DeepSeek gained this level of recognition because, while its competitors spent billions on building, training, and enhancing their models before offering them at a higher cost, the startup developed and trained its model at a fraction of the cost, kept it open-source, and made it available to the public at a significantly lower price
The sudden rise of China’s AI start-up DeepSeek took the entire tech industry by surprise, especially the giants sitting in California’s Bay area. Founded in 2023, the AI start-up launched its new large language model (LLM), DeepSeek- R1 in January 2025. This model shook the AI ecosystem, witnessing an unprecedented and exponential growth.
At a time when almost all of the substantial advancements in the AI industry were coming from the US, a two year old East Asian AI start-up came out of nowhere causing a disruption in the market.
Imagine the level of disruption- within a week of its release, DeepSeek topped the Apple Store charts in the US and caused Nvidia to lose $600 billion in value, marking the largest single-day financial collapse in history.
DeepSeek gained this level of recognition because, while its competitors spent billions on building, training, and enhancing their models before offering them at a higher cost, the startup developed and trained its model at a fraction of the cost, kept it open-source, and made it available to the public at a significantly lower price. DeepSeek made its model more collaborative and economical without compromising on the quality and efficiency of the model.
The "Sputnik moment" refers to a pivotal event that sparks a nation's realisation that it is falling behind in a critical area, leading to urgent efforts to catch up or surpass the competition. The term originates from the Soviet Union's launch of Sputnik 1 in 1957, the world's first artificial satellite to orbit the Earth. This unanticipated success not only marked the rise of the space age but also rattled the US, which had assumed its technological superiority. This moment spurred massive investments in science, technology, and education, leading to innovations like the Apollo program. Today, the phrase is used metaphorically to describe any wake-up call that comes from an unprecedented advancement in a field.
Similarly, before 2025, the US was too confident about its global AI dominance, fairly because the monster models of the world were based out of the country. Tech experts and major US publications had earlier labeled the US AGI (Artificial Generative Intelligence) system as “well ahead of the game.”
In November 2024, A Stanford University index revealed that the US is way ahead of China in AI development, leading in research, investment, and innovation. Computer scientist Ray Perrault noted the gap between the US and China was widening due to increased US investment.
The US had a simple strategy to maintain its edge over competitors, especially China. It kept its most powerful models closed source and, in addition, restricted China from sourcing Nvidia computing chips. The vision behind this strategy was to curb China’s attempts to replicate US models, thereby avoiding the popular quip, ‘America innovates and China imitates.’
However, this strategy backfired as it compelled China to devise methods for building models of comparable strength to those of the giants, but at a fraction of their cost. This led to China releasing DeepSeek’s models, which were open source, available at a much lower cost than standard models, and designed using an algorithm that ran efficiently on limited chips.
DeepSeek's latest LLMs are on par with models of AI giants like OpenAI, Google DeepMind, xAI, and others. DeepSeek’s R-1 and V-3 models have outperformed OpenAI’s GPT-4o and O3 Preview, Google’s Gemini Pro Flash, and Anthropic’s Claude 3.5 Sonnet across various benchmarks.
The rise of DeepSeek is being termed as AI's Sputnik moment. Silicon Valley investor Marc Andreessen popularised the phrase through an X post that read, “One of the most amazing and impressive breakthroughs I’ve ever seen,” he tweeted, before adding that it was “AI’s Sputnik moment”.
In 2017, China unveiled a strategic plan called the "New Generation Artificial Intelligence Development Plan" with the explicit goal of becoming the global leader in artificial intelligence (AI) by 2030; this plan outlined ambitions to advance AI technology across various sectors like industry, governance, and defense, establishing China as a major AI innovation hub.
At the inaugural meeting of the Association for the Advancement of Artificial Intelligence in 1980, no papers were submitted by Chinese researchers. However by 2018, China had submitted 25% more papers than the US.
This shows that the sudden rise of DeepSeek is not a mere coincidence, but a result of constant research and investment.
China's AI success is driven by its emphasis on developing and retaining talent. AI education is integrated into university curricula, while research institutes ensure a steady pipeline of skilled professionals. Initiatives like the ‘Thousand Talents Plan’ attract overseas experts, while substantial funding and improved research infrastructure encourage domestic talent to stay and contribute.
According to a 2023 MacroPolo report on global AI talent, nearly half of the world’s top AI researchers completed their undergraduate studies in China. Chinese universities, state-backed labs, and research divisions of American tech giants such as the Beijing-based Microsoft Research Asia have played a crucial role in training a large pool of local researchers.
