The rise of DeepSeek, a tech startup in China, in the AI industry marks a significant shift in how high-performance models are developed, trained, and deployed.
The latest Omdia report said DeepSeek’s AI assistant, which topped app store charts in the US and Europe, demonstrated not only market demand but also the efficiency and effectiveness of DeepSeek’s underlying models.
DeepSeek-V3
The DeepSeek-V3 model, with 671 billion parameters, has proven capable of competing with and, in some cases, outperforming leading US-based open and closed-source AI models. This achievement is particularly noteworthy given its training efficiency — using 2,048 NVIDIA H800 GPUs at a fraction of the cost typically required for models of similar scale.
The release of DeepSeek-R1, a reasoning-focused model integrating advancements from multiple open-source models, further solidifies the company’s ability to deliver cost-effective and high-performing AI systems.
DeepSeek’s success challenges the long-standing belief that only companies with access to the most advanced hardware and massive financial resources can lead AI innovation. By optimizing its training processes and leveraging open-source foundations, DeepSeek has demonstrated that AI development can be more accessible and affordable.
This shift is likely to influence broader AI research and industry dynamics, encouraging other companies to explore similar cost-efficient approaches. The rapid emergence of DeepSeek has also sparked intense discussion in the AI community regarding originality, innovation, and the evolving role of open-source collaboration.
While some debate the uniqueness of DeepSeek’s technological contributions, its ability to refine, optimize, and scale existing methodologies is undeniable. The company’s approach exemplifies a new wave of AI development that prioritizes agility, creativity, and strategic use of available resources.
As enterprises and AI researchers assess DeepSeek’s impact, the industry may witness a reevaluation of what is necessary to build powerful AI systems. The company’s growth suggests that the future of AI is not solely dependent on proprietary models from dominant players but may increasingly involve a diverse range of contributors leveraging open-source advancements in innovative ways. DeepSeek’s rise signals a transformation in AI economics, competition, and accessibility, making it a crucial player in the ongoing evolution of artificial intelligence.
DeepSeek-V3 is a game changer because it highlights and accelerates a shift in the AI industry — from massive, proprietary models to smaller, more efficient, open-source alternatives. While this transformation has been unfolding since the leaked Llama weights in 2023, DeepSeek-V3 and R1 have dramatized its impact by delivering high performance at a fraction of the cost traditionally associated with large-scale AI training. The success of these models challenges the dominance of proprietary AI giants by proving that capabilities can be achieved with strategic optimization rather than brute-force computing power.
The rise of DeepSeek-V3 aligns with industry trends, where advancements in quantization, model merging, and distillation have made it possible to create smaller, highly capable models. While DeepSeek’s achievements are significant, it is not alone in this movement — companies like Meta, Google, and Microsoft have also contributed to the development of more compute-efficient AI.
However, DeepSeek’s ability to rapidly scale and deploy such models in real-world applications has set a new benchmark, demonstrating that cost-effective AI solutions can rival the performance of traditionally dominant players. This shift is redefining AI accessibility, making advanced capabilities available to a wider range of developers and enterprises, and reshaping the competitive landscape of the AI industry.
Data privacy concerns
Data privacy concerns surrounding DeepSeek mirror anxieties about AI services and data security, particularly given geopolitical tensions and past controversies involving Chinese tech companies like TikTok. While DeepSeek’s AI services raise questions about data access and potential exploitation, these concerns are not unique — cloud-based AI services from OpenAI, Google, Microsoft, and Apple also collect user data depending on their licensing agreements. The key issue is not whether AI models collect data, but who has access to it and how it is used, whether for service improvement, advertising, or governmental oversight.
One major advantage of DeepSeek-R1 is its availability under an open-source MIT license, offering enterprises greater control over data privacy. Unlike proprietary AI services, which can involve undisclosed data collection, open-source models allow companies to self-host, modify, and audit the software to ensure data security. This flexibility enables organizations to deploy AI solutions without relying on external providers, reducing the risk of data leaks or unauthorized data harvesting.
As AI adoption grows, enterprises must balance usability with privacy by implementing secure, internally controlled solutions to mitigate the risks associated with commercial AI services. The broader shift toward open-source AI underscores the importance of transparency and control in an era of increasing data privacy concerns.
AI supremacy
DeepSeek’s achievements have not upended the race for AI supremacy, despite claims to the contrary. While its ability to produce high-performance models using lower-spec GPUs is impressive — especially in light of US export bans on high-end chips to China — it has not given China a definitive lead in AI. The AI race remains dynamic, with innovations rapidly integrated and iterated upon by competitors. Open-source AI, in particular, serves as a great equalizer, ensuring that no single breakthrough remains an exclusive advantage for long.
Rather than ending the AI race, DeepSeek’s success highlights the evolving nature of AI innovation. It challenges the assumption that brute-force computing power is the primary driver of progress, emphasizing instead the importance of agility, creativity, and strategic resource use. Nations aiming for AI leadership should recognize that infrastructure alone is not enough — fostering innovation, talent, and adaptability is equally critical.
DeepSeek’s rise should prompt a reassessment of what AI supremacy truly means. If AI development focuses solely on competition, it risks prioritizing speed over safety, ethics, and responsible deployment. No AI company, regardless of its current dominance, can assume long-term superiority. The lesson from DeepSeek is clear: the future of AI belongs not just to those with the most computing power, but to those who can innovate efficiently and responsibly, Omdia analysts said.
Eden Zoller (Chief Analyst, Applied AI), Bradley Shimmin (Chief Analyst, AI Platforms), Lian Jye Su (Chief Analyst, Applied Intelligence) and Alexander Harrowell (Principal Analyst, Advanced Computing for AI) at Omdia are the authors of the report.
Rajani Baburajan
