(Kitco News) – The announcement of Chinese upstart AI platform DeepSeek at the start of the week sent shockwaves through the artificial intelligence industry – and drove the market value of other AI-based tech equities sharply lower. Wednesday’s announcement that Chinese retail giant Alibaba has their own AI platform which outperforms even DeepSeek underscores how quickly and fundamentally this disruptive industry is itself being disrupted. But, AI experts believe the technological and financial implications of open-source AI at a fraction of the price are only just beginning to be grasped.
Dariia Porechna is the Head of Protocol at Autonomys, a network that provides the infrastructure to scale decentralized AI applications on-chain. She told Kitco News that DeepSeek has already disrupted the AI ecosystem at a fundamental level, by undercutting the proprietary and closed models of the dominant players.
“The rise of DeepSeek has effectively challenged the status quo narrative built by the top U.S. companies that open-source AI will always be inferior to proprietary models developed in big data centers spending billions of dollars on massive training runs,” Porechna said. “DeepSeek has shown that open-source tech has not only caught up with the reigning OpenAI and Anthropic models but has also done so in a much more cost-effective way.”
Porechna believes the next big disruption in the space will come from decentralized AI. “DeepSeek, based on open-source research, has, in a way, already benefitted from decentralized science, but they still had to rely only on hardware available domestically,” she said. “When we can tap into idle computational capacity worldwide and coordinate training of frontier models globally, we may uncover a technology with unprecedented capabilities, accessible to anyone.”
As is often the case in the tech space, Porechna said the key will be watching to see if DeepSeek can successfully transition from a white-hot startup to a bread-and-butter service provider.
“The DeepSeek lab is a relatively new player in the space, so we must see if they can keep up with the success,” she said. “At the same time, an ‘unknown’ team being able to produce such high-quality models and openly publish their research particularly lowers the barrier to frontier AI capabilities. It exposes the budget moat of Big Tech as much less significant.”
“By demonstrating that high performance can be achieved with significantly fewer resources, the DeepSeek precedent encourages the AI community to explore more innovative and sustainable approaches to development,” she added. “Many in the wider industry will have to reconsider their priorities in spending and research.”
With so much of the U.S. equity market’s all-time high valuations based on the promise of AI to not only create enormous value for American businesses but also to successfully capture and monetize that value, the implications of a much cheaper model outside the control of Big Tech is profound.
“The implications are undoubtedly significant for the end users and the developer,” Porechna said. “The DeepSeek success story as a company could lead to a more democratized AI landscape, where smaller entities and individuals have greater access to advanced AI technologies. The end users realize they can have access to a compelling technology practically for free and unlock enormous productivity gains.”
“AI researchers who have so far been left out of the competition due to the prohibitive costs of training cutting-edge models are now more than ever encouraged to develop the models they have only theoretically thought about,” she added. “We may see many more specialized models seeing light that were deemed economically unreasonable before.”
Ben James is the Founder of 404, Bittensor Subnet 17, which enables users to build virtual worlds, games, and AR/VR/XR experiences. He is also the CEO and Founder of Atlas, a 3D generative AI platform for gaming and virtual world-building.
James believes that even if DeepSeek’s value ends up being integrated into the existing AI landscape, its proof-of-concept for speed of development and cost, and its accessibility to and example for smaller development teams will still have a massive impact.
“DeepSeek disrupts commonly accepted ideas about the costs of foundational model training,” James told Kitco News. “It brings national concerns into the AI race given that this model is from China… will OpenAI ask to ban it for national security? Will the US government fund Meta’s open-source AI lab?”
That said, James doesn’t believe DeepSeek will necessarily disrupt the AI space, but rather the technology could serve to magnify the disruptive impacts that are already underway.
“Decentralized AI models tend to work best in combination with open source since the community of builders can be larger, so in a way this can amplify disruption,” he said. “The market is potentially overreacting (e.g. NVIDIA GPUs will still be used by the application and agent layers for inference), but it is significant especially for crypto and AI at the application/agent layer where DeepSeek facilitates innovation.”
James believes the open-source nature of DeepSeek will unlock significant app and agent innovations. “The builder cost suggests that new foundational models can be built by significantly smaller enterprises,” he added.