Additionally, a growing number of China’s AI elite are turning down opportunities in Silicon Valley in favor of the domestic AI industry, which offers lower living costs, proximity to family, and the chance to take on significant roles early in their careers. The country’s tech talent is increasingly uninterested in full-time jobs in the U.S.
For instance, Zizheng Pan, a young AI researcher from China, interned at Nvidia in 2023. Upon completing his internship, he declined a full-time position at Nvidia and instead joined an emerging Chinese startup called DeepSeek.
Government and private-sector collaboration has accelerated AI innovation. Leading tech firms like Alibaba, Baidu, and Tencent align their research efforts with national goals, driving advancements in machine learning.
Alibaba has pledged over $52bn to AI and cloud computing, reinforcing strong public-private collaboration. Meanwhile, AI startups like DeepSeek are developing competitive models, solidifying China's position in the global AI race.
DeepSeek’s founder, Liang Wenfeng, received crucial support from the Chinese government that accelerated the company’s growth. The government provided high-level political endorsement, strengthening DeepSeek’s credibility by facilitating meetings with top leaders as well as infrastructure support through initiatives like the "Eastern Data, Western Computing" project, which offered access to state-subsidised, cost-effective data centres and computing resources.
Additionally, a favorable policy environment with subsidies and reduced regulatory barriers helped attract investment and lower operational costs, enabling DeepSeek to scale rapidly and compete globally.
China’s strategy in the global AI race relies heavily on developing cost-effective, open-source AI models that democratise access and drive rapid innovation. By training models like DeepSeek at a fraction of the cost of Western counterparts, China is overcoming US export controls on advanced chips, accelerating adoption across government, industry, and research, and reducing reliance on expensive foreign technology.
The pricing structure shows a significant difference, with DeepSeek-V3 being substantially more economical at $0.14 per mn input tokens and $0.28 per mn output tokens, compared to o1 Preview's $15.00 and $60.00 respectively.
This approach not only fuels widespread diffusion and collaboration within its domestic AI ecosystem but also supports China’s long-term goal of achieving global AI leadership by 2030.
The US government has imposed export restrictions on AI chips to China to limit its access to advanced technology, particularly GPUs. This strategy is designed to maintain US dominance in the AI chip industry while preventing Chinese companies and the military from acquiring cutting-edge hardware.
To regulate GPU exports, the US has established a tiered system. Tier 1 includes the US and 18 allied nations, such as the UK, Japan, and EU member states, which have unrestricted access to advanced AI chips. Tier 2 covers most other countries, allowing them to purchase up to 100,000 GPU units. Meanwhile, Tier 3, which includes China and Russia, is entirely banned from importing advanced GPUs.
DeepSeek primarily runs on Nvidia H800 GPUs, modified versions of the standard H100 GPUs designed to comply with the US export restrictions. However, Chinese companies like Moore Threads are actively developing GPUs designed specifically for DeepSeek’s AI models.
Additionally, Chinese tech giant Baidu has developed an in-house AI cluster powered by its Kunlun Core P800 AI chips. The Core P800 outperforms comparable mainstream GPUs by 20–50%, supports 8-bit inference, and offers significantly lower deployment and maintenance costs. The chip is reportedly fully compatible with DeepSeek’s V3/R1 AI models, enabling seamless inference deployment.
This marks a significant step toward China’s goal of AI self-sufficiency, particularly in response to US sanctions limiting access to high-end Nvidia GPUs. By advancing its own GPU capabilities, China is reducing its reliance on foreign hardware to support powerful AI models like DeepSeek.
The US has adopted a defensive AI strategy, focusing on regulating AI development, restricting access to computing power, and keeping its models closed-source. While this approach aims to safeguard its technology and prevent replication, it is proving counterproductive.
The Framework for AI Diffusion, introduced by the US is a well-intentioned policy designed to maintain AI leadership by controlling the outflow of technology. However, in an era of rapid innovation, it has already become outdated and fails to account for China's adaptive strategies.
China has shifted its focus to aggressively advancing more efficient and cost-effective AI technologies, recognising its limitation in developing powerful computing chips. It is also working to establish dominance in open-source AI, cloud infrastructure, and global data ecosystems.
These efforts allow China to offer cheaper, unrestricted AI access to countries frustrated by US policies, embedding itself in emerging markets in ways that will be difficult to dislodge.