The Decentralization of AI: DeepSeek, Qwen, and Cold War 2.0

The Decentralization of AI: DeepSeek, Qwen, and Cold War 2.0

These past three weeks, the artificial intelligence (AI) sector has experienced significant transformations with the emergence of two groundbreaking models: DeepSeek and Qwen. These innovations have redefined the AI landscape globally, particularly in terms of cost-effectiveness, performance, and decentralization.

Democratizing AI with Cost-Effective Models

Traditionally, developing advanced AI models required substantial financial investments, often reaching billions of dollars. DeepSeek, a Chinese AI startup, has challenged this norm by creating sophisticated AI models at a fraction of the cost. 

For instance, their R1 reasoning model was trained for approximately $6 million, significantly less than the $100 million spent on OpenAI's GPT-4 and is priced at $2.19 per million output tokens, compared to OpenAI's comparable model priced at $60.00—a cost reduction to 1/30th of similar OpenAI models.

This cost efficiency has made AI technology more accessible to a broader range of organizations and individuals, fostering innovation and competition.

Delivering Competitive Performance

Despite their cost-effective development, both DeepSeek and Qwen have demonstrated exceptional performance. 

DeepSeek's models, such as the R1 reasoning model, have outperformed those from major companies like OpenAI, Meta, and Google in various benchmarks:

Note: Training times and costs for OpenAI's models are proprietary and estimated based on industry standards.

The open-source nature of both DeepSeek and Qwen has been instrumental in promoting decentralization in AI development. By making their models publicly accessible, they have empowered a global community of developers and researchers to contribute to and benefit from AI advancements. This approach contrasts with the traditional model of proprietary AI development, where access is often restricted to a select few organizations.

While the advancements brought by DeepSeek and Qwen are commendable, they also introduce new geopolitical considerations.

Decentralization or Cold War 2.0?

In 2025, Russia’s Sberbank launched a notable joint research initiative with China, aiming to develop advanced AI models that could rival Western counterparts like OpenAI’s GPT and Anthropic’s Claude. This collaboration seeks to leverage the strengths and expertise of both nations, advancing AI technologies with applications across diverse sectors such as finance and security.

While this partnership accelerates the pace of AI development, it also sparks discussions about the ethical and strategic implications of these advancements. As AI models like DeepSeek and Qwen rapidly gain global traction, they are becoming influential players in what some are calling a “Cold War 2.0” between global superpowers, particularly the United States and China, with artificial intelligence at its core.

Concerns have emerged internationally about their potential application in surveillance and state-driven narratives. A recent AP report noted that the U.S. government has restricted the use of DeepSeek models on federal devices, citing potential risks tied to data security and influence operations.

As these developments unfold, they reveal the pivotal role AI plays in shaping the future of global governance and technological cooperation. Far from being solely a competitive endeavor, advancements like DeepSeek and Qwen also open opportunities for collaboration and dialogue between nations, as they navigate the complexities of integrating AI into society responsibly. 

This technological evolution offers the potential to bridge gaps, foster innovation, and address shared challenges on the global stage.