DeepSeek’s AI Revolution: Will TSMC, Nvidia, and Big Tech Hold Their Ground or Lose the Throne?

DeepSeek’s Development Journey and Its Distinctive Open-Source Strategy

DeepSeek quickly captured global attention with its low-cost deployment and open-source strategy. Unlike traditional AI services tied to Big Tech giants like Google, Microsoft, and Amazon, open-source frees users from dependence on these companies or the need to pay for their services. Instead, DeepSeek offers developers and consumers unrestricted access to top-

tier AI tools, making innovation more affordable, adaptable, and accessible. The story of DeepSeek begins with its CEO, Liang Wenfeng, whose bold vision transformed his journey from finance to AI. In 2021, Liang made a decisive move—purchasing 10,000 Nvidia H800 chips in preparation for his next big AI venture just before the H800 faced export restrictions. By 2023, DeepSeek AI was officially established, gathering a team of 200+ PhDs to push the limits of AI development. In 2024, the company achieved a major milestone by building Large Language Models (LLMs) at only one-twentieth of OpenAI’s cost, setting a new benchmark for efficiency in the AI industry. The table below provides a detailed comparison, showcasing how DeepSeek maintains its competitiveness with OpenAI by strategically utilizing H800 chips, a downgraded version of Nvidia’s H100.

An overview of 2 main competitive Models: DeepSeek V.S OpenAI 

DeepSeek’s Challenge to Big Tech and the Resilient Stronghold of Semiconductor Giants:

Market Share: Nvidia - 90%, ASML - 100%, TSMC - 90%

DeepSeek’s rise is disrupting Big Tech’s monopoly, while Nvidia and TSMC continue to solidify their dominance in the semiconductor supply chain. By adopting an open-source approach, DeepSeek is reshaping the AI industry and directly challenging the control that companies like Google, Microsoft, and Amazon have maintained over 70% of the global market through their cloud services. Historically, developers and consumers had little choice but to depend on these tech giants for AI computing resources.However, DeepSeek’s low-cost, open-access models are changing the game. By allowing developers to run advanced AI applications on mid-range GPUs, DeepSeek removes the need for expensive cloud services, fueling the growth of edge AI and opening the door to independent innovation. This shift is giving smaller companies and individual developers the tools to innovate without being locked into Big Tech’s ecosystem.

Despite the disruption DeepSeek brings, Nvidia and TSMC still hold critical positions in the global semiconductor industry. Nvidia controls 85-90% of the chip design market, producing essential AI chips like A100 and H100, which are crucial for AI development. Besides, ASML, with its 100% market share in semiconductor equipment manufacturing, is the only company capable of producing EUV lithography machines, essential for advanced chip production. Meanwhile, TSMC dominates AI chip manufacturing with a 90% market share in the foundry sector. Even with DeepSeek pushing to democratize AI tools, the dominance of Nvidia, TSMC, and ASML ensures their central role in the industry’s infrastructure. However, DeepSeek’s growing influence has also caught the attention of the U.S. government, prompting stricter export restrictions on Nvidia’s chip sales to China. 

Resources: https://www.bnext.com.tw/article/82142/ai-market-success-keypoint

Navigating U.S. Trade Restrictions: Nvidia’s Adaptive Strategy

The U.S. government-imposed export bans on Nvidia’s A100 and H100 chips in September 2022, aiming to prevent their use in military applications in China. In response, Nvidia developed A800 and H800, reduced-performance versions to comply with regulations. These models were later banned in October 2023 as the U.S. tightened restrictions.

Despite these challenges, Nvidia’s strategy to develop compliant AI chips for the Chinese market has paid off. H20, L20, and L2 are projected to generate $12 billion in revenue in 2024, surpassing expectations and cementing Nvidia’s position in the market. (Refer to the table for a detailed timeline of the restrictions.)

Conflicts Between Nvidia and U.S. Updated Regulations

Resources: https://tecex.com/ai-chips-trade-wars-sanctions/

Although complying with U.S. export restrictions, Nvidia has continuously run into conflicts over its push to keep supplying China with downgraded versions of its chips. The company has argued that losing access to the Chinese market would carry serious economic consequences and hurt U.S. leadership in AI. Nvidia’s CFO, Colette Kress, openly warned that these sanctions could result in a “permanent loss of opportunities” for the company. In response to the tightened regulations, Nvidia introduced the H20, L20, and L2 chips to navigate these challenges. However, this move has sparked ongoing debates about how to balance national security concerns with maintaining global competitiveness. Nvidia’s determination to remain in the Chinese market reflects just how critical China is to its overall revenue.

The Shifting Battlefield – From Hardware Control to Software Dominance

As DeepSeek’s open-source models continue to grow, the focus in AI is shifting from hardware dominance to software and cloud services. This open-source movement, driven by global developers, is challenging Big Tech’s centralized control. Nvidia and TSMC remain strong players in the hardware sector, but the real question lies in whether Big Tech can adapt to this evolving, decentralized AI environment. Restricting chip supply may work as a short term measure, but it also highlights the need for Big Tech to develop longer term strategies in response to these changes.

In light of this, the U.S. government tightened regulations on Nvidia’s exports to China, raising broader questions about how open-source technology combined with downgraded but still powerful chips like H800 could influence AI development over time. While the full extent of this transformation remains uncertain, the rise of DeepSeek represents a significant shift, one that may gradually reshape how AI innovation unfolds and is shared worldwide.