSeek’s launch, coinciding with the US presidential inauguration, suggests it may be a strategic move to assert China’s capabilities in this space. This could escalate the AI “arms race” between the two superpowers.
  3. The ability of a Chinese company to rapidly develop an AI model that can rival established Western players like OpenAI raises concerns about technology transfer and the potential for China to leapfrog the US in certain AI capabilities.

Bottom line

DeepSeek stirred the market and called into question the models behind ChatGPT and other AI models. AI is still in its infancy, however, and many more surprises lay ahead. How semiconductor suppliers will fare is still an open question.

Watch the discussion with Carr below:

Bolaji Ojo
Publisher and Managing Editor, Ojo-Yoshida Report


This article was published by the The Ojo-Yoshida Report. For more in-depth analysis, register today and get a free two-month all-access subscription.

Here you’ll find a wealth of practical technical insights and expert advice to help you bring AI and visual intelligence into your products without flying blind.

Contact

Address

Berkeley Design Technology, Inc.
PO Box #4446
Walnut Creek, CA 94596

Phone
Phone: +1 (925) 954-1411
Scroll to Top